<?xml version="1.0"?> <net name="Model0" version="11"> <layers> <layer id="3" name="input_ids" type="Parameter" version="opset1"> <data shape="?,?" element_type="i64" /> <output> <port id="0" precision="I64" names="input_ids"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2" name="attention_mask" type="Parameter" version="opset1"> <data shape="?,?" element_type="i64" /> <output> <port id="0" precision="I64" names="attention_mask"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1" name="position_ids" type="Parameter" version="opset1"> <data shape="?,?" element_type="i64" /> <output> <port id="0" precision="I64" names="position_ids"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="0" name="beam_idx" type="Parameter" version="opset1"> <data shape="?" element_type="i32" /> <output> <port id="0" precision="I32" names="beam_idx"> <dim>-1</dim> </port> </output> </layer> <layer id="4" name="Constant_534991" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5" name="ShapeOf_534977" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>2</dim> </port> </output> </layer> <layer id="6" name="Constant_534979" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7" name="Constant_534981" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8" name="Gather_534982" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>2</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9" name="Constant_534984" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="10" name="Constant_534986" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="11" name="Constant_534988" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="12" name="Concat_534989" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="13" name="Broadcast_534992" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="14" name="ReadValue_522477" type="ReadValue" version="opset6"> <data variable_id="past_key_values.9.valuepresent.9.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.9.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="15" name="Constant_509599" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="16" name="Gather_509600" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="17" name="Constant_201" type="Const" version="opset1"> <data element_type="u8" shape="50432, 6144" offset="28" size="309854208" /> <output> <port id="0" precision="U8"> <dim>50432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="18" name="Convert_548891" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>50432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="19" name="Constant_201/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="50432, 1" offset="309854236" size="50432" /> <output> <port id="0" precision="U8"> <dim>50432</dim> <dim>1</dim> </port> </output> </layer> <layer id="20" name="Convert_548894" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>50432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>50432</dim> <dim>1</dim> </port> </output> </layer> <layer id="21" name="Subtract_548895" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>50432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="22" name="Constant_548896" type="Const" version="opset1"> <data element_type="f32" shape="50432, 1" offset="309904668" size="201728" /> <output> <port id="0" precision="FP32"> <dim>50432</dim> <dim>1</dim> </port> </output> </layer> <layer id="23" name="Constant_201/fq_weights_0" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>50432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="24" name="__module.model.gpt_neox.embed_in/aten::embedding/Convert" type="Convert" version="opset1"> <data destination_type="i32" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="25" name="__module.model.gpt_neox.embed_in/aten::embedding/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="26" name="__module.model.gpt_neox.embed_in/aten::embedding/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="406,input.1"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="27" name="__module.model.gpt_neox.layers.0.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="28" name="__module.model.gpt_neox.layers.0.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="29" name="Constant_466783" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="310106400" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="30" name="__module.model.gpt_neox.layers.0.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="31" name="Constant_466784" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="310130976" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="32" name="__module.model.gpt_neox.layers.0.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="540,input.11"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="33" name="Constant_750" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="310155552" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="34" name="Convert_548918" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="35" name="Constant_750/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="461150496" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="36" name="Convert_548921" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="37" name="Subtract_548922" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="38" name="Constant_548923" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="461175072" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="39" name="Constant_750/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="40" name="__module.model.gpt_neox.layers.0.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="41" name="Constant_466785" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="461273376" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="42" name="__module.model.gpt_neox.layers.0.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="545,input.13"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="43" name="__module.model.gpt_neox.layers.0.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="554,input.15"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="44" name="Constant_770" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="461371680" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="45" name="Convert_548927" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="46" name="Constant_770/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="612366624" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="47" name="Convert_548930" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="48" name="Subtract_548931" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="49" name="Constant_548932" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="612372768" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="50" name="Constant_770/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="51" name="__module.model.gpt_neox.layers.0.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="52" name="Constant_466786" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="612397344" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="53" name="__module.model.gpt_neox.layers.0.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="557,input.17"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="54" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="55" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="56" name="Constant_466787" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="612421920" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="57" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="58" name="Constant_466788" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="612446496" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="59" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="415,input.5"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="60" name="Constant_229" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="612471072" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="61" name="Convert_548900" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="62" name="Constant_229/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="725717280" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="63" name="Convert_548903" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="64" name="Subtract_548904" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="65" name="Constant_548905" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="725735712" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="66" name="Constant_229/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="67" name="__module.model.gpt_neox.layers.0.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="68" name="Constant_466789" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="725809440" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="69" name="__module.model.gpt_neox.layers.0.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="422,qkv.1"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="70" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="71" name="__module.model.gpt_neox.layers.0.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="426,qkv.3"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="72" name="Constant_423264" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="73" name="Constant_423267" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="74" name="Constant_423270" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="75" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="427"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="76" name="Constant_257" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="428"> <dim>4</dim> </port> </output> </layer> <layer id="77" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="429,query.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="78" name="Constant_423276" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="79" name="Constant_423279" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="80" name="Constant_423282" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="81" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="436,q.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="82" name="Constant_333" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="448,cos_cached.1"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="83" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="84" name="Constant_423288" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="85" name="Constant_423291" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="86" name="Constant_423294" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="87" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="430"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="88" name="Constant_271" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="431"> <dim>4</dim> </port> </output> </layer> <layer id="89" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="432,key.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="90" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="91" name="278" type="Const" version="opset1"> <data element_type="i32" shape="" offset="726079888" size="4" /> <output> <port id="0" precision="I32" names="278" /> </output> </layer> <layer id="92" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="93" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="440" /> </output> </layer> <layer id="94" name="Constant_535004" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="95" name="Constant_534997" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="96" name="Constant_534999" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="97" name="Constant_535001" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="98" name="Concat_535002" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="99" name="Broadcast_535005" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="100" name="ReadValue_522303" type="ReadValue" version="opset6"> <data variable_id="past_key_values.0.keypresent.0.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.0.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="101" name="Constant_509542" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="102" name="Gather_509543" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="103" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="104" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="105" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="442" /> </output> </layer> <layer id="106" name="__module.model.gpt_neox.layers.0.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="441,445,446,seq_len.1" /> </output> </layer> <layer id="107" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="108" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="109" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="110" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="449,452,cos.1"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="111" name="__module.model.gpt_neox.layers.0.attention/aten::index/Convert" type="Convert" version="opset1"> <data destination_type="i32" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="112" name="Constant_51903" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="113" name="__module.model.gpt_neox.layers.0.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="455"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="114" name="277" type="Const" version="opset1"> <data element_type="i32" shape="" offset="726079892" size="4" /> <output> <port id="0" precision="I32" names="277" /> </output> </layer> <layer id="115" name="__module.model.gpt_neox.layers.0.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="456,cos.3"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="116" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="460"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="117" name="Constant_423299" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="118" name="Constant_423298" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="119" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="120" name="279" type="Const" version="opset1"> <data element_type="i32" shape="" offset="726079904" size="4" /> <output> <port id="0" precision="I32" names="279" /> </output> </layer> <layer id="121" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="122" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="461,466" /> </output> </layer> <layer id="123" name="295" type="Const" version="opset1"> <data element_type="i32" shape="" offset="726079888" size="4" /> <output> <port id="0" precision="I32" names="295" /> </output> </layer> <layer id="124" name="__module.model.gpt_neox.layers.0.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="125" name="__module.model.gpt_neox.layers.0.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="463,464,468,469" /> </output> </layer> <layer id="126" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="127" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="128" name="Constant_423297" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="129" name="ScatterUpdate_423300" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="130" name="Constant_423303" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="131" name="Constant_423306" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="132" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="470,x2.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="133" name="Constant_466790" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="134" name="__module.model.gpt_neox.layers.0.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="471"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="135" name="Constant_423352" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="136" name="Constant_423357" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="137" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="465,x1.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="138" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="473"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="139" name="Constant_331" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="447,sin_cached.1"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="140" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="141" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="142" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="450,453,sin.1"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="143" name="Constant_51906" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="144" name="__module.model.gpt_neox.layers.0.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="458"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="145" name="__module.model.gpt_neox.layers.0.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="459,sin.3"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="146" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="474"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="147" name="__module.model.gpt_neox.layers.0.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="475,query.3"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="148" name="Constant_423404" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="149" name="Constant_423407" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="150" name="Constant_423410" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="151" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="437,query_pass.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="152" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="493,503,query.5,query.7"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="153" name="Constant_423416" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="154" name="Constant_423419" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="155" name="Constant_423422" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="156" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="438,k.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="157" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="476"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="158" name="Constant_423426" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="159" name="Constant_423425" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="160" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="161" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="162" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="477,482" /> </output> </layer> <layer id="163" name="__module.model.gpt_neox.layers.0.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="164" name="__module.model.gpt_neox.layers.0.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="479,480,484,485" /> </output> </layer> <layer id="165" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="166" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="167" name="Constant_423424" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="168" name="ScatterUpdate_423427" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="169" name="Constant_423430" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="170" name="Constant_423433" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="171" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="486,x2.3"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="172" name="Constant_466791" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="173" name="__module.model.gpt_neox.layers.0.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="487"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="174" name="Constant_423479" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="175" name="Constant_423484" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="176" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="481,x1.3"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="177" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="489"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="178" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="490"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="179" name="__module.model.gpt_neox.layers.0.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="491,key.3"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="180" name="Constant_423530" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="181" name="Constant_423533" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="182" name="Constant_423536" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="183" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="439,key_pass.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="184" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="495,key.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="185" name="__module.model.gpt_neox.layers.0.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.0.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="186" name="Constant_535017" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="187" name="Constant_535010" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="188" name="Constant_535012" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="189" name="Constant_535014" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="190" name="Concat_535015" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="191" name="Broadcast_535018" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="192" name="ReadValue_522305" type="ReadValue" version="opset6"> <data variable_id="past_key_values.0.valuepresent.0.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.0.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="193" name="Constant_509545" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="194" name="Gather_509546" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="195" name="Constant_423542" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="196" name="Constant_423545" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="197" name="Constant_423548" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="198" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="433"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="199" name="Constant_285" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="434"> <dim>4</dim> </port> </output> </layer> <layer id="200" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="435,value.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="201" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.0.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="202" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="512"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="203" name="Constant_423552" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="204" name="Constant_423551" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="205" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="206" name="296" type="Const" version="opset1"> <data element_type="i32" shape="" offset="730470876" size="4" /> <output> <port id="0" precision="I32" names="296" /> </output> </layer> <layer id="207" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="208" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="507" /> </output> </layer> <layer id="209" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="210" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="211" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="505" /> </output> </layer> <layer id="212" name="__module.model.gpt_neox.layers.0.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="509,510" /> </output> </layer> <layer id="213" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="214" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="215" name="Constant_423550" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="216" name="ScatterUpdate_423553" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="217" name="Constant_423554" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="218" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="219" name="ScatterUpdate_423555" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="220" name="Constant_423558" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="221" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="513"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="222" name="Constant_423630" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="223" name="Constant_423631" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="224" name="Constant_423627" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="225" name="Constant_423626" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="226" name="ScatterUpdate_423632" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="227" name="Constant_423635" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="228" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="514,515,causal_mask.1"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="229" name="__module.model.gpt_neox.layers.0.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="230" name="__module.model.gpt_neox.layers.0.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="502,mask_value.1" /> </output> </layer> <layer id="231" name="__module.model.gpt_neox.layers.0.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="516,causal_mask.3"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="232" name="Constant_466793" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="730470896" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="233" name="__module.model.gpt_neox/aten::size/ShapeOf" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>2</dim> </port> </output> </layer> <layer id="234" name="Constant_402275" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="235" name="__module.model.gpt_neox/aten::size/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="236" name="__module.model.gpt_neox/aten::size/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>2</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="395"> <dim>1</dim> </port> </output> </layer> <layer id="237" name="__module.model.gpt_neox/prim::ListConstruct/Reshape_0" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="238" name="__module.model.gpt_neox/prim::ListConstruct/Concat" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="I32" names="396"> <dim>2</dim> </port> </output> </layer> <layer id="239" name="__module.model.gpt_neox/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>2</dim> </port> </input> <output> <port id="2" precision="I64" names="397,attention_mask.3"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="240" name="__module.model.gpt_neox/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="241" name="__module.model.gpt_neox/aten::slice/Unsqueeze_1" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470900" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="242" name="__module.model.gpt_neox/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="243" name="__module.model.gpt_neox/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64" names="398"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="244" name="__module.model.gpt_neox/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I64" names="399"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="245" name="__module.model.gpt_neox/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I64" names="400"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="246" name="Constant_423644" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="247" name="Constant_423647" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="248" name="Constant_423650" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="249" name="__module.model.gpt_neox/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="I64" names="401,attention_mask.5"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="250" name="__module.model.gpt_neox/aten::to/Convert" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="FP32" names="402"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="251" name="Constant_466792" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="730470896" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="252" name="__module.model.gpt_neox/aten::rsub/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="253" name="__module.model.gpt_neox/aten::rsub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32" names="403"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="254" name="Constant_466794" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="730470892" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="255" name="__module.model.gpt_neox/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="404,attention_mask.7"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="256" name="Add_453239" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="257" name="ShapeOf_453240" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="258" name="Constant_402272" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="259" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="260" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="500"> <dim>1</dim> </port> </output> </layer> <layer id="261" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="262" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="263" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="264" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="517"> <dim>4</dim> </port> </output> </layer> <layer id="265" name="__module.model.gpt_neox.layers.0.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="266" name="__module.model.gpt_neox.layers.0.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="267" name="__module.model.gpt_neox.layers.0.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="268" name="__module.model.gpt_neox.layers.0.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="269" name="Maximum_453241" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="270" name="__module.model.gpt_neox.layers.0.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="519,attention_mask.9"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="271" name="__module.model.gpt_neox.layers.0.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="520,521,sdpa_result.1,tensor.1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="272" name="Constant_710" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="522"> <dim>4</dim> </port> </output> </layer> <layer id="273" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="523"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="274" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="275" name="__module.model.gpt_neox.layers.0.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="528,input.7"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="276" name="Constant_728" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="730470936" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="277" name="Convert_548909" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="278" name="Constant_728/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="768219672" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="279" name="Convert_548912" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="280" name="Subtract_548913" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="281" name="Constant_548914" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="768225816" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="282" name="Constant_728/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="283" name="__module.model.gpt_neox.layers.0.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="284" name="Constant_466795" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="612397344" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="285" name="__module.model.gpt_neox.layers.0.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="531,533,input.9"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="286" name="__module.model.gpt_neox.layers.0/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="559"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="287" name="__module.model.gpt_neox.layers.0/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="560,562,input.19"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="288" name="__module.model.gpt_neox.layers.1.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="289" name="__module.model.gpt_neox.layers.1.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="290" name="Constant_466796" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="768250392" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="291" name="__module.model.gpt_neox.layers.1.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="292" name="Constant_466797" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="768274968" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="293" name="__module.model.gpt_neox.layers.1.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="697,input.27"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="294" name="Constant_1328" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="768299544" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="295" name="Convert_548954" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="296" name="Constant_1328/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="919294488" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="297" name="Convert_548957" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="298" name="Subtract_548958" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="299" name="Constant_548959" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="919319064" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="300" name="Constant_1328/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="301" name="__module.model.gpt_neox.layers.1.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="302" name="Constant_466798" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="919417368" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="303" name="__module.model.gpt_neox.layers.1.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="702,input.29"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="304" name="__module.model.gpt_neox.layers.1.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="711,input.31"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="305" name="Constant_1348" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="919515672" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="306" name="Convert_548963" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="307" name="Constant_1348/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="1070510616" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="308" name="Convert_548966" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="309" name="Subtract_548967" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="310" name="Constant_548968" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="1070516760" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="311" name="Constant_1348/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="312" name="__module.model.gpt_neox.layers.1.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="313" name="Constant_466799" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1070541336" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="314" name="__module.model.gpt_neox.layers.1.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="714,input.33"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="315" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="316" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="317" name="Constant_466800" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1070565912" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="318" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="319" name="Constant_466801" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1070590488" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="320" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="572,input.21"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="321" name="Constant_807" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="1070615064" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="322" name="Convert_548936" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="323" name="Constant_807/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="1183861272" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="324" name="Convert_548939" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="325" name="Subtract_548940" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="326" name="Constant_548941" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="1183879704" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="327" name="Constant_807/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="328" name="__module.model.gpt_neox.layers.1.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="329" name="Constant_466802" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="1183953432" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="330" name="__module.model.gpt_neox.layers.1.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="579,qkv.5"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="331" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="332" name="__module.model.gpt_neox.layers.1.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="583,qkv.7"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="333" name="Constant_423656" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="334" name="Constant_423659" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="335" name="Constant_423662" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="336" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="584"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="337" name="Constant_835" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="585"> <dim>4</dim> </port> </output> </layer> <layer id="338" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="586,query.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="339" name="Constant_423668" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="340" name="Constant_423671" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="341" name="Constant_423674" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="342" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="593,q.3"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="343" name="Constant_911" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="605,cos_cached.3"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="344" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="345" name="Constant_423680" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="346" name="Constant_423683" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="347" name="Constant_423686" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="348" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="587"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="349" name="Constant_849" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="588"> <dim>4</dim> </port> </output> </layer> <layer id="350" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="589,key.11"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="351" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="352" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="353" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="597" /> </output> </layer> <layer id="354" name="Constant_535030" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="355" name="Constant_535023" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="356" name="Constant_535025" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="357" name="Constant_535027" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="358" name="Concat_535028" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="359" name="Broadcast_535031" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="360" name="ReadValue_522307" type="ReadValue" version="opset6"> <data variable_id="past_key_values.1.keypresent.1.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.1.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="361" name="Constant_509548" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="362" name="Gather_509549" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="363" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="364" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="365" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="599" /> </output> </layer> <layer id="366" name="__module.model.gpt_neox.layers.1.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="598,602,603,seq_len.5" /> </output> </layer> <layer id="367" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="368" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="369" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="370" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="606,609,cos.5"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="371" name="Constant_51909" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="372" name="__module.model.gpt_neox.layers.1.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="612"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="373" name="__module.model.gpt_neox.layers.1.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="613,cos.7"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="374" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="617"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="375" name="Constant_423691" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="376" name="Constant_423690" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="377" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="378" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="379" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="618,623" /> </output> </layer> <layer id="380" name="__module.model.gpt_neox.layers.1.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="381" name="__module.model.gpt_neox.layers.1.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="620,621,625,626" /> </output> </layer> <layer id="382" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="383" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="384" name="Constant_423689" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="385" name="ScatterUpdate_423692" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="386" name="Constant_423695" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="387" name="Constant_423698" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="388" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="627,x2.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="389" name="Constant_466803" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="390" name="__module.model.gpt_neox.layers.1.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="628"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="391" name="Constant_423744" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="392" name="Constant_423749" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="393" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="622,x1.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="394" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="630"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="395" name="Constant_909" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="604,sin_cached.3"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="396" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="397" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="398" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="607,610,sin.5"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="399" name="Constant_51912" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="400" name="__module.model.gpt_neox.layers.1.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="615"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="401" name="__module.model.gpt_neox.layers.1.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="616,sin.7"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="402" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="631"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="403" name="__module.model.gpt_neox.layers.1.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="632,query.11"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="404" name="Constant_423796" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="405" name="Constant_423799" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="406" name="Constant_423802" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="407" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="594,query_pass.3"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="408" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="650,660,query.13,query.15"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="409" name="Constant_423808" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="410" name="Constant_423811" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="411" name="Constant_423814" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="412" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="595,k.3"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="413" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="633"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="414" name="Constant_423818" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="415" name="Constant_423817" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="416" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="417" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="418" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="634,639" /> </output> </layer> <layer id="419" name="__module.model.gpt_neox.layers.1.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="420" name="__module.model.gpt_neox.layers.1.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="636,637,641,642" /> </output> </layer> <layer id="421" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="422" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="423" name="Constant_423816" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="424" name="ScatterUpdate_423819" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="425" name="Constant_423822" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="426" name="Constant_423825" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="427" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="643,x2.7"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="428" name="Constant_466804" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="429" name="__module.model.gpt_neox.layers.1.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="644"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="430" name="Constant_423871" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="431" name="Constant_423876" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="432" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="638,x1.7"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="433" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="646"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="434" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="647"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="435" name="__module.model.gpt_neox.layers.1.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="648,key.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="436" name="Constant_423922" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="437" name="Constant_423925" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="438" name="Constant_423928" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="439" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="596,key_pass.3"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="440" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="652,key.15"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="441" name="__module.model.gpt_neox.layers.1.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.1.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="442" name="Constant_535043" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="443" name="Constant_535036" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="444" name="Constant_535038" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="445" name="Constant_535040" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="446" name="Concat_535041" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="447" name="Broadcast_535044" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="448" name="ReadValue_522309" type="ReadValue" version="opset6"> <data variable_id="past_key_values.1.valuepresent.1.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.1.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="449" name="Constant_509551" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="450" name="Gather_509552" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="451" name="Constant_423934" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="452" name="Constant_423937" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="453" name="Constant_423940" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="454" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="590"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="455" name="Constant_863" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="591"> <dim>4</dim> </port> </output> </layer> <layer id="456" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="592,value.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="457" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.1.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="458" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="669"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="459" name="Constant_423944" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="460" name="Constant_423943" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="461" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="462" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="463" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="664" /> </output> </layer> <layer id="464" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="465" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="466" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="662" /> </output> </layer> <layer id="467" name="__module.model.gpt_neox.layers.1.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="666,667" /> </output> </layer> <layer id="468" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="469" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="470" name="Constant_423942" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="471" name="ScatterUpdate_423945" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="472" name="Constant_423946" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="473" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="474" name="ScatterUpdate_423947" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="475" name="Constant_423950" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="476" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="670"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="477" name="Constant_424022" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="478" name="Constant_424023" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="479" name="Constant_424019" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="480" name="Constant_424018" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="481" name="ScatterUpdate_424024" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="482" name="Constant_424027" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="483" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="671,672,causal_mask.7"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="484" name="__module.model.gpt_neox.layers.1.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="485" name="__module.model.gpt_neox.layers.1.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="659,mask_value.3" /> </output> </layer> <layer id="486" name="__module.model.gpt_neox.layers.1.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="673,causal_mask.9"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="487" name="Add_453315" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="488" name="ShapeOf_453316" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="489" name="Constant_402290" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="490" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="491" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="657"> <dim>1</dim> </port> </output> </layer> <layer id="492" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="493" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="494" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="495" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="674"> <dim>4</dim> </port> </output> </layer> <layer id="496" name="__module.model.gpt_neox.layers.1.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="497" name="__module.model.gpt_neox.layers.1.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="498" name="__module.model.gpt_neox.layers.1.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="499" name="__module.model.gpt_neox.layers.1.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="500" name="Maximum_453317" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="501" name="__module.model.gpt_neox.layers.1.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="676,attention_mask.11"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="502" name="__module.model.gpt_neox.layers.1.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="677,678,sdpa_result.3,tensor.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="503" name="Constant_1288" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="679"> <dim>4</dim> </port> </output> </layer> <layer id="504" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="680"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="505" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="506" name="__module.model.gpt_neox.layers.1.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="685,input.23"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="507" name="Constant_1306" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="1184027160" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="508" name="Convert_548945" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="509" name="Constant_1306/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="1221775896" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="510" name="Convert_548948" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="511" name="Subtract_548949" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="512" name="Constant_548950" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="1221782040" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="513" name="Constant_1306/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="514" name="__module.model.gpt_neox.layers.1.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="515" name="Constant_466805" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1221806616" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="516" name="__module.model.gpt_neox.layers.1.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="688,690,input.25"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="517" name="__module.model.gpt_neox.layers.1/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="716"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="518" name="__module.model.gpt_neox.layers.1/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="717,719,input.35"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="519" name="__module.model.gpt_neox.layers.2.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="520" name="__module.model.gpt_neox.layers.2.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="521" name="Constant_466806" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1221831192" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="522" name="__module.model.gpt_neox.layers.2.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="523" name="Constant_466807" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1221855768" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="524" name="__module.model.gpt_neox.layers.2.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="854,input.43"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="525" name="Constant_1906" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="1221880344" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="526" name="Convert_548990" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="527" name="Constant_1906/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="1372875288" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="528" name="Convert_548993" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="529" name="Subtract_548994" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="530" name="Constant_548995" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="1372899864" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="531" name="Constant_1906/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="532" name="__module.model.gpt_neox.layers.2.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="533" name="Constant_466808" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="1372998168" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="534" name="__module.model.gpt_neox.layers.2.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="859,input.45"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="535" name="__module.model.gpt_neox.layers.2.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="868,input.47"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="536" name="Constant_1926" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="1373096472" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="537" name="Convert_548999" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="538" name="Constant_1926/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="1524091416" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="539" name="Convert_549002" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="540" name="Subtract_549003" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="541" name="Constant_549004" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="1524097560" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="542" name="Constant_1926/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="543" name="__module.model.gpt_neox.layers.2.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="544" name="Constant_466809" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1524122136" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="545" name="__module.model.gpt_neox.layers.2.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="871,input.49"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="546" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="547" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="548" name="Constant_466810" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1524146712" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="549" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="550" name="Constant_466811" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1524171288" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="551" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="729,input.37"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="552" name="Constant_1385" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="1524195864" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="553" name="Convert_548972" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="554" name="Constant_1385/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="1637442072" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="555" name="Convert_548975" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="556" name="Subtract_548976" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="557" name="Constant_548977" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="1637460504" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="558" name="Constant_1385/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="559" name="__module.model.gpt_neox.layers.2.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="560" name="Constant_466812" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="1637534232" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="561" name="__module.model.gpt_neox.layers.2.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="736,qkv.9"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="562" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="563" name="__module.model.gpt_neox.layers.2.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="740,qkv.11"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="564" name="Constant_424035" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="565" name="Constant_424038" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="566" name="Constant_424041" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="567" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="741"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="568" name="Constant_1413" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="742"> <dim>4</dim> </port> </output> </layer> <layer id="569" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="743,query.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="570" name="Constant_424047" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="571" name="Constant_424050" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="572" name="Constant_424053" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="573" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="750,q.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="574" name="Constant_1489" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="762,cos_cached.5"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="575" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="576" name="Constant_424059" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="577" name="Constant_424062" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="578" name="Constant_424065" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="579" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="744"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="580" name="Constant_1427" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="745"> <dim>4</dim> </port> </output> </layer> <layer id="581" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="746,key.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="582" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="583" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="584" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="754" /> </output> </layer> <layer id="585" name="Constant_535056" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="586" name="Constant_535049" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="587" name="Constant_535051" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="588" name="Constant_535053" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="589" name="Concat_535054" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="590" name="Broadcast_535057" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="591" name="ReadValue_522351" type="ReadValue" version="opset6"> <data variable_id="past_key_values.2.keypresent.2.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.2.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="592" name="Constant_509554" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="593" name="Gather_509555" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="594" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="595" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="596" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="756" /> </output> </layer> <layer id="597" name="__module.model.gpt_neox.layers.2.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="755,759,760,seq_len.9" /> </output> </layer> <layer id="598" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="599" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="600" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="601" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="763,766,cos.9"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="602" name="Constant_51915" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="603" name="__module.model.gpt_neox.layers.2.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="769"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="604" name="__module.model.gpt_neox.layers.2.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="770,cos.11"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="605" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="774"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="606" name="Constant_424070" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="607" name="Constant_424069" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="608" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="609" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="610" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="775,780" /> </output> </layer> <layer id="611" name="__module.model.gpt_neox.layers.2.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="612" name="__module.model.gpt_neox.layers.2.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="777,778,782,783" /> </output> </layer> <layer id="613" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="614" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="615" name="Constant_424068" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="616" name="ScatterUpdate_424071" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="617" name="Constant_424074" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="618" name="Constant_424077" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="619" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="784,x2.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="620" name="Constant_466813" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="621" name="__module.model.gpt_neox.layers.2.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="785"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="622" name="Constant_424123" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="623" name="Constant_424128" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="624" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="779,x1.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="625" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="787"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="626" name="Constant_1487" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="761,sin_cached.5"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="627" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="628" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="629" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="764,767,sin.9"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="630" name="Constant_51918" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="631" name="__module.model.gpt_neox.layers.2.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="772"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="632" name="__module.model.gpt_neox.layers.2.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="773,sin.11"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="633" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="788"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="634" name="__module.model.gpt_neox.layers.2.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="789,query.19"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="635" name="Constant_424175" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="636" name="Constant_424178" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="637" name="Constant_424181" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="638" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="751,query_pass.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="639" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="807,817,query.21,query.23"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="640" name="Constant_424187" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="641" name="Constant_424190" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="642" name="Constant_424193" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="643" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="752,k.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="644" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="790"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="645" name="Constant_424197" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="646" name="Constant_424196" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="647" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="648" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="649" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="791,796" /> </output> </layer> <layer id="650" name="__module.model.gpt_neox.layers.2.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="651" name="__module.model.gpt_neox.layers.2.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="793,794,798,799" /> </output> </layer> <layer id="652" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="653" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="654" name="Constant_424195" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="655" name="ScatterUpdate_424198" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="656" name="Constant_424201" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="657" name="Constant_424204" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="658" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="800,x2.11"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="659" name="Constant_466814" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="660" name="__module.model.gpt_neox.layers.2.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="801"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="661" name="Constant_424250" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="662" name="Constant_424255" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="663" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="795,x1.11"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="664" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="803"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="665" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="804"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="666" name="__module.model.gpt_neox.layers.2.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="805,key.23"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="667" name="Constant_424301" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="668" name="Constant_424304" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="669" name="Constant_424307" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="670" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="753,key_pass.5"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="671" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="809,key.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="672" name="__module.model.gpt_neox.layers.2.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.2.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="673" name="Constant_535069" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="674" name="Constant_535062" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="675" name="Constant_535064" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="676" name="Constant_535066" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="677" name="Concat_535067" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="678" name="Broadcast_535070" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="679" name="ReadValue_522353" type="ReadValue" version="opset6"> <data variable_id="past_key_values.2.valuepresent.2.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.2.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="680" name="Constant_509557" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="681" name="Gather_509558" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="682" name="Constant_424313" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="683" name="Constant_424316" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="684" name="Constant_424319" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="685" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="747"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="686" name="Constant_1441" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="748"> <dim>4</dim> </port> </output> </layer> <layer id="687" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="749,value.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="688" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.2.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="689" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="826"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="690" name="Constant_424323" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="691" name="Constant_424322" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="692" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="693" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="694" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="821" /> </output> </layer> <layer id="695" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="696" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="697" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="819" /> </output> </layer> <layer id="698" name="__module.model.gpt_neox.layers.2.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="823,824" /> </output> </layer> <layer id="699" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="700" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="701" name="Constant_424321" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="702" name="ScatterUpdate_424324" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="703" name="Constant_424325" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="704" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="705" name="ScatterUpdate_424326" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="706" name="Constant_424329" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="707" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="827"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="708" name="Constant_424401" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="709" name="Constant_424402" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="710" name="Constant_424398" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="711" name="Constant_424397" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="712" name="ScatterUpdate_424403" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="713" name="Constant_424406" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="714" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="828,829,causal_mask.13"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="715" name="__module.model.gpt_neox.layers.2.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="716" name="__module.model.gpt_neox.layers.2.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="816,mask_value.5" /> </output> </layer> <layer id="717" name="__module.model.gpt_neox.layers.2.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="830,causal_mask.15"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="718" name="Add_453391" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="719" name="ShapeOf_453392" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="720" name="Constant_402305" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="721" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="722" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="814"> <dim>1</dim> </port> </output> </layer> <layer id="723" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="724" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="725" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="726" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="831"> <dim>4</dim> </port> </output> </layer> <layer id="727" name="__module.model.gpt_neox.layers.2.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="728" name="__module.model.gpt_neox.layers.2.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="729" name="__module.model.gpt_neox.layers.2.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="730" name="__module.model.gpt_neox.layers.2.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="731" name="Maximum_453393" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="732" name="__module.model.gpt_neox.layers.2.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="833,attention_mask.13"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="733" name="__module.model.gpt_neox.layers.2.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="834,835,sdpa_result.5,tensor.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="734" name="Constant_1866" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="836"> <dim>4</dim> </port> </output> </layer> <layer id="735" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="837"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="736" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="737" name="__module.model.gpt_neox.layers.2.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="842,input.39"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="738" name="Constant_1884" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="1637607960" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="739" name="Convert_548981" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="740" name="Constant_1884/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="1675356696" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="741" name="Convert_548984" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="742" name="Subtract_548985" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="743" name="Constant_548986" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="1675362840" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="744" name="Constant_1884/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="745" name="__module.model.gpt_neox.layers.2.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="746" name="Constant_466815" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1524122136" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="747" name="__module.model.gpt_neox.layers.2.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="845,847,input.41"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="748" name="__module.model.gpt_neox.layers.2/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="873"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="749" name="__module.model.gpt_neox.layers.2/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="874,876,input.51"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="750" name="__module.model.gpt_neox.layers.3.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="751" name="__module.model.gpt_neox.layers.3.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="752" name="Constant_466816" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1675387416" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="753" name="__module.model.gpt_neox.layers.3.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="754" name="Constant_466817" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1675411992" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="755" name="__module.model.gpt_neox.layers.3.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1011,input.59"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="756" name="Constant_2484" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="1675436568" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="757" name="Convert_549026" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="758" name="Constant_2484/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="1826431512" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="759" name="Convert_549029" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="760" name="Subtract_549030" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="761" name="Constant_549031" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="1826456088" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="762" name="Constant_2484/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="763" name="__module.model.gpt_neox.layers.3.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="764" name="Constant_466818" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="1826554392" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="765" name="__module.model.gpt_neox.layers.3.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="1016,input.61"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="766" name="__module.model.gpt_neox.layers.3.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="1025,input.63"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="767" name="Constant_2504" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="1826652696" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="768" name="Convert_549035" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="769" name="Constant_2504/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="1977647640" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="770" name="Convert_549038" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="771" name="Subtract_549039" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="772" name="Constant_549040" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="1977653784" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="773" name="Constant_2504/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="774" name="__module.model.gpt_neox.layers.3.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="775" name="Constant_466819" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1977678360" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="776" name="__module.model.gpt_neox.layers.3.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1028,input.65"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="777" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="778" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="779" name="Constant_466820" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1977702936" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="780" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="781" name="Constant_466821" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1977727512" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="782" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="886,input.53"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="783" name="Constant_1963" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="1977752088" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="784" name="Convert_549008" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="785" name="Constant_1963/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="2090998296" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="786" name="Convert_549011" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="787" name="Subtract_549012" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="788" name="Constant_549013" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="2091016728" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="789" name="Constant_1963/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="790" name="__module.model.gpt_neox.layers.3.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="791" name="Constant_466822" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="2091090456" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="792" name="__module.model.gpt_neox.layers.3.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="893,qkv.13"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="793" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="794" name="__module.model.gpt_neox.layers.3.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="897,qkv.15"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="795" name="Constant_424414" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="796" name="Constant_424417" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="797" name="Constant_424420" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="798" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="898"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="799" name="Constant_1991" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="899"> <dim>4</dim> </port> </output> </layer> <layer id="800" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="900,query.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="801" name="Constant_424426" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="802" name="Constant_424429" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="803" name="Constant_424432" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="804" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="907,q.7"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="805" name="Constant_2067" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="919,cos_cached.7"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="806" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="807" name="Constant_424438" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="808" name="Constant_424441" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="809" name="Constant_424444" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="810" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="901"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="811" name="Constant_2005" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="902"> <dim>4</dim> </port> </output> </layer> <layer id="812" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="903,key.31"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="813" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="814" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="815" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="911" /> </output> </layer> <layer id="816" name="Constant_535082" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="817" name="Constant_535075" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="818" name="Constant_535077" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="819" name="Constant_535079" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="820" name="Concat_535080" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="821" name="Broadcast_535083" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="822" name="ReadValue_522395" type="ReadValue" version="opset6"> <data variable_id="past_key_values.3.keypresent.3.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.3.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="823" name="Constant_509560" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="824" name="Gather_509561" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="825" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="826" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="827" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="913" /> </output> </layer> <layer id="828" name="__module.model.gpt_neox.layers.3.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="912,916,917,seq_len.13" /> </output> </layer> <layer id="829" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="830" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="831" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="832" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="920,923,cos.13"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="833" name="Constant_51921" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="834" name="__module.model.gpt_neox.layers.3.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="926"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="835" name="__module.model.gpt_neox.layers.3.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="927,cos.15"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="836" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="931"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="837" name="Constant_424449" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="838" name="Constant_424448" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="839" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="840" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="841" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="932,937" /> </output> </layer> <layer id="842" name="__module.model.gpt_neox.layers.3.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="843" name="__module.model.gpt_neox.layers.3.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="934,935,939,940" /> </output> </layer> <layer id="844" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="845" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="846" name="Constant_424447" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="847" name="ScatterUpdate_424450" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="848" name="Constant_424453" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="849" name="Constant_424456" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="850" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="941,x2.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="851" name="Constant_466823" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="852" name="__module.model.gpt_neox.layers.3.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="942"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="853" name="Constant_424502" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="854" name="Constant_424507" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="855" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="936,x1.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="856" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="944"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="857" name="Constant_2065" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="918,sin_cached.7"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="858" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="859" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="860" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="921,924,sin.13"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="861" name="Constant_51924" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="862" name="__module.model.gpt_neox.layers.3.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="929"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="863" name="__module.model.gpt_neox.layers.3.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="930,sin.15"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="864" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="945"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="865" name="__module.model.gpt_neox.layers.3.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="946,query.27"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="866" name="Constant_424554" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="867" name="Constant_424557" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="868" name="Constant_424560" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="869" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="908,query_pass.7"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="870" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="964,974,query.29,query.31"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="871" name="Constant_424566" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="872" name="Constant_424569" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="873" name="Constant_424572" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="874" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="909,k.7"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="875" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="947"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="876" name="Constant_424576" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="877" name="Constant_424575" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="878" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="879" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="880" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="948,953" /> </output> </layer> <layer id="881" name="__module.model.gpt_neox.layers.3.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="882" name="__module.model.gpt_neox.layers.3.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="950,951,955,956" /> </output> </layer> <layer id="883" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="884" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="885" name="Constant_424574" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="886" name="ScatterUpdate_424577" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="887" name="Constant_424580" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="888" name="Constant_424583" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="889" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="957,x2.15"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="890" name="Constant_466824" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="891" name="__module.model.gpt_neox.layers.3.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="958"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="892" name="Constant_424629" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="893" name="Constant_424634" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="894" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="952,x1.15"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="895" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="960"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="896" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="961"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="897" name="__module.model.gpt_neox.layers.3.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="962,key.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="898" name="Constant_424680" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="899" name="Constant_424683" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="900" name="Constant_424686" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="901" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="910,key_pass.7"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="902" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="966,key.35"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="903" name="__module.model.gpt_neox.layers.3.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.3.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="904" name="Constant_535095" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="905" name="Constant_535088" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="906" name="Constant_535090" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="907" name="Constant_535092" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="908" name="Concat_535093" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="909" name="Broadcast_535096" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="910" name="ReadValue_522397" type="ReadValue" version="opset6"> <data variable_id="past_key_values.3.valuepresent.3.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.3.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="911" name="Constant_509563" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="912" name="Gather_509564" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="913" name="Constant_424692" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="914" name="Constant_424695" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="915" name="Constant_424698" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="916" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="904"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="917" name="Constant_2019" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="905"> <dim>4</dim> </port> </output> </layer> <layer id="918" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="906,value.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="919" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.3.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="920" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="983"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="921" name="Constant_424702" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="922" name="Constant_424701" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="923" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="924" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="925" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="978" /> </output> </layer> <layer id="926" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="927" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="928" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="976" /> </output> </layer> <layer id="929" name="__module.model.gpt_neox.layers.3.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="980,981" /> </output> </layer> <layer id="930" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="931" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="932" name="Constant_424700" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="933" name="ScatterUpdate_424703" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="934" name="Constant_424704" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="935" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="936" name="ScatterUpdate_424705" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="937" name="Constant_424708" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="938" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="984"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="939" name="Constant_424780" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="940" name="Constant_424781" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="941" name="Constant_424777" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="942" name="Constant_424776" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="943" name="ScatterUpdate_424782" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="944" name="Constant_424785" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="945" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="985,986,causal_mask.19"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="946" name="__module.model.gpt_neox.layers.3.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="947" name="__module.model.gpt_neox.layers.3.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="973,mask_value.7" /> </output> </layer> <layer id="948" name="__module.model.gpt_neox.layers.3.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="987,causal_mask.21"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="949" name="Add_453467" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="950" name="ShapeOf_453468" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="951" name="Constant_402320" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="952" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="953" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="971"> <dim>1</dim> </port> </output> </layer> <layer id="954" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="955" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="956" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="957" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="988"> <dim>4</dim> </port> </output> </layer> <layer id="958" name="__module.model.gpt_neox.layers.3.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="959" name="__module.model.gpt_neox.layers.3.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="960" name="__module.model.gpt_neox.layers.3.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="961" name="__module.model.gpt_neox.layers.3.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="962" name="Maximum_453469" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="963" name="__module.model.gpt_neox.layers.3.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="990,attention_mask.15"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="964" name="__module.model.gpt_neox.layers.3.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="991,992,sdpa_result.7,tensor.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="965" name="Constant_2444" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="993"> <dim>4</dim> </port> </output> </layer> <layer id="966" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="994"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="967" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="968" name="__module.model.gpt_neox.layers.3.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="999,input.55"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="969" name="Constant_2462" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="2091164184" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="970" name="Convert_549017" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="971" name="Constant_2462/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="2128912920" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="972" name="Convert_549020" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="973" name="Subtract_549021" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="974" name="Constant_549022" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="2128919064" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="975" name="Constant_2462/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="976" name="__module.model.gpt_neox.layers.3.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="977" name="Constant_466825" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="1977678360" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="978" name="__module.model.gpt_neox.layers.3.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1002,1004,input.57"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="979" name="__module.model.gpt_neox.layers.3/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1030"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="980" name="__module.model.gpt_neox.layers.3/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1031,1033,input.67"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="981" name="__module.model.gpt_neox.layers.4.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="982" name="__module.model.gpt_neox.layers.4.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="983" name="Constant_466826" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2128943640" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="984" name="__module.model.gpt_neox.layers.4.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="985" name="Constant_466827" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2128968216" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="986" name="__module.model.gpt_neox.layers.4.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1168,input.75"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="987" name="Constant_3062" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="2128992792" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="988" name="Convert_549062" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="989" name="Constant_3062/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="2279987736" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="990" name="Convert_549065" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="991" name="Subtract_549066" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="992" name="Constant_549067" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="2280012312" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="993" name="Constant_3062/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="994" name="__module.model.gpt_neox.layers.4.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="995" name="Constant_466828" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="2280110616" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="996" name="__module.model.gpt_neox.layers.4.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="1173,input.77"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="997" name="__module.model.gpt_neox.layers.4.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="1182,input.79"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="998" name="Constant_3082" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="2280208920" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="999" name="Convert_549071" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1000" name="Constant_3082/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="2431203864" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1001" name="Convert_549074" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1002" name="Subtract_549075" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1003" name="Constant_549076" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="2431210008" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1004" name="Constant_3082/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1005" name="__module.model.gpt_neox.layers.4.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1006" name="Constant_466829" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2431234584" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1007" name="__module.model.gpt_neox.layers.4.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1185,input.81"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1008" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1009" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1010" name="Constant_466830" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2431259160" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1011" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1012" name="Constant_466831" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2431283736" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1013" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1043,input.69"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1014" name="Constant_2541" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="2431308312" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1015" name="Convert_549044" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1016" name="Constant_2541/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="2544554520" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1017" name="Convert_549047" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1018" name="Subtract_549048" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1019" name="Constant_549049" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="2544572952" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1020" name="Constant_2541/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1021" name="__module.model.gpt_neox.layers.4.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1022" name="Constant_466832" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="2544646680" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1023" name="__module.model.gpt_neox.layers.4.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="1050,qkv.17"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1024" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1025" name="__module.model.gpt_neox.layers.4.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1054,qkv.19"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="1026" name="Constant_424793" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1027" name="Constant_424796" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1028" name="Constant_424799" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1029" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1055"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1030" name="Constant_2569" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1056"> <dim>4</dim> </port> </output> </layer> <layer id="1031" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1057,query.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1032" name="Constant_424805" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1033" name="Constant_424808" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1034" name="Constant_424811" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1035" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1064,q.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1036" name="Constant_2645" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="1076,cos_cached.9"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1037" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1038" name="Constant_424817" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1039" name="Constant_424820" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1040" name="Constant_424823" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1041" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1058"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1042" name="Constant_2583" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1059"> <dim>4</dim> </port> </output> </layer> <layer id="1043" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1060,key.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1044" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1045" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1046" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1068" /> </output> </layer> <layer id="1047" name="Constant_535108" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1048" name="Constant_535101" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1049" name="Constant_535103" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1050" name="Constant_535105" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1051" name="Concat_535106" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1052" name="Broadcast_535109" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1053" name="ReadValue_522439" type="ReadValue" version="opset6"> <data variable_id="past_key_values.4.keypresent.4.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.4.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1054" name="Constant_509566" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1055" name="Gather_509567" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1056" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1057" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1058" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1070" /> </output> </layer> <layer id="1059" name="__module.model.gpt_neox.layers.4.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1069,1073,1074,seq_len.17" /> </output> </layer> <layer id="1060" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1061" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1062" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1063" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1077,1080,cos.17"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1064" name="Constant_51927" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1065" name="__module.model.gpt_neox.layers.4.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1083"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1066" name="__module.model.gpt_neox.layers.4.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1084,cos.19"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1067" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1088"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1068" name="Constant_424828" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1069" name="Constant_424827" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1070" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1071" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1072" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1089,1094" /> </output> </layer> <layer id="1073" name="__module.model.gpt_neox.layers.4.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1074" name="__module.model.gpt_neox.layers.4.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1091,1092,1096,1097" /> </output> </layer> <layer id="1075" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1076" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1077" name="Constant_424826" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1078" name="ScatterUpdate_424829" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1079" name="Constant_424832" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1080" name="Constant_424835" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1081" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1098,x2.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1082" name="Constant_466833" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="1083" name="__module.model.gpt_neox.layers.4.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1099"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1084" name="Constant_424881" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1085" name="Constant_424886" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1086" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1093,x1.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1087" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1101"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1088" name="Constant_2643" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="1075,sin_cached.9"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1089" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1090" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1091" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1078,1081,sin.17"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1092" name="Constant_51930" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1093" name="__module.model.gpt_neox.layers.4.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1086"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1094" name="__module.model.gpt_neox.layers.4.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1087,sin.19"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1095" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1102"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1096" name="__module.model.gpt_neox.layers.4.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1103,query.35"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1097" name="Constant_424933" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1098" name="Constant_424936" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1099" name="Constant_424939" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1100" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1065,query_pass.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="1101" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1121,1131,query.37,query.39"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1102" name="Constant_424945" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1103" name="Constant_424948" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1104" name="Constant_424951" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1105" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1066,k.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1106" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1104"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1107" name="Constant_424955" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1108" name="Constant_424954" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1109" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1110" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1111" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1105,1110" /> </output> </layer> <layer id="1112" name="__module.model.gpt_neox.layers.4.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1113" name="__module.model.gpt_neox.layers.4.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1107,1108,1112,1113" /> </output> </layer> <layer id="1114" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1115" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1116" name="Constant_424953" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1117" name="ScatterUpdate_424956" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1118" name="Constant_424959" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1119" name="Constant_424962" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1120" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1114,x2.19"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1121" name="Constant_466834" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="1122" name="__module.model.gpt_neox.layers.4.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1115"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1123" name="Constant_425008" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1124" name="Constant_425013" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1125" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1109,x1.19"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1126" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1117"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1127" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1118"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1128" name="__module.model.gpt_neox.layers.4.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1119,key.43"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1129" name="Constant_425059" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1130" name="Constant_425062" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1131" name="Constant_425065" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1132" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1067,key_pass.9"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="1133" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1123,key.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1134" name="__module.model.gpt_neox.layers.4.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.4.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1135" name="Constant_535121" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1136" name="Constant_535114" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1137" name="Constant_535116" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1138" name="Constant_535118" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1139" name="Concat_535119" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1140" name="Broadcast_535122" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1141" name="ReadValue_522441" type="ReadValue" version="opset6"> <data variable_id="past_key_values.4.valuepresent.4.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.4.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1142" name="Constant_509569" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1143" name="Gather_509570" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1144" name="Constant_425071" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1145" name="Constant_425074" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1146" name="Constant_425077" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1147" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1061"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1148" name="Constant_2597" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1062"> <dim>4</dim> </port> </output> </layer> <layer id="1149" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1063,value.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1150" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.4.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1151" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="1140"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="1152" name="Constant_425081" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1153" name="Constant_425080" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1154" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1155" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1156" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1135" /> </output> </layer> <layer id="1157" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1158" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1159" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1133" /> </output> </layer> <layer id="1160" name="__module.model.gpt_neox.layers.4.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1137,1138" /> </output> </layer> <layer id="1161" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1162" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1163" name="Constant_425079" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1164" name="ScatterUpdate_425082" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1165" name="Constant_425083" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1166" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1167" name="ScatterUpdate_425084" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1168" name="Constant_425087" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1169" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1141"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="1170" name="Constant_425159" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1171" name="Constant_425160" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1172" name="Constant_425156" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1173" name="Constant_425155" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1174" name="ScatterUpdate_425161" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1175" name="Constant_425164" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1176" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1142,1143,causal_mask.25"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1177" name="__module.model.gpt_neox.layers.4.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1178" name="__module.model.gpt_neox.layers.4.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="1130,mask_value.9" /> </output> </layer> <layer id="1179" name="__module.model.gpt_neox.layers.4.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="1144,causal_mask.27"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1180" name="Add_453543" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1181" name="ShapeOf_453544" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1182" name="Constant_402335" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1183" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1184" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1128"> <dim>1</dim> </port> </output> </layer> <layer id="1185" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1186" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1187" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1188" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="1145"> <dim>4</dim> </port> </output> </layer> <layer id="1189" name="__module.model.gpt_neox.layers.4.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1190" name="__module.model.gpt_neox.layers.4.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="1191" name="__module.model.gpt_neox.layers.4.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1192" name="__module.model.gpt_neox.layers.4.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1193" name="Maximum_453545" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1194" name="__module.model.gpt_neox.layers.4.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1147,attention_mask.17"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1195" name="__module.model.gpt_neox.layers.4.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1148,1149,sdpa_result.9,tensor.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1196" name="Constant_3022" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1150"> <dim>4</dim> </port> </output> </layer> <layer id="1197" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1151"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1198" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1199" name="__module.model.gpt_neox.layers.4.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="1156,input.71"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1200" name="Constant_3040" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="2544720408" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1201" name="Convert_549053" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1202" name="Constant_3040/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="2582469144" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1203" name="Convert_549056" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1204" name="Subtract_549057" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1205" name="Constant_549058" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="2582475288" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1206" name="Constant_3040/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1207" name="__module.model.gpt_neox.layers.4.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1208" name="Constant_466835" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2431234584" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1209" name="__module.model.gpt_neox.layers.4.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1159,1161,input.73"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1210" name="__module.model.gpt_neox.layers.4/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1187"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1211" name="__module.model.gpt_neox.layers.4/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1188,1190,input.83"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1212" name="__module.model.gpt_neox.layers.5.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1213" name="__module.model.gpt_neox.layers.5.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1214" name="Constant_466836" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2582499864" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1215" name="__module.model.gpt_neox.layers.5.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1216" name="Constant_466837" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2582524440" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1217" name="__module.model.gpt_neox.layers.5.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1325,input.91"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1218" name="Constant_3640" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="2582549016" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1219" name="Convert_549098" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1220" name="Constant_3640/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="2733543960" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1221" name="Convert_549101" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1222" name="Subtract_549102" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1223" name="Constant_549103" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="2733568536" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1224" name="Constant_3640/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1225" name="__module.model.gpt_neox.layers.5.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1226" name="Constant_466838" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="2733666840" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1227" name="__module.model.gpt_neox.layers.5.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="1330,input.93"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1228" name="__module.model.gpt_neox.layers.5.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="1339,input.95"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1229" name="Constant_3660" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="2733765144" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1230" name="Convert_549107" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1231" name="Constant_3660/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="2884760088" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1232" name="Convert_549110" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1233" name="Subtract_549111" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1234" name="Constant_549112" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="2884766232" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1235" name="Constant_3660/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1236" name="__module.model.gpt_neox.layers.5.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1237" name="Constant_466839" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2884790808" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1238" name="__module.model.gpt_neox.layers.5.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1342,input.97"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1239" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1240" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1241" name="Constant_466840" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2884815384" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1242" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1243" name="Constant_466841" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2884839960" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1244" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1200,input.85"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1245" name="Constant_3119" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="2884864536" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1246" name="Convert_549080" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1247" name="Constant_3119/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="2998110744" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1248" name="Convert_549083" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1249" name="Subtract_549084" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1250" name="Constant_549085" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="2998129176" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1251" name="Constant_3119/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1252" name="__module.model.gpt_neox.layers.5.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1253" name="Constant_466842" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="2998202904" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1254" name="__module.model.gpt_neox.layers.5.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="1207,qkv.21"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1255" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1256" name="__module.model.gpt_neox.layers.5.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1211,qkv.23"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="1257" name="Constant_425172" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1258" name="Constant_425175" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1259" name="Constant_425178" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1260" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1212"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1261" name="Constant_3147" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1213"> <dim>4</dim> </port> </output> </layer> <layer id="1262" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1214,query.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1263" name="Constant_425184" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1264" name="Constant_425187" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1265" name="Constant_425190" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1266" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1221,q.11"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1267" name="Constant_3223" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="1233,cos_cached.11"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1268" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1269" name="Constant_425196" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1270" name="Constant_425199" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1271" name="Constant_425202" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1272" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1215"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1273" name="Constant_3161" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1216"> <dim>4</dim> </port> </output> </layer> <layer id="1274" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1217,key.51"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1275" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1276" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1277" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1225" /> </output> </layer> <layer id="1278" name="Constant_535134" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1279" name="Constant_535127" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1280" name="Constant_535129" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1281" name="Constant_535131" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1282" name="Concat_535132" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1283" name="Broadcast_535135" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1284" name="ReadValue_522459" type="ReadValue" version="opset6"> <data variable_id="past_key_values.5.keypresent.5.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.5.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1285" name="Constant_509572" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1286" name="Gather_509573" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1287" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1288" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1289" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1227" /> </output> </layer> <layer id="1290" name="__module.model.gpt_neox.layers.5.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1226,1230,1231,seq_len.21" /> </output> </layer> <layer id="1291" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1292" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1293" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1294" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1234,1237,cos.21"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1295" name="Constant_51933" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1296" name="__module.model.gpt_neox.layers.5.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1240"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1297" name="__module.model.gpt_neox.layers.5.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1241,cos.23"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1298" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1245"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1299" name="Constant_425207" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1300" name="Constant_425206" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1301" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1302" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1303" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1246,1251" /> </output> </layer> <layer id="1304" name="__module.model.gpt_neox.layers.5.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1305" name="__module.model.gpt_neox.layers.5.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1248,1249,1253,1254" /> </output> </layer> <layer id="1306" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1307" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1308" name="Constant_425205" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1309" name="ScatterUpdate_425208" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1310" name="Constant_425211" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1311" name="Constant_425214" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1312" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1255,x2.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1313" name="Constant_466843" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="1314" name="__module.model.gpt_neox.layers.5.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1256"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1315" name="Constant_425260" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1316" name="Constant_425265" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1317" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1250,x1.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1318" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1258"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1319" name="Constant_3221" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="1232,sin_cached.11"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1320" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1321" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1322" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1235,1238,sin.21"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1323" name="Constant_51936" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1324" name="__module.model.gpt_neox.layers.5.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1243"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1325" name="__module.model.gpt_neox.layers.5.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1244,sin.23"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1326" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1259"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1327" name="__module.model.gpt_neox.layers.5.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1260,query.43"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1328" name="Constant_425312" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1329" name="Constant_425315" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1330" name="Constant_425318" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1331" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1222,query_pass.11"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="1332" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1278,1288,query.45,query.47"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1333" name="Constant_425324" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1334" name="Constant_425327" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1335" name="Constant_425330" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1336" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1223,k.11"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1337" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1261"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1338" name="Constant_425334" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1339" name="Constant_425333" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1340" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1341" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1342" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1262,1267" /> </output> </layer> <layer id="1343" name="__module.model.gpt_neox.layers.5.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1344" name="__module.model.gpt_neox.layers.5.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1264,1265,1269,1270" /> </output> </layer> <layer id="1345" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1346" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1347" name="Constant_425332" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1348" name="ScatterUpdate_425335" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1349" name="Constant_425338" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1350" name="Constant_425341" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1351" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1271,x2.23"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1352" name="Constant_466844" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="1353" name="__module.model.gpt_neox.layers.5.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1272"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1354" name="Constant_425387" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1355" name="Constant_425392" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1356" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1266,x1.23"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1357" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1274"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1358" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1275"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1359" name="__module.model.gpt_neox.layers.5.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1276,key.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1360" name="Constant_425438" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1361" name="Constant_425441" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1362" name="Constant_425444" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1363" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1224,key_pass.11"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="1364" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1280,key.55"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1365" name="__module.model.gpt_neox.layers.5.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.5.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1366" name="Constant_535147" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1367" name="Constant_535140" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1368" name="Constant_535142" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1369" name="Constant_535144" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1370" name="Concat_535145" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1371" name="Broadcast_535148" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1372" name="ReadValue_522461" type="ReadValue" version="opset6"> <data variable_id="past_key_values.5.valuepresent.5.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.5.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1373" name="Constant_509575" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1374" name="Gather_509576" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1375" name="Constant_425450" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1376" name="Constant_425453" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1377" name="Constant_425456" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1378" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1218"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1379" name="Constant_3175" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1219"> <dim>4</dim> </port> </output> </layer> <layer id="1380" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1220,value.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1381" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.5.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1382" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="1297"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="1383" name="Constant_425460" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1384" name="Constant_425459" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1385" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1386" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1387" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1292" /> </output> </layer> <layer id="1388" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1389" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1390" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1290" /> </output> </layer> <layer id="1391" name="__module.model.gpt_neox.layers.5.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1294,1295" /> </output> </layer> <layer id="1392" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1393" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1394" name="Constant_425458" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1395" name="ScatterUpdate_425461" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1396" name="Constant_425462" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1397" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1398" name="ScatterUpdate_425463" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1399" name="Constant_425466" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1400" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1298"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="1401" name="Constant_425538" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1402" name="Constant_425539" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1403" name="Constant_425535" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1404" name="Constant_425534" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1405" name="ScatterUpdate_425540" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1406" name="Constant_425543" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1407" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1299,1300,causal_mask.31"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1408" name="__module.model.gpt_neox.layers.5.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1409" name="__module.model.gpt_neox.layers.5.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="1287,mask_value.11" /> </output> </layer> <layer id="1410" name="__module.model.gpt_neox.layers.5.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="1301,causal_mask.33"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1411" name="Add_453619" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1412" name="ShapeOf_453620" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1413" name="Constant_402350" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1414" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1415" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1285"> <dim>1</dim> </port> </output> </layer> <layer id="1416" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1417" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1418" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1419" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="1302"> <dim>4</dim> </port> </output> </layer> <layer id="1420" name="__module.model.gpt_neox.layers.5.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1421" name="__module.model.gpt_neox.layers.5.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="1422" name="__module.model.gpt_neox.layers.5.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1423" name="__module.model.gpt_neox.layers.5.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1424" name="Maximum_453621" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1425" name="__module.model.gpt_neox.layers.5.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1304,attention_mask.19"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1426" name="__module.model.gpt_neox.layers.5.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1305,1306,sdpa_result.11,tensor.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1427" name="Constant_3600" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1307"> <dim>4</dim> </port> </output> </layer> <layer id="1428" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1308"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1429" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1430" name="__module.model.gpt_neox.layers.5.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="1313,input.87"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1431" name="Constant_3618" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="2998276632" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1432" name="Convert_549089" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1433" name="Constant_3618/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="3036025368" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1434" name="Convert_549092" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1435" name="Subtract_549093" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1436" name="Constant_549094" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="3036031512" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1437" name="Constant_3618/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1438" name="__module.model.gpt_neox.layers.5.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1439" name="Constant_466845" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="2884790808" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1440" name="__module.model.gpt_neox.layers.5.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1316,1318,input.89"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1441" name="__module.model.gpt_neox.layers.5/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1344"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1442" name="__module.model.gpt_neox.layers.5/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1345,1347,input.99"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1443" name="__module.model.gpt_neox.layers.6.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1444" name="__module.model.gpt_neox.layers.6.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1445" name="Constant_466846" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3036056088" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1446" name="__module.model.gpt_neox.layers.6.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1447" name="Constant_466847" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3036080664" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1448" name="__module.model.gpt_neox.layers.6.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1482,input.107"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1449" name="Constant_4218" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="3036105240" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1450" name="Convert_549134" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1451" name="Constant_4218/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="3187100184" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1452" name="Convert_549137" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1453" name="Subtract_549138" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1454" name="Constant_549139" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="3187124760" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1455" name="Constant_4218/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1456" name="__module.model.gpt_neox.layers.6.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1457" name="Constant_466848" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="3187223064" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1458" name="__module.model.gpt_neox.layers.6.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="1487,input.109"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1459" name="__module.model.gpt_neox.layers.6.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="1496,input.111"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1460" name="Constant_4238" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="3187321368" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1461" name="Convert_549143" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1462" name="Constant_4238/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="3338316312" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1463" name="Convert_549146" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1464" name="Subtract_549147" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1465" name="Constant_549148" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="3338322456" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1466" name="Constant_4238/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1467" name="__module.model.gpt_neox.layers.6.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1468" name="Constant_466849" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3338347032" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1469" name="__module.model.gpt_neox.layers.6.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1499,input.113"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1470" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1471" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1472" name="Constant_466850" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3338371608" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1473" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1474" name="Constant_466851" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3338396184" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1475" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1357,input.101"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1476" name="Constant_3697" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="3338420760" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1477" name="Convert_549116" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1478" name="Constant_3697/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="3451666968" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1479" name="Convert_549119" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1480" name="Subtract_549120" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1481" name="Constant_549121" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="3451685400" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1482" name="Constant_3697/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1483" name="__module.model.gpt_neox.layers.6.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1484" name="Constant_466852" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="3451759128" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1485" name="__module.model.gpt_neox.layers.6.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="1364,qkv.25"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1486" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1487" name="__module.model.gpt_neox.layers.6.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1368,qkv.27"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="1488" name="Constant_425551" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1489" name="Constant_425554" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1490" name="Constant_425557" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1491" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1369"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1492" name="Constant_3725" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1370"> <dim>4</dim> </port> </output> </layer> <layer id="1493" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1371,query.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1494" name="Constant_425563" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1495" name="Constant_425566" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1496" name="Constant_425569" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1497" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1378,q.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1498" name="Constant_3801" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="1390,cos_cached.13"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1499" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1500" name="Constant_425575" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1501" name="Constant_425578" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1502" name="Constant_425581" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1503" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1372"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1504" name="Constant_3739" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1373"> <dim>4</dim> </port> </output> </layer> <layer id="1505" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1374,key.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1506" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1507" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1508" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1382" /> </output> </layer> <layer id="1509" name="Constant_535160" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1510" name="Constant_535153" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1511" name="Constant_535155" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1512" name="Constant_535157" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1513" name="Concat_535158" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1514" name="Broadcast_535161" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1515" name="ReadValue_522463" type="ReadValue" version="opset6"> <data variable_id="past_key_values.6.keypresent.6.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.6.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1516" name="Constant_509578" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1517" name="Gather_509579" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1518" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1519" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1520" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1384" /> </output> </layer> <layer id="1521" name="__module.model.gpt_neox.layers.6.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1383,1387,1388,seq_len.25" /> </output> </layer> <layer id="1522" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1523" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1524" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1525" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1391,1394,cos.25"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1526" name="Constant_51939" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1527" name="__module.model.gpt_neox.layers.6.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1397"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1528" name="__module.model.gpt_neox.layers.6.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1398,cos.27"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1529" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1402"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1530" name="Constant_425586" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1531" name="Constant_425585" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1532" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1533" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1534" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1403,1408" /> </output> </layer> <layer id="1535" name="__module.model.gpt_neox.layers.6.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1536" name="__module.model.gpt_neox.layers.6.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1405,1406,1410,1411" /> </output> </layer> <layer id="1537" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1538" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1539" name="Constant_425584" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1540" name="ScatterUpdate_425587" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1541" name="Constant_425590" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1542" name="Constant_425593" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1543" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1412,x2.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1544" name="Constant_466853" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="1545" name="__module.model.gpt_neox.layers.6.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1413"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1546" name="Constant_425639" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1547" name="Constant_425644" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1548" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1407,x1.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1549" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1415"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1550" name="Constant_3799" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="1389,sin_cached.13"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1551" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1552" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1553" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1392,1395,sin.25"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1554" name="Constant_51942" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1555" name="__module.model.gpt_neox.layers.6.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1400"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1556" name="__module.model.gpt_neox.layers.6.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1401,sin.27"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1557" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1416"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1558" name="__module.model.gpt_neox.layers.6.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1417,query.51"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1559" name="Constant_425691" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1560" name="Constant_425694" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1561" name="Constant_425697" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1562" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1379,query_pass.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="1563" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1435,1445,query.53,query.55"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1564" name="Constant_425703" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1565" name="Constant_425706" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1566" name="Constant_425709" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1567" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1380,k.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1568" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1418"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1569" name="Constant_425713" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1570" name="Constant_425712" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1571" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1572" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1573" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1419,1424" /> </output> </layer> <layer id="1574" name="__module.model.gpt_neox.layers.6.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1575" name="__module.model.gpt_neox.layers.6.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1421,1422,1426,1427" /> </output> </layer> <layer id="1576" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1577" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1578" name="Constant_425711" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1579" name="ScatterUpdate_425714" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1580" name="Constant_425717" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1581" name="Constant_425720" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1582" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1428,x2.27"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1583" name="Constant_466854" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="1584" name="__module.model.gpt_neox.layers.6.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1429"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1585" name="Constant_425766" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1586" name="Constant_425771" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1587" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1423,x1.27"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1588" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1431"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1589" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1432"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1590" name="__module.model.gpt_neox.layers.6.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1433,key.63"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1591" name="Constant_425817" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1592" name="Constant_425820" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1593" name="Constant_425823" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1594" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1381,key_pass.13"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="1595" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1437,key.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1596" name="__module.model.gpt_neox.layers.6.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.6.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1597" name="Constant_535173" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1598" name="Constant_535166" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1599" name="Constant_535168" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1600" name="Constant_535170" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1601" name="Concat_535171" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1602" name="Broadcast_535174" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1603" name="ReadValue_522465" type="ReadValue" version="opset6"> <data variable_id="past_key_values.6.valuepresent.6.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.6.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1604" name="Constant_509581" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1605" name="Gather_509582" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1606" name="Constant_425829" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1607" name="Constant_425832" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1608" name="Constant_425835" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1609" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1375"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1610" name="Constant_3753" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1376"> <dim>4</dim> </port> </output> </layer> <layer id="1611" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1377,value.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1612" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.6.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1613" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="1454"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="1614" name="Constant_425839" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1615" name="Constant_425838" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1616" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1617" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1618" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1449" /> </output> </layer> <layer id="1619" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1620" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1621" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1447" /> </output> </layer> <layer id="1622" name="__module.model.gpt_neox.layers.6.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1451,1452" /> </output> </layer> <layer id="1623" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1624" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1625" name="Constant_425837" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1626" name="ScatterUpdate_425840" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1627" name="Constant_425841" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1628" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1629" name="ScatterUpdate_425842" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1630" name="Constant_425845" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1631" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1455"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="1632" name="Constant_425917" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1633" name="Constant_425918" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1634" name="Constant_425914" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1635" name="Constant_425913" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1636" name="ScatterUpdate_425919" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1637" name="Constant_425922" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1638" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1456,1457,causal_mask.37"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1639" name="__module.model.gpt_neox.layers.6.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1640" name="__module.model.gpt_neox.layers.6.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="1444,mask_value.13" /> </output> </layer> <layer id="1641" name="__module.model.gpt_neox.layers.6.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="1458,causal_mask.39"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1642" name="Add_453695" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1643" name="ShapeOf_453696" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1644" name="Constant_402365" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1645" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1646" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1442"> <dim>1</dim> </port> </output> </layer> <layer id="1647" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1648" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1649" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1650" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="1459"> <dim>4</dim> </port> </output> </layer> <layer id="1651" name="__module.model.gpt_neox.layers.6.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1652" name="__module.model.gpt_neox.layers.6.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="1653" name="__module.model.gpt_neox.layers.6.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1654" name="__module.model.gpt_neox.layers.6.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1655" name="Maximum_453697" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1656" name="__module.model.gpt_neox.layers.6.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1461,attention_mask.21"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1657" name="__module.model.gpt_neox.layers.6.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1462,1463,sdpa_result.13,tensor.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1658" name="Constant_4178" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1464"> <dim>4</dim> </port> </output> </layer> <layer id="1659" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1465"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1660" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1661" name="__module.model.gpt_neox.layers.6.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="1470,input.103"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1662" name="Constant_4196" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="3451832856" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1663" name="Convert_549125" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1664" name="Constant_4196/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="3489581592" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1665" name="Convert_549128" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1666" name="Subtract_549129" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1667" name="Constant_549130" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="3489587736" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1668" name="Constant_4196/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1669" name="__module.model.gpt_neox.layers.6.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1670" name="Constant_466855" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3338347032" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1671" name="__module.model.gpt_neox.layers.6.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1473,1475,input.105"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1672" name="__module.model.gpt_neox.layers.6/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1501"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1673" name="__module.model.gpt_neox.layers.6/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1502,1504,input.115"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1674" name="__module.model.gpt_neox.layers.7.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1675" name="__module.model.gpt_neox.layers.7.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1676" name="Constant_466856" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3489612312" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1677" name="__module.model.gpt_neox.layers.7.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1678" name="Constant_466857" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3489636888" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1679" name="__module.model.gpt_neox.layers.7.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1639,input.123"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1680" name="Constant_4796" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="3489661464" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1681" name="Convert_549170" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1682" name="Constant_4796/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="3640656408" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1683" name="Convert_549173" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1684" name="Subtract_549174" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1685" name="Constant_549175" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="3640680984" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1686" name="Constant_4796/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1687" name="__module.model.gpt_neox.layers.7.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1688" name="Constant_466858" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="3640779288" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1689" name="__module.model.gpt_neox.layers.7.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="1644,input.125"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1690" name="__module.model.gpt_neox.layers.7.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="1653,input.127"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1691" name="Constant_4816" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="3640877592" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1692" name="Convert_549179" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1693" name="Constant_4816/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="3791872536" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1694" name="Convert_549182" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1695" name="Subtract_549183" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1696" name="Constant_549184" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="3791878680" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1697" name="Constant_4816/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1698" name="__module.model.gpt_neox.layers.7.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1699" name="Constant_466859" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3791903256" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1700" name="__module.model.gpt_neox.layers.7.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1656,input.129"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1701" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1702" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1703" name="Constant_466860" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3791927832" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1704" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1705" name="Constant_466861" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3791952408" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1706" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1514,input.117"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1707" name="Constant_4275" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="3791976984" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1708" name="Convert_549152" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1709" name="Constant_4275/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="3905223192" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1710" name="Convert_549155" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1711" name="Subtract_549156" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1712" name="Constant_549157" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="3905241624" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1713" name="Constant_4275/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1714" name="__module.model.gpt_neox.layers.7.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1715" name="Constant_466862" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="3905315352" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1716" name="__module.model.gpt_neox.layers.7.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="1521,qkv.29"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1717" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1718" name="__module.model.gpt_neox.layers.7.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1525,qkv.31"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="1719" name="Constant_425930" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1720" name="Constant_425933" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1721" name="Constant_425936" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1722" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1526"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1723" name="Constant_4303" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1527"> <dim>4</dim> </port> </output> </layer> <layer id="1724" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1528,query.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1725" name="Constant_425942" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1726" name="Constant_425945" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1727" name="Constant_425948" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1728" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1535,q.15"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1729" name="Constant_4379" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="1547,cos_cached.15"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1730" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1731" name="Constant_425954" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1732" name="Constant_425957" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1733" name="Constant_425960" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1734" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1529"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1735" name="Constant_4317" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1530"> <dim>4</dim> </port> </output> </layer> <layer id="1736" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1531,key.71"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1737" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1738" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1739" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1539" /> </output> </layer> <layer id="1740" name="Constant_535186" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1741" name="Constant_535179" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1742" name="Constant_535181" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1743" name="Constant_535183" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1744" name="Concat_535184" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1745" name="Broadcast_535187" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1746" name="ReadValue_522467" type="ReadValue" version="opset6"> <data variable_id="past_key_values.7.keypresent.7.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.7.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1747" name="Constant_509584" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1748" name="Gather_509585" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1749" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1750" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1751" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1541" /> </output> </layer> <layer id="1752" name="__module.model.gpt_neox.layers.7.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1540,1544,1545,seq_len.29" /> </output> </layer> <layer id="1753" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1754" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1755" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1756" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1548,1551,cos.29"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1757" name="Constant_51945" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1758" name="__module.model.gpt_neox.layers.7.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1554"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1759" name="__module.model.gpt_neox.layers.7.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1555,cos.31"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1760" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1559"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1761" name="Constant_425965" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1762" name="Constant_425964" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1763" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1764" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1765" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1560,1565" /> </output> </layer> <layer id="1766" name="__module.model.gpt_neox.layers.7.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1767" name="__module.model.gpt_neox.layers.7.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1562,1563,1567,1568" /> </output> </layer> <layer id="1768" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1769" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1770" name="Constant_425963" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1771" name="ScatterUpdate_425966" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1772" name="Constant_425969" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1773" name="Constant_425972" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1774" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1569,x2.29"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1775" name="Constant_466863" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="1776" name="__module.model.gpt_neox.layers.7.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1570"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1777" name="Constant_426018" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1778" name="Constant_426023" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1779" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1564,x1.29"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1780" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1572"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1781" name="Constant_4377" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="1546,sin_cached.15"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1782" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1783" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1784" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1549,1552,sin.29"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1785" name="Constant_51948" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1786" name="__module.model.gpt_neox.layers.7.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1557"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1787" name="__module.model.gpt_neox.layers.7.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1558,sin.31"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1788" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1573"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1789" name="__module.model.gpt_neox.layers.7.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1574,query.59"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1790" name="Constant_426070" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1791" name="Constant_426073" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1792" name="Constant_426076" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1793" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1536,query_pass.15"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="1794" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1592,1602,query.61,query.63"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1795" name="Constant_426082" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1796" name="Constant_426085" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1797" name="Constant_426088" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1798" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1537,k.15"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1799" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1575"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1800" name="Constant_426092" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1801" name="Constant_426091" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1802" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1803" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1804" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1576,1581" /> </output> </layer> <layer id="1805" name="__module.model.gpt_neox.layers.7.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1806" name="__module.model.gpt_neox.layers.7.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1578,1579,1583,1584" /> </output> </layer> <layer id="1807" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1808" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1809" name="Constant_426090" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1810" name="ScatterUpdate_426093" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1811" name="Constant_426096" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1812" name="Constant_426099" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1813" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1585,x2.31"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1814" name="Constant_466864" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="1815" name="__module.model.gpt_neox.layers.7.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1586"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1816" name="Constant_426145" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1817" name="Constant_426150" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1818" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1580,x1.31"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="1819" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1588"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1820" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1589"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1821" name="__module.model.gpt_neox.layers.7.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1590,key.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1822" name="Constant_426196" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1823" name="Constant_426199" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1824" name="Constant_426202" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1825" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1538,key_pass.15"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="1826" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1594,key.75"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1827" name="__module.model.gpt_neox.layers.7.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.7.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1828" name="Constant_535199" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1829" name="Constant_535192" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1830" name="Constant_535194" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1831" name="Constant_535196" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1832" name="Concat_535197" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1833" name="Broadcast_535200" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1834" name="ReadValue_522469" type="ReadValue" version="opset6"> <data variable_id="past_key_values.7.valuepresent.7.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.7.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1835" name="Constant_509587" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1836" name="Gather_509588" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1837" name="Constant_426208" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1838" name="Constant_426211" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1839" name="Constant_426214" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1840" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1532"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1841" name="Constant_4331" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1533"> <dim>4</dim> </port> </output> </layer> <layer id="1842" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1534,value.29"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1843" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.7.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1844" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="1611"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="1845" name="Constant_426218" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1846" name="Constant_426217" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1847" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1848" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1849" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1606" /> </output> </layer> <layer id="1850" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1851" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1852" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1604" /> </output> </layer> <layer id="1853" name="__module.model.gpt_neox.layers.7.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1608,1609" /> </output> </layer> <layer id="1854" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1855" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1856" name="Constant_426216" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1857" name="ScatterUpdate_426219" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1858" name="Constant_426220" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1859" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1860" name="ScatterUpdate_426221" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1861" name="Constant_426224" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1862" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1612"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="1863" name="Constant_426296" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1864" name="Constant_426297" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1865" name="Constant_426293" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1866" name="Constant_426292" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1867" name="ScatterUpdate_426298" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1868" name="Constant_426301" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1869" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1613,1614,causal_mask.43"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1870" name="__module.model.gpt_neox.layers.7.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1871" name="__module.model.gpt_neox.layers.7.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="1601,mask_value.15" /> </output> </layer> <layer id="1872" name="__module.model.gpt_neox.layers.7.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="1615,causal_mask.45"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1873" name="Add_453771" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1874" name="ShapeOf_453772" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1875" name="Constant_402380" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1876" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1877" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1599"> <dim>1</dim> </port> </output> </layer> <layer id="1878" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1879" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1880" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1881" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="1616"> <dim>4</dim> </port> </output> </layer> <layer id="1882" name="__module.model.gpt_neox.layers.7.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1883" name="__module.model.gpt_neox.layers.7.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="1884" name="__module.model.gpt_neox.layers.7.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1885" name="__module.model.gpt_neox.layers.7.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1886" name="Maximum_453773" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1887" name="__module.model.gpt_neox.layers.7.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1618,attention_mask.23"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1888" name="__module.model.gpt_neox.layers.7.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1619,1620,sdpa_result.15,tensor.29"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1889" name="Constant_4756" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1621"> <dim>4</dim> </port> </output> </layer> <layer id="1890" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1622"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1891" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="1892" name="__module.model.gpt_neox.layers.7.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="1627,input.119"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1893" name="Constant_4774" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="3905389080" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1894" name="Convert_549161" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1895" name="Constant_4774/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="3943137816" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1896" name="Convert_549164" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1897" name="Subtract_549165" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1898" name="Constant_549166" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="3943143960" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1899" name="Constant_4774/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1900" name="__module.model.gpt_neox.layers.7.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1901" name="Constant_466865" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3791903256" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1902" name="__module.model.gpt_neox.layers.7.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1630,1632,input.121"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1903" name="__module.model.gpt_neox.layers.7/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1658"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1904" name="__module.model.gpt_neox.layers.7/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1659,1661,input.131"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1905" name="__module.model.gpt_neox.layers.8.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1906" name="__module.model.gpt_neox.layers.8.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1907" name="Constant_466866" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3943168536" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1908" name="__module.model.gpt_neox.layers.8.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1909" name="Constant_466867" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="3943193112" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1910" name="__module.model.gpt_neox.layers.8.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1796,input.139"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1911" name="Constant_5374" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="3943217688" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1912" name="Convert_549206" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1913" name="Constant_5374/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="4094212632" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1914" name="Convert_549209" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1915" name="Subtract_549210" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1916" name="Constant_549211" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="4094237208" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="1917" name="Constant_5374/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1918" name="__module.model.gpt_neox.layers.8.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1919" name="Constant_466868" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="4094335512" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1920" name="__module.model.gpt_neox.layers.8.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="1801,input.141"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1921" name="__module.model.gpt_neox.layers.8.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="1810,input.143"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1922" name="Constant_5394" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="4094433816" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1923" name="Convert_549215" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1924" name="Constant_5394/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="4245428760" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1925" name="Convert_549218" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1926" name="Subtract_549219" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1927" name="Constant_549220" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="4245434904" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="1928" name="Constant_5394/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="1929" name="__module.model.gpt_neox.layers.8.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1930" name="Constant_466869" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4245459480" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1931" name="__module.model.gpt_neox.layers.8.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1813,input.145"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1932" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1933" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1934" name="Constant_466870" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4245484056" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1935" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1936" name="Constant_466871" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4245508632" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1937" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1671,input.133"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1938" name="Constant_4853" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="4245533208" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1939" name="Convert_549188" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1940" name="Constant_4853/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="4358779416" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1941" name="Convert_549191" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1942" name="Subtract_549192" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1943" name="Constant_549193" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="4358797848" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="1944" name="Constant_4853/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="1945" name="__module.model.gpt_neox.layers.8.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1946" name="Constant_466872" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="4358871576" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1947" name="__module.model.gpt_neox.layers.8.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="1678,qkv.33"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="1948" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1949" name="__module.model.gpt_neox.layers.8.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1682,qkv.35"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="1950" name="Constant_426309" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1951" name="Constant_426312" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1952" name="Constant_426315" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1953" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1683"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1954" name="Constant_4881" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1684"> <dim>4</dim> </port> </output> </layer> <layer id="1955" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1685,query.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1956" name="Constant_426321" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1957" name="Constant_426324" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1958" name="Constant_426327" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1959" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1692,q.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1960" name="Constant_4957" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="1704,cos_cached.17"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="1961" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1962" name="Constant_426333" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1963" name="Constant_426336" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1964" name="Constant_426339" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1965" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1686"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="1966" name="Constant_4895" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1687"> <dim>4</dim> </port> </output> </layer> <layer id="1967" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1688,key.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1968" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1969" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1970" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1696" /> </output> </layer> <layer id="1971" name="Constant_535212" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="1972" name="Constant_535205" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1973" name="Constant_535207" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1974" name="Constant_535209" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1975" name="Concat_535210" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="1976" name="Broadcast_535213" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="1977" name="ReadValue_522471" type="ReadValue" version="opset6"> <data variable_id="past_key_values.8.keypresent.8.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.8.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1978" name="Constant_509590" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="1979" name="Gather_509591" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="1980" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1981" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1982" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1698" /> </output> </layer> <layer id="1983" name="__module.model.gpt_neox.layers.8.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1697,1701,1702,seq_len.33" /> </output> </layer> <layer id="1984" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1985" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1986" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="1987" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1705,1708,cos.33"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1988" name="Constant_51951" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1989" name="__module.model.gpt_neox.layers.8.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1711"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1990" name="__module.model.gpt_neox.layers.8.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1712,cos.35"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1991" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1716"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="1992" name="Constant_426344" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1993" name="Constant_426343" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="1994" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="1995" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="1996" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1717,1722" /> </output> </layer> <layer id="1997" name="__module.model.gpt_neox.layers.8.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="1998" name="__module.model.gpt_neox.layers.8.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1719,1720,1724,1725" /> </output> </layer> <layer id="1999" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2000" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2001" name="Constant_426342" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2002" name="ScatterUpdate_426345" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2003" name="Constant_426348" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2004" name="Constant_426351" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2005" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1726,x2.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2006" name="Constant_466873" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2007" name="__module.model.gpt_neox.layers.8.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1727"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2008" name="Constant_426397" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2009" name="Constant_426402" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2010" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1721,x1.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2011" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1729"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2012" name="Constant_4955" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="1703,sin_cached.17"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2013" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2014" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2015" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1706,1709,sin.33"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2016" name="Constant_51954" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2017" name="__module.model.gpt_neox.layers.8.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1714"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2018" name="__module.model.gpt_neox.layers.8.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1715,sin.35"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2019" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1730"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2020" name="__module.model.gpt_neox.layers.8.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1731,query.67"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2021" name="Constant_426449" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2022" name="Constant_426452" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2023" name="Constant_426455" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2024" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1693,query_pass.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2025" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1749,1759,query.69,query.71"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2026" name="Constant_426461" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2027" name="Constant_426464" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2028" name="Constant_426467" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2029" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1694,k.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2030" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1732"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2031" name="Constant_426471" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2032" name="Constant_426470" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2033" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2034" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2035" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1733,1738" /> </output> </layer> <layer id="2036" name="__module.model.gpt_neox.layers.8.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2037" name="__module.model.gpt_neox.layers.8.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1735,1736,1740,1741" /> </output> </layer> <layer id="2038" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2039" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2040" name="Constant_426469" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2041" name="ScatterUpdate_426472" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2042" name="Constant_426475" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2043" name="Constant_426478" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2044" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1742,x2.35"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2045" name="Constant_466874" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2046" name="__module.model.gpt_neox.layers.8.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1743"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2047" name="Constant_426524" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2048" name="Constant_426529" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2049" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1737,x1.35"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2050" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1745"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2051" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1746"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2052" name="__module.model.gpt_neox.layers.8.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1747,key.83"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2053" name="Constant_426575" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2054" name="Constant_426578" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2055" name="Constant_426581" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2056" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1695,key_pass.17"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2057" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1751,key.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2058" name="__module.model.gpt_neox.layers.8.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.8.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2059" name="Constant_535225" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2060" name="Constant_535218" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2061" name="Constant_535220" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2062" name="Constant_535222" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2063" name="Concat_535223" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2064" name="Broadcast_535226" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2065" name="ReadValue_522473" type="ReadValue" version="opset6"> <data variable_id="past_key_values.8.valuepresent.8.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.8.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2066" name="Constant_509593" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="2067" name="Gather_509594" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2068" name="Constant_426587" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2069" name="Constant_426590" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2070" name="Constant_426593" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2071" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1689"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2072" name="Constant_4909" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1690"> <dim>4</dim> </port> </output> </layer> <layer id="2073" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1691,value.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2074" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.8.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2075" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="1768"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="2076" name="Constant_426597" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2077" name="Constant_426596" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2078" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2079" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2080" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1763" /> </output> </layer> <layer id="2081" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2082" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2083" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1761" /> </output> </layer> <layer id="2084" name="__module.model.gpt_neox.layers.8.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1765,1766" /> </output> </layer> <layer id="2085" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2086" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2087" name="Constant_426595" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2088" name="ScatterUpdate_426598" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2089" name="Constant_426599" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2090" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2091" name="ScatterUpdate_426600" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2092" name="Constant_426603" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2093" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1769"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="2094" name="Constant_426675" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2095" name="Constant_426676" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2096" name="Constant_426672" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2097" name="Constant_426671" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2098" name="ScatterUpdate_426677" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2099" name="Constant_426680" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2100" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1770,1771,causal_mask.49"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2101" name="__module.model.gpt_neox.layers.8.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2102" name="__module.model.gpt_neox.layers.8.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="1758,mask_value.17" /> </output> </layer> <layer id="2103" name="__module.model.gpt_neox.layers.8.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="1772,causal_mask.51"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2104" name="Add_453847" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2105" name="ShapeOf_453848" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2106" name="Constant_402395" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2107" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2108" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1756"> <dim>1</dim> </port> </output> </layer> <layer id="2109" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2110" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2111" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2112" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="1773"> <dim>4</dim> </port> </output> </layer> <layer id="2113" name="__module.model.gpt_neox.layers.8.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2114" name="__module.model.gpt_neox.layers.8.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="2115" name="__module.model.gpt_neox.layers.8.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2116" name="__module.model.gpt_neox.layers.8.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2117" name="Maximum_453849" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2118" name="__module.model.gpt_neox.layers.8.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1775,attention_mask.25"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2119" name="__module.model.gpt_neox.layers.8.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1776,1777,sdpa_result.17,tensor.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2120" name="Constant_5334" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1778"> <dim>4</dim> </port> </output> </layer> <layer id="2121" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1779"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2122" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2123" name="__module.model.gpt_neox.layers.8.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="1784,input.135"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2124" name="Constant_5352" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="4358945304" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2125" name="Convert_549197" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2126" name="Constant_5352/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="4396694040" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2127" name="Convert_549200" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2128" name="Subtract_549201" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2129" name="Constant_549202" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="4396700184" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2130" name="Constant_5352/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2131" name="__module.model.gpt_neox.layers.8.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2132" name="Constant_466875" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4245459480" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2133" name="__module.model.gpt_neox.layers.8.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1787,1789,input.137"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2134" name="__module.model.gpt_neox.layers.8/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1815"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2135" name="__module.model.gpt_neox.layers.8/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1816,1818,input.147"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2136" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2137" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2138" name="Constant_466880" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4396724760" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2139" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2140" name="Constant_466881" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4396749336" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2141" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1828,input.149"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2142" name="Constant_5431" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="4396773912" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2143" name="Convert_549224" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2144" name="Constant_5431/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="4510020120" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2145" name="Convert_549227" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2146" name="Subtract_549228" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2147" name="Constant_549229" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="4510038552" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2148" name="Constant_5431/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2149" name="__module.model.gpt_neox.layers.9.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2150" name="Constant_466882" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="4510112280" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2151" name="__module.model.gpt_neox.layers.9.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="1835,qkv.37"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2152" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2153" name="__module.model.gpt_neox.layers.9.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1839,qkv.39"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="2154" name="Constant_426966" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2155" name="Constant_426969" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2156" name="Constant_426972" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2157" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1846"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2158" name="Constant_5487" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1847"> <dim>4</dim> </port> </output> </layer> <layer id="2159" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1848,value.37"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2160" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.9.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2162" name="Constant_535238" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2163" name="Constant_535231" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2164" name="Constant_535233" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2165" name="Constant_535235" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2166" name="Concat_535236" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2167" name="Broadcast_535239" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2168" name="ReadValue_522475" type="ReadValue" version="opset6"> <data variable_id="past_key_values.9.keypresent.9.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.9.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2169" name="Constant_509596" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="2170" name="Gather_509597" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2171" name="Constant_426712" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2172" name="Constant_426715" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2173" name="Constant_426718" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2174" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1843"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2175" name="Constant_5473" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1844"> <dim>4</dim> </port> </output> </layer> <layer id="2176" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1845,key.91"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2177" name="Constant_426840" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2178" name="Constant_426843" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2179" name="Constant_426846" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2180" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1851,k.19"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2181" name="Constant_5535" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="1861,cos_cached.19"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2182" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2183" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2184" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2185" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1853" /> </output> </layer> <layer id="2186" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2187" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2188" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1855" /> </output> </layer> <layer id="2189" name="__module.model.gpt_neox.layers.9.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1854,1858,1859,seq_len.37" /> </output> </layer> <layer id="2190" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2191" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2192" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2193" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1862,1865,cos.37"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2194" name="Constant_51957" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2195" name="__module.model.gpt_neox.layers.9.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1868"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2196" name="__module.model.gpt_neox.layers.9.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1869,cos.39"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2197" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1889"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2198" name="Constant_426850" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2199" name="Constant_426849" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2200" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2201" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2202" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1890,1895" /> </output> </layer> <layer id="2203" name="__module.model.gpt_neox.layers.9.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2204" name="__module.model.gpt_neox.layers.9.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1892,1893,1897,1898" /> </output> </layer> <layer id="2205" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2206" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2207" name="Constant_426848" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2208" name="ScatterUpdate_426851" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2209" name="Constant_426854" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2210" name="Constant_426857" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2211" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1899,x2.39"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2212" name="Constant_466884" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2213" name="__module.model.gpt_neox.layers.9.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1900"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2214" name="Constant_426903" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2215" name="Constant_426908" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2216" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1894,x1.39"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2217" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1902"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2218" name="Constant_5533" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="1860,sin_cached.19"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2219" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2220" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2221" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1863,1866,sin.37"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2222" name="Constant_51960" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2223" name="__module.model.gpt_neox.layers.9.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="1871"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2224" name="__module.model.gpt_neox.layers.9.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="1872,sin.39"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2225" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1903"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2226" name="__module.model.gpt_neox.layers.9.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1904,key.93"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2227" name="Constant_426954" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2228" name="Constant_426957" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2229" name="Constant_426960" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2230" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1852,key_pass.19"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2231" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1908,key.95"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2232" name="__module.model.gpt_neox.layers.9.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.9.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2242" name="Constant_535251" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2243" name="Constant_535244" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2244" name="Constant_535246" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2245" name="Constant_535248" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2246" name="Concat_535249" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2247" name="Broadcast_535252" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2248" name="ReadValue_522457" type="ReadValue" version="opset6"> <data variable_id="past_key_values.43.valuepresent.43.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.43.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2249" name="Constant_509803" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="2250" name="Gather_509804" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2251" name="__module.model.gpt_neox.layers.9.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2252" name="__module.model.gpt_neox.layers.9.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2253" name="Constant_466876" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4510186008" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2254" name="__module.model.gpt_neox.layers.9.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2255" name="Constant_466877" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4510210584" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2256" name="__module.model.gpt_neox.layers.9.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1953,input.155"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2257" name="Constant_5952" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="4510235160" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2258" name="Convert_549242" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2259" name="Constant_5952/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="4661230104" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2260" name="Convert_549245" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2261" name="Subtract_549246" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2262" name="Constant_549247" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="4661254680" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2263" name="Constant_5952/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2264" name="__module.model.gpt_neox.layers.9.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2265" name="Constant_466878" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="4661352984" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2266" name="__module.model.gpt_neox.layers.9.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="1958,input.157"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2267" name="__module.model.gpt_neox.layers.9.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="1967,input.159"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2268" name="Constant_5972" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="4661451288" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2269" name="Convert_549251" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2270" name="Constant_5972/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="4812446232" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2271" name="Convert_549254" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2272" name="Subtract_549255" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2273" name="Constant_549256" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="4812452376" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2274" name="Constant_5972/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2275" name="__module.model.gpt_neox.layers.9.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2276" name="Constant_466879" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4812476952" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2277" name="__module.model.gpt_neox.layers.9.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1970,input.161"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2278" name="Constant_426688" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2279" name="Constant_426691" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2280" name="Constant_426694" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2281" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1840"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2282" name="Constant_5459" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1841"> <dim>4</dim> </port> </output> </layer> <layer id="2283" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1842,query.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2284" name="Constant_426700" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2285" name="Constant_426703" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2286" name="Constant_426706" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2287" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1849,q.19"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2288" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1873"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2289" name="Constant_426723" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2290" name="Constant_426722" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2291" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2292" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2293" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1874,1879" /> </output> </layer> <layer id="2294" name="__module.model.gpt_neox.layers.9.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2295" name="__module.model.gpt_neox.layers.9.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="1876,1877,1881,1882" /> </output> </layer> <layer id="2296" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2297" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2298" name="Constant_426721" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2299" name="ScatterUpdate_426724" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2300" name="Constant_426727" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2301" name="Constant_426730" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2302" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1883,x2.37"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2303" name="Constant_466883" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2304" name="__module.model.gpt_neox.layers.9.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="1884"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2305" name="Constant_426776" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2306" name="Constant_426781" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2307" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1878,x1.37"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2308" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="1886"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2309" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1887"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2310" name="__module.model.gpt_neox.layers.9.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="1888,query.75"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2311" name="Constant_426828" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2312" name="Constant_426831" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2313" name="Constant_426834" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2314" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1850,query_pass.19"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2315" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="1906,1916,query.77,query.79"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2316" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="1925"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="2317" name="Constant_426976" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2318" name="Constant_426975" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2319" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2320" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2321" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1920" /> </output> </layer> <layer id="2322" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2323" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2324" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1918" /> </output> </layer> <layer id="2325" name="__module.model.gpt_neox.layers.9.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="1922,1923" /> </output> </layer> <layer id="2326" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2327" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2328" name="Constant_426974" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2329" name="ScatterUpdate_426977" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2330" name="Constant_426978" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2331" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2332" name="ScatterUpdate_426979" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2333" name="Constant_426982" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2334" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1926"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="2335" name="Constant_427054" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2336" name="Constant_427055" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2337" name="Constant_427051" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2338" name="Constant_427050" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2339" name="ScatterUpdate_427056" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2340" name="Constant_427059" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2341" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="1927,1928,causal_mask.55"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2342" name="__module.model.gpt_neox.layers.9.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2343" name="__module.model.gpt_neox.layers.9.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="1915,mask_value.19" /> </output> </layer> <layer id="2344" name="__module.model.gpt_neox.layers.9.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="1929,causal_mask.57"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2345" name="Add_453923" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2346" name="ShapeOf_453924" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2347" name="Constant_402410" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2348" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2349" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="1913"> <dim>1</dim> </port> </output> </layer> <layer id="2350" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2351" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2352" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2353" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="1930"> <dim>4</dim> </port> </output> </layer> <layer id="2354" name="__module.model.gpt_neox.layers.9.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2355" name="__module.model.gpt_neox.layers.9.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="2356" name="__module.model.gpt_neox.layers.9.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2357" name="__module.model.gpt_neox.layers.9.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2358" name="Maximum_453925" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2359" name="__module.model.gpt_neox.layers.9.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1932,attention_mask.27"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2360" name="__module.model.gpt_neox.layers.9.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="1933,1934,sdpa_result.19,tensor.37"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2361" name="Constant_5912" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1935"> <dim>4</dim> </port> </output> </layer> <layer id="2362" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1936"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2363" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2364" name="__module.model.gpt_neox.layers.9.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="1941,input.151"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2365" name="Constant_5930" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="4812501528" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2366" name="Convert_549233" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2367" name="Constant_5930/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="4850250264" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2368" name="Convert_549236" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2369" name="Subtract_549237" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2370" name="Constant_549238" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="4850256408" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2371" name="Constant_5930/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2372" name="__module.model.gpt_neox.layers.9.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2373" name="Constant_466885" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4812476952" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2374" name="__module.model.gpt_neox.layers.9.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1944,1946,input.153"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2375" name="__module.model.gpt_neox.layers.9/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1972"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2376" name="__module.model.gpt_neox.layers.9/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1973,1975,input.163"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2377" name="__module.model.gpt_neox.layers.10.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2378" name="__module.model.gpt_neox.layers.10.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2379" name="Constant_466886" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4850280984" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2380" name="__module.model.gpt_neox.layers.10.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2381" name="Constant_466887" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="4850305560" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2382" name="__module.model.gpt_neox.layers.10.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2110,input.171"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2383" name="Constant_6530" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="4850330136" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2384" name="Convert_549278" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2385" name="Constant_6530/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="5001325080" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2386" name="Convert_549281" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2387" name="Subtract_549282" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2388" name="Constant_549283" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="5001349656" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2389" name="Constant_6530/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2390" name="__module.model.gpt_neox.layers.10.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2391" name="Constant_466888" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="5001447960" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2392" name="__module.model.gpt_neox.layers.10.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="2115,input.173"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2393" name="__module.model.gpt_neox.layers.10.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="2124,input.175"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2394" name="Constant_6550" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="5001546264" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2395" name="Convert_549287" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2396" name="Constant_6550/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="5152541208" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2397" name="Convert_549290" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2398" name="Subtract_549291" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2399" name="Constant_549292" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="5152547352" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2400" name="Constant_6550/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2401" name="__module.model.gpt_neox.layers.10.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2402" name="Constant_466889" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5152571928" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2403" name="__module.model.gpt_neox.layers.10.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2127,input.177"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2404" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2405" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2406" name="Constant_466890" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5152596504" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2407" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2408" name="Constant_466891" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5152621080" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2409" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="1985,input.165"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2410" name="Constant_6009" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="5152645656" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2411" name="Convert_549260" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2412" name="Constant_6009/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="5265891864" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2413" name="Convert_549263" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2414" name="Subtract_549264" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2415" name="Constant_549265" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="5265910296" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2416" name="Constant_6009/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2417" name="__module.model.gpt_neox.layers.10.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2418" name="Constant_466892" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="5265984024" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2419" name="__module.model.gpt_neox.layers.10.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="1992,qkv.41"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2420" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2421" name="__module.model.gpt_neox.layers.10.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1996,qkv.43"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="2422" name="Constant_427067" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2423" name="Constant_427070" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2424" name="Constant_427073" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2425" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="1997"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2426" name="Constant_6037" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="1998"> <dim>4</dim> </port> </output> </layer> <layer id="2427" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="1999,query.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2428" name="Constant_427079" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2429" name="Constant_427082" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2430" name="Constant_427085" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2431" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2006,q.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2432" name="Constant_6113" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="2018,cos_cached.21"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2433" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2434" name="Constant_427091" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2435" name="Constant_427094" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2436" name="Constant_427097" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2437" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2000"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2438" name="Constant_6051" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2001"> <dim>4</dim> </port> </output> </layer> <layer id="2439" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2002,key.101"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2440" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2441" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2442" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2010" /> </output> </layer> <layer id="2443" name="Constant_535264" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2444" name="Constant_535257" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2445" name="Constant_535259" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2446" name="Constant_535261" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2447" name="Concat_535262" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2448" name="Broadcast_535265" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2449" name="ReadValue_522311" type="ReadValue" version="opset6"> <data variable_id="past_key_values.10.keypresent.10.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.10.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2450" name="Constant_509602" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="2451" name="Gather_509603" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2452" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2453" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2454" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2012" /> </output> </layer> <layer id="2455" name="__module.model.gpt_neox.layers.10.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2011,2015,2016,seq_len.41" /> </output> </layer> <layer id="2456" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2457" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2458" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2459" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2019,2022,cos.41"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2460" name="Constant_51963" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2461" name="__module.model.gpt_neox.layers.10.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2025"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2462" name="__module.model.gpt_neox.layers.10.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2026,cos.43"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2463" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2030"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2464" name="Constant_427102" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2465" name="Constant_427101" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2466" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2467" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2468" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2031,2036" /> </output> </layer> <layer id="2469" name="__module.model.gpt_neox.layers.10.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2470" name="__module.model.gpt_neox.layers.10.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2033,2034,2038,2039" /> </output> </layer> <layer id="2471" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2472" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2473" name="Constant_427100" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2474" name="ScatterUpdate_427103" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2475" name="Constant_427106" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2476" name="Constant_427109" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2477" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2040,x2.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2478" name="Constant_466893" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2479" name="__module.model.gpt_neox.layers.10.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2041"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2480" name="Constant_427155" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2481" name="Constant_427160" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2482" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2035,x1.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2483" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2043"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2484" name="Constant_6111" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="2017,sin_cached.21"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2485" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2486" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2487" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2020,2023,sin.41"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2488" name="Constant_51966" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2489" name="__module.model.gpt_neox.layers.10.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2028"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2490" name="__module.model.gpt_neox.layers.10.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2029,sin.43"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2491" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2044"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2492" name="__module.model.gpt_neox.layers.10.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2045,query.83"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2493" name="Constant_427207" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2494" name="Constant_427210" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2495" name="Constant_427213" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2496" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2007,query_pass.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2497" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2063,2073,query.85,query.87"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2498" name="Constant_427219" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2499" name="Constant_427222" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2500" name="Constant_427225" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2501" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2008,k.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2502" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2046"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2503" name="Constant_427229" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2504" name="Constant_427228" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2505" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2506" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2507" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2047,2052" /> </output> </layer> <layer id="2508" name="__module.model.gpt_neox.layers.10.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2509" name="__module.model.gpt_neox.layers.10.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2049,2050,2054,2055" /> </output> </layer> <layer id="2510" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2511" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2512" name="Constant_427227" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2513" name="ScatterUpdate_427230" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2514" name="Constant_427233" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2515" name="Constant_427236" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2516" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2056,x2.43"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2517" name="Constant_466894" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2518" name="__module.model.gpt_neox.layers.10.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2057"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2519" name="Constant_427282" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2520" name="Constant_427287" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2521" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2051,x1.43"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2522" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2059"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2523" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2060"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2524" name="__module.model.gpt_neox.layers.10.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2061,key.103"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2525" name="Constant_427333" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2526" name="Constant_427336" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2527" name="Constant_427339" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2528" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2009,key_pass.21"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2529" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2065,key.105"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2530" name="__module.model.gpt_neox.layers.10.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.10.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2531" name="Constant_535277" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2532" name="Constant_535270" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2533" name="Constant_535272" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2534" name="Constant_535274" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2535" name="Concat_535275" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2536" name="Broadcast_535278" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2537" name="ReadValue_522313" type="ReadValue" version="opset6"> <data variable_id="past_key_values.10.valuepresent.10.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.10.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2538" name="Constant_509605" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="2539" name="Gather_509606" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2540" name="Constant_427345" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2541" name="Constant_427348" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2542" name="Constant_427351" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2543" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2003"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2544" name="Constant_6065" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2004"> <dim>4</dim> </port> </output> </layer> <layer id="2545" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2005,value.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2546" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.10.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2547" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="2082"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="2548" name="Constant_427355" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2549" name="Constant_427354" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2550" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2551" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2552" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2077" /> </output> </layer> <layer id="2553" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2554" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2555" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2075" /> </output> </layer> <layer id="2556" name="__module.model.gpt_neox.layers.10.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2079,2080" /> </output> </layer> <layer id="2557" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2558" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2559" name="Constant_427353" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2560" name="ScatterUpdate_427356" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2561" name="Constant_427357" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2562" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2563" name="ScatterUpdate_427358" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2564" name="Constant_427361" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2565" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2083"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="2566" name="Constant_427433" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2567" name="Constant_427434" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2568" name="Constant_427430" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2569" name="Constant_427429" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2570" name="ScatterUpdate_427435" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2571" name="Constant_427438" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2572" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2084,2085,causal_mask.61"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2573" name="__module.model.gpt_neox.layers.10.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2574" name="__module.model.gpt_neox.layers.10.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="2072,mask_value.21" /> </output> </layer> <layer id="2575" name="__module.model.gpt_neox.layers.10.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="2086,causal_mask.63"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2576" name="Add_453999" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2577" name="ShapeOf_454000" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2578" name="Constant_402425" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2579" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2580" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2070"> <dim>1</dim> </port> </output> </layer> <layer id="2581" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2582" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2583" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2584" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="2087"> <dim>4</dim> </port> </output> </layer> <layer id="2585" name="__module.model.gpt_neox.layers.10.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2586" name="__module.model.gpt_neox.layers.10.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="2587" name="__module.model.gpt_neox.layers.10.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2588" name="__module.model.gpt_neox.layers.10.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2589" name="Maximum_454001" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2590" name="__module.model.gpt_neox.layers.10.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2089,attention_mask.29"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2591" name="__module.model.gpt_neox.layers.10.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2090,2091,sdpa_result.21,tensor.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2592" name="Constant_6490" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2092"> <dim>4</dim> </port> </output> </layer> <layer id="2593" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2093"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2594" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2595" name="__module.model.gpt_neox.layers.10.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="2098,input.167"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2596" name="Constant_6508" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="5266057752" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2597" name="Convert_549269" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2598" name="Constant_6508/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="5303806488" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2599" name="Convert_549272" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2600" name="Subtract_549273" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2601" name="Constant_549274" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="5303812632" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2602" name="Constant_6508/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2603" name="__module.model.gpt_neox.layers.10.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2604" name="Constant_466895" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5303837208" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2605" name="__module.model.gpt_neox.layers.10.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2101,2103,input.169"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2606" name="__module.model.gpt_neox.layers.10/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2129"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2607" name="__module.model.gpt_neox.layers.10/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2130,2132,input.179"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2608" name="__module.model.gpt_neox.layers.11.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2609" name="__module.model.gpt_neox.layers.11.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2610" name="Constant_466896" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5303861784" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2611" name="__module.model.gpt_neox.layers.11.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2612" name="Constant_466897" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5303886360" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2613" name="__module.model.gpt_neox.layers.11.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2267,input.187"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2614" name="Constant_7108" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="5303910936" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2615" name="Convert_549314" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2616" name="Constant_7108/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="5454905880" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2617" name="Convert_549317" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2618" name="Subtract_549318" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2619" name="Constant_549319" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="5454930456" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2620" name="Constant_7108/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2621" name="__module.model.gpt_neox.layers.11.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2622" name="Constant_466898" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="5455028760" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2623" name="__module.model.gpt_neox.layers.11.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="2272,input.189"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2624" name="__module.model.gpt_neox.layers.11.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="2281,input.191"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2625" name="Constant_7128" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="5455127064" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2626" name="Convert_549323" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2627" name="Constant_7128/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="5606122008" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2628" name="Convert_549326" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2629" name="Subtract_549327" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2630" name="Constant_549328" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="5606128152" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2631" name="Constant_7128/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2632" name="__module.model.gpt_neox.layers.11.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2633" name="Constant_466899" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5606152728" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2634" name="__module.model.gpt_neox.layers.11.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2284,input.193"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2635" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2636" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2637" name="Constant_466900" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5606177304" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2638" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2639" name="Constant_466901" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5606201880" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2640" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2142,input.181"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2641" name="Constant_6587" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="5606226456" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2642" name="Convert_549296" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2643" name="Constant_6587/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="5719472664" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2644" name="Convert_549299" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2645" name="Subtract_549300" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2646" name="Constant_549301" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="5719491096" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2647" name="Constant_6587/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2648" name="__module.model.gpt_neox.layers.11.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2649" name="Constant_466902" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="5719564824" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2650" name="__module.model.gpt_neox.layers.11.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="2149,qkv.45"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2651" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2652" name="__module.model.gpt_neox.layers.11.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2153,qkv.47"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="2653" name="Constant_427446" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2654" name="Constant_427449" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2655" name="Constant_427452" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2656" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2154"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2657" name="Constant_6615" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2155"> <dim>4</dim> </port> </output> </layer> <layer id="2658" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2156,query.89"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2659" name="Constant_427458" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2660" name="Constant_427461" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2661" name="Constant_427464" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2662" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2163,q.23"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2663" name="Constant_6691" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="2175,cos_cached.23"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2664" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2665" name="Constant_427470" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2666" name="Constant_427473" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2667" name="Constant_427476" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2668" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2157"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2669" name="Constant_6629" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2158"> <dim>4</dim> </port> </output> </layer> <layer id="2670" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2159,key.111"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2671" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2672" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2673" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2167" /> </output> </layer> <layer id="2674" name="Constant_535290" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2675" name="Constant_535283" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2676" name="Constant_535285" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2677" name="Constant_535287" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2678" name="Concat_535288" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2679" name="Broadcast_535291" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2680" name="ReadValue_522315" type="ReadValue" version="opset6"> <data variable_id="past_key_values.11.keypresent.11.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.11.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2681" name="Constant_509608" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="2682" name="Gather_509609" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2683" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2684" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2685" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2169" /> </output> </layer> <layer id="2686" name="__module.model.gpt_neox.layers.11.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2168,2172,2173,seq_len.45" /> </output> </layer> <layer id="2687" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2688" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2689" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2690" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2176,2179,cos.45"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2691" name="Constant_51969" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2692" name="__module.model.gpt_neox.layers.11.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2182"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2693" name="__module.model.gpt_neox.layers.11.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2183,cos.47"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2694" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2187"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2695" name="Constant_427481" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2696" name="Constant_427480" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2697" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2698" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2699" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2188,2193" /> </output> </layer> <layer id="2700" name="__module.model.gpt_neox.layers.11.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2701" name="__module.model.gpt_neox.layers.11.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2190,2191,2195,2196" /> </output> </layer> <layer id="2702" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2703" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2704" name="Constant_427479" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2705" name="ScatterUpdate_427482" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2706" name="Constant_427485" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2707" name="Constant_427488" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2708" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2197,x2.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2709" name="Constant_466903" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2710" name="__module.model.gpt_neox.layers.11.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2198"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2711" name="Constant_427534" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2712" name="Constant_427539" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2713" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2192,x1.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2714" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2200"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2715" name="Constant_6689" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="2174,sin_cached.23"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2716" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2717" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2718" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2177,2180,sin.45"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2719" name="Constant_51972" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2720" name="__module.model.gpt_neox.layers.11.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2185"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2721" name="__module.model.gpt_neox.layers.11.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2186,sin.47"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2722" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2201"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2723" name="__module.model.gpt_neox.layers.11.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2202,query.91"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2724" name="Constant_427586" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2725" name="Constant_427589" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2726" name="Constant_427592" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2727" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2164,query_pass.23"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2728" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2220,2230,query.93,query.95"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2729" name="Constant_427598" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2730" name="Constant_427601" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2731" name="Constant_427604" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2732" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2165,k.23"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2733" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2203"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2734" name="Constant_427608" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2735" name="Constant_427607" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2736" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2737" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2738" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2204,2209" /> </output> </layer> <layer id="2739" name="__module.model.gpt_neox.layers.11.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2740" name="__module.model.gpt_neox.layers.11.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2206,2207,2211,2212" /> </output> </layer> <layer id="2741" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2742" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2743" name="Constant_427606" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2744" name="ScatterUpdate_427609" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2745" name="Constant_427612" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2746" name="Constant_427615" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2747" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2213,x2.47"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2748" name="Constant_466904" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2749" name="__module.model.gpt_neox.layers.11.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2214"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2750" name="Constant_427661" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2751" name="Constant_427666" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2752" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2208,x1.47"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2753" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2216"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2754" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2217"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2755" name="__module.model.gpt_neox.layers.11.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2218,key.113"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2756" name="Constant_427712" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2757" name="Constant_427715" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2758" name="Constant_427718" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2759" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2166,key_pass.23"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2760" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2222,key.115"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2761" name="__module.model.gpt_neox.layers.11.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.11.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2762" name="Constant_535303" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2763" name="Constant_535296" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2764" name="Constant_535298" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2765" name="Constant_535300" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2766" name="Concat_535301" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2767" name="Broadcast_535304" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2768" name="ReadValue_522317" type="ReadValue" version="opset6"> <data variable_id="past_key_values.11.valuepresent.11.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.11.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2769" name="Constant_509611" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="2770" name="Gather_509612" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2771" name="Constant_427724" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2772" name="Constant_427727" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2773" name="Constant_427730" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2774" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2160"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2775" name="Constant_6643" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2161"> <dim>4</dim> </port> </output> </layer> <layer id="2776" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2162,value.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2777" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.11.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2778" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="2239"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="2779" name="Constant_427734" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2780" name="Constant_427733" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2781" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2782" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2783" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2234" /> </output> </layer> <layer id="2784" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2785" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2786" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2232" /> </output> </layer> <layer id="2787" name="__module.model.gpt_neox.layers.11.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2236,2237" /> </output> </layer> <layer id="2788" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2789" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2790" name="Constant_427732" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2791" name="ScatterUpdate_427735" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2792" name="Constant_427736" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2793" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2794" name="ScatterUpdate_427737" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2795" name="Constant_427740" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2796" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2240"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="2797" name="Constant_427812" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2798" name="Constant_427813" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2799" name="Constant_427809" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2800" name="Constant_427808" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2801" name="ScatterUpdate_427814" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2802" name="Constant_427817" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2803" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2241,2242,causal_mask.67"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2804" name="__module.model.gpt_neox.layers.11.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2805" name="__module.model.gpt_neox.layers.11.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="2229,mask_value.23" /> </output> </layer> <layer id="2806" name="__module.model.gpt_neox.layers.11.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="2243,causal_mask.69"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2807" name="Add_454075" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2808" name="ShapeOf_454076" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2809" name="Constant_402440" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2810" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2811" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2227"> <dim>1</dim> </port> </output> </layer> <layer id="2812" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2813" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2814" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2815" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="2244"> <dim>4</dim> </port> </output> </layer> <layer id="2816" name="__module.model.gpt_neox.layers.11.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2817" name="__module.model.gpt_neox.layers.11.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="2818" name="__module.model.gpt_neox.layers.11.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2819" name="__module.model.gpt_neox.layers.11.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2820" name="Maximum_454077" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2821" name="__module.model.gpt_neox.layers.11.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2246,attention_mask.31"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2822" name="__module.model.gpt_neox.layers.11.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2247,2248,sdpa_result.23,tensor.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2823" name="Constant_7068" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2249"> <dim>4</dim> </port> </output> </layer> <layer id="2824" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2250"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2825" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="2826" name="__module.model.gpt_neox.layers.11.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="2255,input.183"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2827" name="Constant_7086" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="5719638552" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2828" name="Convert_549305" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2829" name="Constant_7086/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="5757387288" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2830" name="Convert_549308" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2831" name="Subtract_549309" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2832" name="Constant_549310" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="5757393432" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2833" name="Constant_7086/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2834" name="__module.model.gpt_neox.layers.11.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2835" name="Constant_466905" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5606152728" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2836" name="__module.model.gpt_neox.layers.11.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2258,2260,input.185"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2837" name="__module.model.gpt_neox.layers.11/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2286"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2838" name="__module.model.gpt_neox.layers.11/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2287,2289,input.195"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2839" name="__module.model.gpt_neox.layers.12.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2840" name="__module.model.gpt_neox.layers.12.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2841" name="Constant_466906" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5757418008" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2842" name="__module.model.gpt_neox.layers.12.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2843" name="Constant_466907" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="5757442584" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2844" name="__module.model.gpt_neox.layers.12.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2424,input.203"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2845" name="Constant_7686" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="5757467160" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2846" name="Convert_549350" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2847" name="Constant_7686/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="5908462104" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2848" name="Convert_549353" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2849" name="Subtract_549354" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2850" name="Constant_549355" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="5908486680" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="2851" name="Constant_7686/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2852" name="__module.model.gpt_neox.layers.12.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2853" name="Constant_466908" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="5908584984" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2854" name="__module.model.gpt_neox.layers.12.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="2429,input.205"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2855" name="__module.model.gpt_neox.layers.12.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="2438,input.207"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2856" name="Constant_7706" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="5908683288" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2857" name="Convert_549359" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2858" name="Constant_7706/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="6059678232" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2859" name="Convert_549362" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2860" name="Subtract_549363" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2861" name="Constant_549364" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="6059684376" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="2862" name="Constant_7706/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="2863" name="__module.model.gpt_neox.layers.12.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2864" name="Constant_466909" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6059708952" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2865" name="__module.model.gpt_neox.layers.12.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2441,input.209"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2866" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2867" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2868" name="Constant_466910" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6059733528" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2869" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2870" name="Constant_466911" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6059758104" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2871" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2299,input.197"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2872" name="Constant_7165" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="6059782680" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2873" name="Convert_549332" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2874" name="Constant_7165/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="6173028888" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2875" name="Convert_549335" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2876" name="Subtract_549336" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2877" name="Constant_549337" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="6173047320" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="2878" name="Constant_7165/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="2879" name="__module.model.gpt_neox.layers.12.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2880" name="Constant_466912" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="6173121048" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2881" name="__module.model.gpt_neox.layers.12.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="2306,qkv.49"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="2882" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2883" name="__module.model.gpt_neox.layers.12.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2310,qkv.51"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="2884" name="Constant_427825" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2885" name="Constant_427828" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2886" name="Constant_427831" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2887" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2311"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2888" name="Constant_7193" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2312"> <dim>4</dim> </port> </output> </layer> <layer id="2889" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2313,query.97"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2890" name="Constant_427837" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2891" name="Constant_427840" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2892" name="Constant_427843" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2893" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2320,q.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2894" name="Constant_7269" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="2332,cos_cached.25"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2895" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2896" name="Constant_427849" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2897" name="Constant_427852" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2898" name="Constant_427855" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2899" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2314"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="2900" name="Constant_7207" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2315"> <dim>4</dim> </port> </output> </layer> <layer id="2901" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2316,key.121"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2902" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2903" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2904" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2324" /> </output> </layer> <layer id="2905" name="Constant_535316" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2906" name="Constant_535309" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2907" name="Constant_535311" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2908" name="Constant_535313" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2909" name="Concat_535314" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2910" name="Broadcast_535317" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2911" name="ReadValue_522319" type="ReadValue" version="opset6"> <data variable_id="past_key_values.12.keypresent.12.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.12.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2912" name="Constant_509614" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="2913" name="Gather_509615" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2914" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2915" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2916" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2326" /> </output> </layer> <layer id="2917" name="__module.model.gpt_neox.layers.12.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2325,2329,2330,seq_len.49" /> </output> </layer> <layer id="2918" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2919" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2920" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2921" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2333,2336,cos.49"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2922" name="Constant_51975" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2923" name="__module.model.gpt_neox.layers.12.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2339"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2924" name="__module.model.gpt_neox.layers.12.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2340,cos.51"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2925" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2344"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2926" name="Constant_427860" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2927" name="Constant_427859" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2928" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2929" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2930" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2345,2350" /> </output> </layer> <layer id="2931" name="__module.model.gpt_neox.layers.12.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2932" name="__module.model.gpt_neox.layers.12.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2347,2348,2352,2353" /> </output> </layer> <layer id="2933" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2934" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2935" name="Constant_427858" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2936" name="ScatterUpdate_427861" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2937" name="Constant_427864" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2938" name="Constant_427867" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2939" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2354,x2.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2940" name="Constant_466913" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2941" name="__module.model.gpt_neox.layers.12.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2355"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2942" name="Constant_427913" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2943" name="Constant_427918" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2944" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2349,x1.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2945" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2357"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2946" name="Constant_7267" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="2331,sin_cached.25"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="2947" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2948" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2949" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2334,2337,sin.49"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2950" name="Constant_51978" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2951" name="__module.model.gpt_neox.layers.12.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2342"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2952" name="__module.model.gpt_neox.layers.12.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2343,sin.51"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2953" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2358"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2954" name="__module.model.gpt_neox.layers.12.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2359,query.99"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2955" name="Constant_427965" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2956" name="Constant_427968" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2957" name="Constant_427971" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2958" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2321,query_pass.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2959" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2377,2387,query.101,query.103"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2960" name="Constant_427977" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2961" name="Constant_427980" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2962" name="Constant_427983" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2963" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2322,k.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2964" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2360"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2965" name="Constant_427987" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2966" name="Constant_427986" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2967" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2968" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2969" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2361,2366" /> </output> </layer> <layer id="2970" name="__module.model.gpt_neox.layers.12.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="2971" name="__module.model.gpt_neox.layers.12.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2363,2364,2368,2369" /> </output> </layer> <layer id="2972" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="2973" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2974" name="Constant_427985" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="2975" name="ScatterUpdate_427988" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2976" name="Constant_427991" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2977" name="Constant_427994" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2978" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2370,x2.51"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2979" name="Constant_466914" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="2980" name="__module.model.gpt_neox.layers.12.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2371"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2981" name="Constant_428040" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2982" name="Constant_428045" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2983" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2365,x1.51"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="2984" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2373"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2985" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2374"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2986" name="__module.model.gpt_neox.layers.12.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2375,key.123"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="2987" name="Constant_428091" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2988" name="Constant_428094" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2989" name="Constant_428097" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="2990" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2323,key_pass.25"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="2991" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2379,key.125"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2992" name="__module.model.gpt_neox.layers.12.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.12.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2993" name="Constant_535329" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="2994" name="Constant_535322" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2995" name="Constant_535324" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2996" name="Constant_535326" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="2997" name="Concat_535327" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="2998" name="Broadcast_535330" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="2999" name="ReadValue_522321" type="ReadValue" version="opset6"> <data variable_id="past_key_values.12.valuepresent.12.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.12.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3000" name="Constant_509617" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3001" name="Gather_509618" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3002" name="Constant_428103" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3003" name="Constant_428106" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3004" name="Constant_428109" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3005" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2317"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3006" name="Constant_7221" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2318"> <dim>4</dim> </port> </output> </layer> <layer id="3007" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2319,value.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3008" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.12.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3009" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="2396"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3010" name="Constant_428113" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3011" name="Constant_428112" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3012" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3013" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3014" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2391" /> </output> </layer> <layer id="3015" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3016" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3017" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2389" /> </output> </layer> <layer id="3018" name="__module.model.gpt_neox.layers.12.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2393,2394" /> </output> </layer> <layer id="3019" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3020" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3021" name="Constant_428111" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3022" name="ScatterUpdate_428114" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3023" name="Constant_428115" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3024" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3025" name="ScatterUpdate_428116" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3026" name="Constant_428119" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3027" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2397"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3028" name="Constant_428191" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3029" name="Constant_428192" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3030" name="Constant_428188" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3031" name="Constant_428187" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3032" name="ScatterUpdate_428193" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3033" name="Constant_428196" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3034" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2398,2399,causal_mask.73"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3035" name="__module.model.gpt_neox.layers.12.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3036" name="__module.model.gpt_neox.layers.12.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="2386,mask_value.25" /> </output> </layer> <layer id="3037" name="__module.model.gpt_neox.layers.12.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="2400,causal_mask.75"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3038" name="Add_454151" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3039" name="ShapeOf_454152" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3040" name="Constant_402455" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3041" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3042" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2384"> <dim>1</dim> </port> </output> </layer> <layer id="3043" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3044" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3045" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3046" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="2401"> <dim>4</dim> </port> </output> </layer> <layer id="3047" name="__module.model.gpt_neox.layers.12.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3048" name="__module.model.gpt_neox.layers.12.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="3049" name="__module.model.gpt_neox.layers.12.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3050" name="__module.model.gpt_neox.layers.12.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3051" name="Maximum_454153" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3052" name="__module.model.gpt_neox.layers.12.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2403,attention_mask.33"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3053" name="__module.model.gpt_neox.layers.12.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2404,2405,sdpa_result.25,tensor.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3054" name="Constant_7646" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2406"> <dim>4</dim> </port> </output> </layer> <layer id="3055" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2407"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3056" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3057" name="__module.model.gpt_neox.layers.12.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="2412,input.199"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3058" name="Constant_7664" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="6173194776" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3059" name="Convert_549341" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3060" name="Constant_7664/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="6210943512" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3061" name="Convert_549344" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3062" name="Subtract_549345" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3063" name="Constant_549346" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="6210949656" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3064" name="Constant_7664/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3065" name="__module.model.gpt_neox.layers.12.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3066" name="Constant_466915" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6059708952" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3067" name="__module.model.gpt_neox.layers.12.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2415,2417,input.201"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3068" name="__module.model.gpt_neox.layers.12/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2443"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3069" name="__module.model.gpt_neox.layers.12/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2444,2446,input.211"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3070" name="__module.model.gpt_neox.layers.13.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3071" name="__module.model.gpt_neox.layers.13.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3072" name="Constant_466916" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6210974232" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3073" name="__module.model.gpt_neox.layers.13.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3074" name="Constant_466917" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6210998808" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3075" name="__module.model.gpt_neox.layers.13.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2581,input.219"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3076" name="Constant_8264" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="6211023384" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3077" name="Convert_549386" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3078" name="Constant_8264/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="6362018328" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3079" name="Convert_549389" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3080" name="Subtract_549390" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3081" name="Constant_549391" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="6362042904" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3082" name="Constant_8264/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3083" name="__module.model.gpt_neox.layers.13.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3084" name="Constant_466918" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="6362141208" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3085" name="__module.model.gpt_neox.layers.13.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="2586,input.221"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3086" name="__module.model.gpt_neox.layers.13.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="2595,input.223"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3087" name="Constant_8284" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="6362239512" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3088" name="Convert_549395" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3089" name="Constant_8284/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="6513234456" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3090" name="Convert_549398" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3091" name="Subtract_549399" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3092" name="Constant_549400" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="6513240600" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3093" name="Constant_8284/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3094" name="__module.model.gpt_neox.layers.13.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3095" name="Constant_466919" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6513265176" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3096" name="__module.model.gpt_neox.layers.13.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2598,input.225"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3097" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3098" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3099" name="Constant_466920" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6513289752" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3100" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3101" name="Constant_466921" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6513314328" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3102" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2456,input.213"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3103" name="Constant_7743" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="6513338904" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3104" name="Convert_549368" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3105" name="Constant_7743/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="6626585112" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3106" name="Convert_549371" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3107" name="Subtract_549372" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3108" name="Constant_549373" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="6626603544" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3109" name="Constant_7743/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3110" name="__module.model.gpt_neox.layers.13.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3111" name="Constant_466922" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="6626677272" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3112" name="__module.model.gpt_neox.layers.13.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="2463,qkv.53"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3113" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3114" name="__module.model.gpt_neox.layers.13.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2467,qkv.55"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="3115" name="Constant_428204" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3116" name="Constant_428207" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3117" name="Constant_428210" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3118" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2468"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3119" name="Constant_7771" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2469"> <dim>4</dim> </port> </output> </layer> <layer id="3120" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2470,query.105"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3121" name="Constant_428216" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3122" name="Constant_428219" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3123" name="Constant_428222" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3124" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2477,q.27"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3125" name="Constant_7847" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="2489,cos_cached.27"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="3126" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3127" name="Constant_428228" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3128" name="Constant_428231" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3129" name="Constant_428234" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3130" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2471"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3131" name="Constant_7785" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2472"> <dim>4</dim> </port> </output> </layer> <layer id="3132" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2473,key.131"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3133" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3134" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3135" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2481" /> </output> </layer> <layer id="3136" name="Constant_535342" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3137" name="Constant_535335" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3138" name="Constant_535337" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3139" name="Constant_535339" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3140" name="Concat_535340" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3141" name="Broadcast_535343" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="3142" name="ReadValue_522323" type="ReadValue" version="opset6"> <data variable_id="past_key_values.13.keypresent.13.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.13.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3143" name="Constant_509620" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3144" name="Gather_509621" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3145" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3146" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3147" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2483" /> </output> </layer> <layer id="3148" name="__module.model.gpt_neox.layers.13.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2482,2486,2487,seq_len.53" /> </output> </layer> <layer id="3149" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3150" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3151" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3152" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2490,2493,cos.53"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3153" name="Constant_51981" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3154" name="__module.model.gpt_neox.layers.13.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2496"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3155" name="__module.model.gpt_neox.layers.13.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2497,cos.55"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3156" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2501"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3157" name="Constant_428239" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3158" name="Constant_428238" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3159" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3160" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3161" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2502,2507" /> </output> </layer> <layer id="3162" name="__module.model.gpt_neox.layers.13.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="3163" name="__module.model.gpt_neox.layers.13.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2504,2505,2509,2510" /> </output> </layer> <layer id="3164" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3165" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3166" name="Constant_428237" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3167" name="ScatterUpdate_428240" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3168" name="Constant_428243" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3169" name="Constant_428246" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3170" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2511,x2.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3171" name="Constant_466923" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="3172" name="__module.model.gpt_neox.layers.13.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2512"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3173" name="Constant_428292" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3174" name="Constant_428297" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3175" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2506,x1.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3176" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2514"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3177" name="Constant_7845" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="2488,sin_cached.27"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="3178" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3179" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3180" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2491,2494,sin.53"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3181" name="Constant_51984" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3182" name="__module.model.gpt_neox.layers.13.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2499"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3183" name="__module.model.gpt_neox.layers.13.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2500,sin.55"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3184" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2515"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3185" name="__module.model.gpt_neox.layers.13.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2516,query.107"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3186" name="Constant_428344" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3187" name="Constant_428347" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3188" name="Constant_428350" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3189" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2478,query_pass.27"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="3190" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2534,2544,query.109,query.111"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3191" name="Constant_428356" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3192" name="Constant_428359" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3193" name="Constant_428362" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3194" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2479,k.27"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3195" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2517"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3196" name="Constant_428366" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3197" name="Constant_428365" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3198" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3199" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3200" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2518,2523" /> </output> </layer> <layer id="3201" name="__module.model.gpt_neox.layers.13.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="3202" name="__module.model.gpt_neox.layers.13.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2520,2521,2525,2526" /> </output> </layer> <layer id="3203" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3204" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3205" name="Constant_428364" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3206" name="ScatterUpdate_428367" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3207" name="Constant_428370" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3208" name="Constant_428373" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3209" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2527,x2.55"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3210" name="Constant_466924" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="3211" name="__module.model.gpt_neox.layers.13.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2528"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3212" name="Constant_428419" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3213" name="Constant_428424" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3214" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2522,x1.55"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3215" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2530"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3216" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2531"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3217" name="__module.model.gpt_neox.layers.13.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2532,key.133"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3218" name="Constant_428470" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3219" name="Constant_428473" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3220" name="Constant_428476" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3221" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2480,key_pass.27"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="3222" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2536,key.135"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3223" name="__module.model.gpt_neox.layers.13.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.13.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3224" name="Constant_535355" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3225" name="Constant_535348" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3226" name="Constant_535350" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3227" name="Constant_535352" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3228" name="Concat_535353" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3229" name="Broadcast_535356" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="3230" name="ReadValue_522325" type="ReadValue" version="opset6"> <data variable_id="past_key_values.13.valuepresent.13.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.13.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3231" name="Constant_509623" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3232" name="Gather_509624" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3233" name="Constant_428482" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3234" name="Constant_428485" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3235" name="Constant_428488" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3236" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2474"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3237" name="Constant_7799" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2475"> <dim>4</dim> </port> </output> </layer> <layer id="3238" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2476,value.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3239" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.13.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3240" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="2553"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3241" name="Constant_428492" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3242" name="Constant_428491" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3243" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3244" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3245" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2548" /> </output> </layer> <layer id="3246" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3247" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3248" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2546" /> </output> </layer> <layer id="3249" name="__module.model.gpt_neox.layers.13.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2550,2551" /> </output> </layer> <layer id="3250" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3251" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3252" name="Constant_428490" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3253" name="ScatterUpdate_428493" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3254" name="Constant_428494" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3255" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3256" name="ScatterUpdate_428495" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3257" name="Constant_428498" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3258" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2554"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3259" name="Constant_428570" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3260" name="Constant_428571" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3261" name="Constant_428567" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3262" name="Constant_428566" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3263" name="ScatterUpdate_428572" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3264" name="Constant_428575" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3265" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2555,2556,causal_mask.79"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3266" name="__module.model.gpt_neox.layers.13.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3267" name="__module.model.gpt_neox.layers.13.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="2543,mask_value.27" /> </output> </layer> <layer id="3268" name="__module.model.gpt_neox.layers.13.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="2557,causal_mask.81"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3269" name="Add_454227" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3270" name="ShapeOf_454228" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3271" name="Constant_402470" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3272" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3273" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2541"> <dim>1</dim> </port> </output> </layer> <layer id="3274" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3275" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3276" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3277" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="2558"> <dim>4</dim> </port> </output> </layer> <layer id="3278" name="__module.model.gpt_neox.layers.13.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3279" name="__module.model.gpt_neox.layers.13.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="3280" name="__module.model.gpt_neox.layers.13.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3281" name="__module.model.gpt_neox.layers.13.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3282" name="Maximum_454229" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3283" name="__module.model.gpt_neox.layers.13.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2560,attention_mask.35"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3284" name="__module.model.gpt_neox.layers.13.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2561,2562,sdpa_result.27,tensor.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3285" name="Constant_8224" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2563"> <dim>4</dim> </port> </output> </layer> <layer id="3286" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2564"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3287" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3288" name="__module.model.gpt_neox.layers.13.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="2569,input.215"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3289" name="Constant_8242" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="6626751000" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3290" name="Convert_549377" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3291" name="Constant_8242/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="6664499736" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3292" name="Convert_549380" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3293" name="Subtract_549381" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3294" name="Constant_549382" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="6664505880" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3295" name="Constant_8242/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3296" name="__module.model.gpt_neox.layers.13.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3297" name="Constant_466925" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6513265176" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3298" name="__module.model.gpt_neox.layers.13.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2572,2574,input.217"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3299" name="__module.model.gpt_neox.layers.13/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2600"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3300" name="__module.model.gpt_neox.layers.13/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2601,2603,input.227"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3301" name="__module.model.gpt_neox.layers.14.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3302" name="__module.model.gpt_neox.layers.14.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3303" name="Constant_466926" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6664530456" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3304" name="__module.model.gpt_neox.layers.14.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3305" name="Constant_466927" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6664555032" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3306" name="__module.model.gpt_neox.layers.14.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2738,input.235"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3307" name="Constant_8842" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="6664579608" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3308" name="Convert_549422" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3309" name="Constant_8842/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="6815574552" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3310" name="Convert_549425" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3311" name="Subtract_549426" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3312" name="Constant_549427" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="6815599128" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3313" name="Constant_8842/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3314" name="__module.model.gpt_neox.layers.14.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3315" name="Constant_466928" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="6815697432" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3316" name="__module.model.gpt_neox.layers.14.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="2743,input.237"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3317" name="__module.model.gpt_neox.layers.14.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="2752,input.239"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3318" name="Constant_8862" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="6815795736" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3319" name="Convert_549431" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3320" name="Constant_8862/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="6966790680" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3321" name="Convert_549434" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3322" name="Subtract_549435" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3323" name="Constant_549436" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="6966796824" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3324" name="Constant_8862/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3325" name="__module.model.gpt_neox.layers.14.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3326" name="Constant_466929" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6966821400" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3327" name="__module.model.gpt_neox.layers.14.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2755,input.241"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3328" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3329" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3330" name="Constant_466930" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6966845976" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3331" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3332" name="Constant_466931" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6966870552" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3333" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2613,input.229"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3334" name="Constant_8321" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="6966895128" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3335" name="Convert_549404" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3336" name="Constant_8321/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="7080141336" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3337" name="Convert_549407" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3338" name="Subtract_549408" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3339" name="Constant_549409" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="7080159768" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3340" name="Constant_8321/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3341" name="__module.model.gpt_neox.layers.14.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3342" name="Constant_466932" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="7080233496" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3343" name="__module.model.gpt_neox.layers.14.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="2620,qkv.57"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3344" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3345" name="__module.model.gpt_neox.layers.14.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2624,qkv.59"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="3346" name="Constant_428583" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3347" name="Constant_428586" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3348" name="Constant_428589" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3349" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2625"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3350" name="Constant_8349" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2626"> <dim>4</dim> </port> </output> </layer> <layer id="3351" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2627,query.113"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3352" name="Constant_428595" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3353" name="Constant_428598" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3354" name="Constant_428601" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3355" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2634,q.29"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3356" name="Constant_8425" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="2646,cos_cached.29"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="3357" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3358" name="Constant_428607" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3359" name="Constant_428610" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3360" name="Constant_428613" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3361" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2628"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3362" name="Constant_8363" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2629"> <dim>4</dim> </port> </output> </layer> <layer id="3363" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2630,key.141"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3364" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3365" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3366" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2638" /> </output> </layer> <layer id="3367" name="Constant_535368" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3368" name="Constant_535361" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3369" name="Constant_535363" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3370" name="Constant_535365" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3371" name="Concat_535366" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3372" name="Broadcast_535369" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="3373" name="ReadValue_522327" type="ReadValue" version="opset6"> <data variable_id="past_key_values.14.keypresent.14.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.14.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3374" name="Constant_509626" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3375" name="Gather_509627" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3376" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3377" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3378" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2640" /> </output> </layer> <layer id="3379" name="__module.model.gpt_neox.layers.14.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2639,2643,2644,seq_len.57" /> </output> </layer> <layer id="3380" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3381" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3382" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3383" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2647,2650,cos.57"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3384" name="Constant_51987" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3385" name="__module.model.gpt_neox.layers.14.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2653"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3386" name="__module.model.gpt_neox.layers.14.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2654,cos.59"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3387" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2658"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3388" name="Constant_428618" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3389" name="Constant_428617" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3390" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3391" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3392" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2659,2664" /> </output> </layer> <layer id="3393" name="__module.model.gpt_neox.layers.14.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="3394" name="__module.model.gpt_neox.layers.14.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2661,2662,2666,2667" /> </output> </layer> <layer id="3395" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3396" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3397" name="Constant_428616" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3398" name="ScatterUpdate_428619" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3399" name="Constant_428622" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3400" name="Constant_428625" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3401" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2668,x2.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3402" name="Constant_466933" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="3403" name="__module.model.gpt_neox.layers.14.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2669"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3404" name="Constant_428671" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3405" name="Constant_428676" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3406" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2663,x1.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3407" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2671"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3408" name="Constant_8423" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="2645,sin_cached.29"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="3409" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3410" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3411" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2648,2651,sin.57"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3412" name="Constant_51990" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3413" name="__module.model.gpt_neox.layers.14.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2656"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3414" name="__module.model.gpt_neox.layers.14.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2657,sin.59"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3415" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2672"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3416" name="__module.model.gpt_neox.layers.14.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2673,query.115"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3417" name="Constant_428723" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3418" name="Constant_428726" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3419" name="Constant_428729" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3420" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2635,query_pass.29"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="3421" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2691,2701,query.117,query.119"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3422" name="Constant_428735" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3423" name="Constant_428738" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3424" name="Constant_428741" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3425" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2636,k.29"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3426" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2674"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3427" name="Constant_428745" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3428" name="Constant_428744" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3429" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3430" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3431" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2675,2680" /> </output> </layer> <layer id="3432" name="__module.model.gpt_neox.layers.14.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="3433" name="__module.model.gpt_neox.layers.14.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2677,2678,2682,2683" /> </output> </layer> <layer id="3434" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3435" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3436" name="Constant_428743" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3437" name="ScatterUpdate_428746" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3438" name="Constant_428749" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3439" name="Constant_428752" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3440" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2684,x2.59"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3441" name="Constant_466934" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="3442" name="__module.model.gpt_neox.layers.14.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2685"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3443" name="Constant_428798" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3444" name="Constant_428803" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3445" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2679,x1.59"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3446" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2687"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3447" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2688"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3448" name="__module.model.gpt_neox.layers.14.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2689,key.143"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3449" name="Constant_428849" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3450" name="Constant_428852" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3451" name="Constant_428855" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3452" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2637,key_pass.29"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="3453" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2693,key.145"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3454" name="__module.model.gpt_neox.layers.14.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.14.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3455" name="Constant_535381" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3456" name="Constant_535374" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3457" name="Constant_535376" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3458" name="Constant_535378" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3459" name="Concat_535379" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3460" name="Broadcast_535382" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="3461" name="ReadValue_522329" type="ReadValue" version="opset6"> <data variable_id="past_key_values.14.valuepresent.14.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.14.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3462" name="Constant_509629" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3463" name="Gather_509630" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3464" name="Constant_428861" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3465" name="Constant_428864" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3466" name="Constant_428867" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3467" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2631"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3468" name="Constant_8377" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2632"> <dim>4</dim> </port> </output> </layer> <layer id="3469" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2633,value.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3470" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.14.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3471" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="2710"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3472" name="Constant_428871" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3473" name="Constant_428870" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3474" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3475" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3476" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2705" /> </output> </layer> <layer id="3477" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3478" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3479" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2703" /> </output> </layer> <layer id="3480" name="__module.model.gpt_neox.layers.14.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2707,2708" /> </output> </layer> <layer id="3481" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3482" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3483" name="Constant_428869" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3484" name="ScatterUpdate_428872" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3485" name="Constant_428873" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3486" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3487" name="ScatterUpdate_428874" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3488" name="Constant_428877" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3489" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2711"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3490" name="Constant_428949" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3491" name="Constant_428950" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3492" name="Constant_428946" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3493" name="Constant_428945" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3494" name="ScatterUpdate_428951" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3495" name="Constant_428954" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3496" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2712,2713,causal_mask.85"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3497" name="__module.model.gpt_neox.layers.14.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3498" name="__module.model.gpt_neox.layers.14.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="2700,mask_value.29" /> </output> </layer> <layer id="3499" name="__module.model.gpt_neox.layers.14.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="2714,causal_mask.87"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3500" name="Add_454303" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3501" name="ShapeOf_454304" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3502" name="Constant_402485" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3503" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3504" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2698"> <dim>1</dim> </port> </output> </layer> <layer id="3505" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3506" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3507" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3508" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="2715"> <dim>4</dim> </port> </output> </layer> <layer id="3509" name="__module.model.gpt_neox.layers.14.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3510" name="__module.model.gpt_neox.layers.14.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="3511" name="__module.model.gpt_neox.layers.14.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3512" name="__module.model.gpt_neox.layers.14.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3513" name="Maximum_454305" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3514" name="__module.model.gpt_neox.layers.14.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2717,attention_mask.37"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3515" name="__module.model.gpt_neox.layers.14.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2718,2719,sdpa_result.29,tensor.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3516" name="Constant_8802" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2720"> <dim>4</dim> </port> </output> </layer> <layer id="3517" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2721"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3518" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3519" name="__module.model.gpt_neox.layers.14.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="2726,input.231"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3520" name="Constant_8820" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="7080307224" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3521" name="Convert_549413" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3522" name="Constant_8820/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="7118055960" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3523" name="Convert_549416" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3524" name="Subtract_549417" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3525" name="Constant_549418" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="7118062104" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3526" name="Constant_8820/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3527" name="__module.model.gpt_neox.layers.14.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3528" name="Constant_466935" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="6966821400" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3529" name="__module.model.gpt_neox.layers.14.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2729,2731,input.233"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3530" name="__module.model.gpt_neox.layers.14/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2757"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3531" name="__module.model.gpt_neox.layers.14/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2758,2760,input.243"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3532" name="__module.model.gpt_neox.layers.15.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3533" name="__module.model.gpt_neox.layers.15.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3534" name="Constant_466936" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7118086680" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3535" name="__module.model.gpt_neox.layers.15.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3536" name="Constant_466937" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7118111256" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3537" name="__module.model.gpt_neox.layers.15.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2895,input.251"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3538" name="Constant_9420" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="7118135832" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3539" name="Convert_549458" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3540" name="Constant_9420/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="7269130776" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3541" name="Convert_549461" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3542" name="Subtract_549462" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3543" name="Constant_549463" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="7269155352" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3544" name="Constant_9420/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3545" name="__module.model.gpt_neox.layers.15.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3546" name="Constant_466938" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="7269253656" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3547" name="__module.model.gpt_neox.layers.15.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="2900,input.253"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3548" name="__module.model.gpt_neox.layers.15.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="2909,input.255"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3549" name="Constant_9440" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="7269351960" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3550" name="Convert_549467" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3551" name="Constant_9440/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="7420346904" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3552" name="Convert_549470" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3553" name="Subtract_549471" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3554" name="Constant_549472" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="7420353048" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3555" name="Constant_9440/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3556" name="__module.model.gpt_neox.layers.15.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3557" name="Constant_466939" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7420377624" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3558" name="__module.model.gpt_neox.layers.15.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2912,input.257"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3559" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3560" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3561" name="Constant_466940" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7420402200" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3562" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3563" name="Constant_466941" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7420426776" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3564" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2770,input.245"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3565" name="Constant_8899" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="7420451352" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3566" name="Convert_549440" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3567" name="Constant_8899/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="7533697560" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3568" name="Convert_549443" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3569" name="Subtract_549444" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3570" name="Constant_549445" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="7533715992" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3571" name="Constant_8899/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3572" name="__module.model.gpt_neox.layers.15.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3573" name="Constant_466942" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="7533789720" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3574" name="__module.model.gpt_neox.layers.15.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="2777,qkv.61"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3575" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3576" name="__module.model.gpt_neox.layers.15.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2781,qkv.63"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="3577" name="Constant_428962" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3578" name="Constant_428965" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3579" name="Constant_428968" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3580" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2782"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3581" name="Constant_8927" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2783"> <dim>4</dim> </port> </output> </layer> <layer id="3582" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2784,query.121"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3583" name="Constant_428974" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3584" name="Constant_428977" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3585" name="Constant_428980" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3586" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2791,q.31"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3587" name="Constant_9003" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="2803,cos_cached.31"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="3588" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3589" name="Constant_428986" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3590" name="Constant_428989" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3591" name="Constant_428992" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3592" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2785"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3593" name="Constant_8941" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2786"> <dim>4</dim> </port> </output> </layer> <layer id="3594" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2787,key.151"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3595" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3596" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3597" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2795" /> </output> </layer> <layer id="3598" name="Constant_535394" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3599" name="Constant_535387" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3600" name="Constant_535389" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3601" name="Constant_535391" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3602" name="Concat_535392" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3603" name="Broadcast_535395" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="3604" name="ReadValue_522331" type="ReadValue" version="opset6"> <data variable_id="past_key_values.15.keypresent.15.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.15.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3605" name="Constant_509632" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3606" name="Gather_509633" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3607" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3608" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3609" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2797" /> </output> </layer> <layer id="3610" name="__module.model.gpt_neox.layers.15.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2796,2800,2801,seq_len.61" /> </output> </layer> <layer id="3611" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3612" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3613" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3614" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2804,2807,cos.61"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3615" name="Constant_51993" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3616" name="__module.model.gpt_neox.layers.15.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2810"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3617" name="__module.model.gpt_neox.layers.15.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2811,cos.63"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3618" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2815"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3619" name="Constant_428997" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3620" name="Constant_428996" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3621" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3622" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3623" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2816,2821" /> </output> </layer> <layer id="3624" name="__module.model.gpt_neox.layers.15.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="3625" name="__module.model.gpt_neox.layers.15.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2818,2819,2823,2824" /> </output> </layer> <layer id="3626" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3627" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3628" name="Constant_428995" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3629" name="ScatterUpdate_428998" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3630" name="Constant_429001" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3631" name="Constant_429004" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3632" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2825,x2.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3633" name="Constant_466943" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="3634" name="__module.model.gpt_neox.layers.15.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2826"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3635" name="Constant_429050" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3636" name="Constant_429055" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3637" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2820,x1.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3638" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2828"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3639" name="Constant_9001" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="2802,sin_cached.31"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="3640" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3641" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3642" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2805,2808,sin.61"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3643" name="Constant_51996" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3644" name="__module.model.gpt_neox.layers.15.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2813"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3645" name="__module.model.gpt_neox.layers.15.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2814,sin.63"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3646" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2829"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3647" name="__module.model.gpt_neox.layers.15.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2830,query.123"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3648" name="Constant_429102" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3649" name="Constant_429105" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3650" name="Constant_429108" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3651" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2792,query_pass.31"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="3652" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2848,2858,query.125,query.127"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3653" name="Constant_429114" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3654" name="Constant_429117" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3655" name="Constant_429120" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3656" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2793,k.31"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3657" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2831"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3658" name="Constant_429124" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3659" name="Constant_429123" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3660" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3661" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3662" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2832,2837" /> </output> </layer> <layer id="3663" name="__module.model.gpt_neox.layers.15.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="3664" name="__module.model.gpt_neox.layers.15.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2834,2835,2839,2840" /> </output> </layer> <layer id="3665" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3666" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3667" name="Constant_429122" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3668" name="ScatterUpdate_429125" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3669" name="Constant_429128" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3670" name="Constant_429131" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3671" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2841,x2.63"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3672" name="Constant_466944" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="3673" name="__module.model.gpt_neox.layers.15.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2842"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3674" name="Constant_429177" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3675" name="Constant_429182" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3676" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2836,x1.63"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3677" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2844"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3678" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2845"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3679" name="__module.model.gpt_neox.layers.15.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2846,key.153"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3680" name="Constant_429228" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3681" name="Constant_429231" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3682" name="Constant_429234" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3683" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2794,key_pass.31"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="3684" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="2850,key.155"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3685" name="__module.model.gpt_neox.layers.15.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.15.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3686" name="Constant_535407" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3687" name="Constant_535400" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3688" name="Constant_535402" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3689" name="Constant_535404" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3690" name="Concat_535405" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3691" name="Broadcast_535408" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="3692" name="ReadValue_522333" type="ReadValue" version="opset6"> <data variable_id="past_key_values.15.valuepresent.15.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.15.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3693" name="Constant_509635" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3694" name="Gather_509636" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3695" name="Constant_429240" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3696" name="Constant_429243" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3697" name="Constant_429246" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3698" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2788"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3699" name="Constant_8955" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2789"> <dim>4</dim> </port> </output> </layer> <layer id="3700" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2790,value.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3701" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.15.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3702" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="2867"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3703" name="Constant_429250" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3704" name="Constant_429249" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3705" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3706" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3707" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2862" /> </output> </layer> <layer id="3708" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3709" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3710" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2860" /> </output> </layer> <layer id="3711" name="__module.model.gpt_neox.layers.15.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2864,2865" /> </output> </layer> <layer id="3712" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3713" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3714" name="Constant_429248" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3715" name="ScatterUpdate_429251" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3716" name="Constant_429252" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3717" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3718" name="ScatterUpdate_429253" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3719" name="Constant_429256" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3720" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2868"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3721" name="Constant_429328" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3722" name="Constant_429329" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3723" name="Constant_429325" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3724" name="Constant_429324" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3725" name="ScatterUpdate_429330" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3726" name="Constant_429333" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3727" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="2869,2870,causal_mask.91"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3728" name="__module.model.gpt_neox.layers.15.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3729" name="__module.model.gpt_neox.layers.15.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="2857,mask_value.31" /> </output> </layer> <layer id="3730" name="__module.model.gpt_neox.layers.15.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="2871,causal_mask.93"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3731" name="Add_454379" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3732" name="ShapeOf_454380" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3733" name="Constant_402500" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3734" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3735" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2855"> <dim>1</dim> </port> </output> </layer> <layer id="3736" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3737" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3738" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3739" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="2872"> <dim>4</dim> </port> </output> </layer> <layer id="3740" name="__module.model.gpt_neox.layers.15.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3741" name="__module.model.gpt_neox.layers.15.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="3742" name="__module.model.gpt_neox.layers.15.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3743" name="__module.model.gpt_neox.layers.15.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3744" name="Maximum_454381" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3745" name="__module.model.gpt_neox.layers.15.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2874,attention_mask.39"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3746" name="__module.model.gpt_neox.layers.15.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2875,2876,sdpa_result.31,tensor.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3747" name="Constant_9380" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2877"> <dim>4</dim> </port> </output> </layer> <layer id="3748" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2878"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3749" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3750" name="__module.model.gpt_neox.layers.15.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="2883,input.247"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3751" name="Constant_9398" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="7533863448" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3752" name="Convert_549449" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3753" name="Constant_9398/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="7571612184" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3754" name="Convert_549452" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3755" name="Subtract_549453" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3756" name="Constant_549454" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="7571618328" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3757" name="Constant_9398/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3758" name="__module.model.gpt_neox.layers.15.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3759" name="Constant_466945" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7420377624" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3760" name="__module.model.gpt_neox.layers.15.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2886,2888,input.249"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3761" name="__module.model.gpt_neox.layers.15/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2914"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3762" name="__module.model.gpt_neox.layers.15/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2915,2917,input.259"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3763" name="__module.model.gpt_neox.layers.16.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3764" name="__module.model.gpt_neox.layers.16.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3765" name="Constant_466946" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7571642904" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3766" name="__module.model.gpt_neox.layers.16.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3767" name="Constant_466947" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7571667480" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3768" name="__module.model.gpt_neox.layers.16.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3052,input.267"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3769" name="Constant_9998" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="7571692056" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3770" name="Convert_549494" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3771" name="Constant_9998/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="7722687000" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3772" name="Convert_549497" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3773" name="Subtract_549498" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3774" name="Constant_549499" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="7722711576" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="3775" name="Constant_9998/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3776" name="__module.model.gpt_neox.layers.16.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3777" name="Constant_466948" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="7722809880" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3778" name="__module.model.gpt_neox.layers.16.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="3057,input.269"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3779" name="__module.model.gpt_neox.layers.16.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="3066,input.271"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3780" name="Constant_10018" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="7722908184" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3781" name="Convert_549503" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3782" name="Constant_10018/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="7873903128" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3783" name="Convert_549506" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3784" name="Subtract_549507" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3785" name="Constant_549508" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="7873909272" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3786" name="Constant_10018/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="3787" name="__module.model.gpt_neox.layers.16.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3788" name="Constant_466949" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7873933848" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3789" name="__module.model.gpt_neox.layers.16.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3069,input.273"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3790" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3791" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3792" name="Constant_466950" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7873958424" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3793" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3794" name="Constant_466951" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7873983000" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3795" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="2927,input.261"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3796" name="Constant_9477" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="7874007576" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3797" name="Convert_549476" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3798" name="Constant_9477/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="7987253784" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3799" name="Convert_549479" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3800" name="Subtract_549480" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3801" name="Constant_549481" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="7987272216" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="3802" name="Constant_9477/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3803" name="__module.model.gpt_neox.layers.16.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3804" name="Constant_466952" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="7987345944" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3805" name="__module.model.gpt_neox.layers.16.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="2934,qkv.65"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="3806" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3807" name="__module.model.gpt_neox.layers.16.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2938,qkv.67"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="3808" name="Constant_429341" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3809" name="Constant_429344" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3810" name="Constant_429347" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3811" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2939"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3812" name="Constant_9505" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2940"> <dim>4</dim> </port> </output> </layer> <layer id="3813" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2941,query.129"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3814" name="Constant_429353" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3815" name="Constant_429356" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3816" name="Constant_429359" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3817" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2948,q.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3818" name="Constant_9581" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="2960,cos_cached.33"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="3819" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3820" name="Constant_429365" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3821" name="Constant_429368" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3822" name="Constant_429371" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3823" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2942"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3824" name="Constant_9519" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2943"> <dim>4</dim> </port> </output> </layer> <layer id="3825" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2944,key.161"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3826" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3827" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3828" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2952" /> </output> </layer> <layer id="3829" name="Constant_535420" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3830" name="Constant_535413" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3831" name="Constant_535415" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3832" name="Constant_535417" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3833" name="Concat_535418" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3834" name="Broadcast_535421" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="3835" name="ReadValue_522335" type="ReadValue" version="opset6"> <data variable_id="past_key_values.16.keypresent.16.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.16.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3836" name="Constant_509638" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3837" name="Gather_509639" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3838" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3839" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3840" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2954" /> </output> </layer> <layer id="3841" name="__module.model.gpt_neox.layers.16.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="2953,2957,2958,seq_len.65" /> </output> </layer> <layer id="3842" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3843" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3844" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3845" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2961,2964,cos.65"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3846" name="Constant_51999" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3847" name="__module.model.gpt_neox.layers.16.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2967"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3848" name="__module.model.gpt_neox.layers.16.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2968,cos.67"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3849" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2972"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3850" name="Constant_429376" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3851" name="Constant_429375" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3852" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3853" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3854" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2973,2978" /> </output> </layer> <layer id="3855" name="__module.model.gpt_neox.layers.16.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="3856" name="__module.model.gpt_neox.layers.16.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2975,2976,2980,2981" /> </output> </layer> <layer id="3857" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3858" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3859" name="Constant_429374" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3860" name="ScatterUpdate_429377" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3861" name="Constant_429380" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3862" name="Constant_429383" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3863" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2982,x2.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3864" name="Constant_466953" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="3865" name="__module.model.gpt_neox.layers.16.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2983"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3866" name="Constant_429429" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3867" name="Constant_429434" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3868" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2977,x1.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3869" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="2985"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3870" name="Constant_9579" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="2959,sin_cached.33"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="3871" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3872" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3873" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="2962,2965,sin.65"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3874" name="Constant_52002" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3875" name="__module.model.gpt_neox.layers.16.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="2970"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3876" name="__module.model.gpt_neox.layers.16.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="2971,sin.67"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3877" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2986"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3878" name="__module.model.gpt_neox.layers.16.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2987,query.131"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3879" name="Constant_429481" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3880" name="Constant_429484" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3881" name="Constant_429487" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3882" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2949,query_pass.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="3883" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3005,3015,query.133,query.135"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3884" name="Constant_429493" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3885" name="Constant_429496" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3886" name="Constant_429499" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3887" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2950,k.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3888" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="2988"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3889" name="Constant_429503" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3890" name="Constant_429502" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3891" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3892" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3893" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="2989,2994" /> </output> </layer> <layer id="3894" name="__module.model.gpt_neox.layers.16.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="3895" name="__module.model.gpt_neox.layers.16.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="2991,2992,2996,2997" /> </output> </layer> <layer id="3896" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3897" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3898" name="Constant_429501" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3899" name="ScatterUpdate_429504" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3900" name="Constant_429507" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3901" name="Constant_429510" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3902" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2998,x2.67"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3903" name="Constant_466954" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="3904" name="__module.model.gpt_neox.layers.16.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="2999"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3905" name="Constant_429556" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3906" name="Constant_429561" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3907" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2993,x1.67"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="3908" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3001"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3909" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3002"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3910" name="__module.model.gpt_neox.layers.16.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3003,key.163"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="3911" name="Constant_429607" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3912" name="Constant_429610" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3913" name="Constant_429613" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3914" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2951,key_pass.33"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="3915" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3007,key.165"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3916" name="__module.model.gpt_neox.layers.16.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.16.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3917" name="Constant_535433" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3918" name="Constant_535426" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3919" name="Constant_535428" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3920" name="Constant_535430" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3921" name="Concat_535431" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3922" name="Broadcast_535434" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="3923" name="ReadValue_522337" type="ReadValue" version="opset6"> <data variable_id="past_key_values.16.valuepresent.16.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.16.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3924" name="Constant_509641" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="3925" name="Gather_509642" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3926" name="Constant_429619" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3927" name="Constant_429622" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="3928" name="Constant_429625" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3929" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="2945"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3930" name="Constant_9533" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="2946"> <dim>4</dim> </port> </output> </layer> <layer id="3931" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="2947,value.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3932" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.16.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3933" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="3024"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3934" name="Constant_429629" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3935" name="Constant_429628" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3936" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3937" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3938" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3019" /> </output> </layer> <layer id="3939" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3940" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3941" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3017" /> </output> </layer> <layer id="3942" name="__module.model.gpt_neox.layers.16.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3021,3022" /> </output> </layer> <layer id="3943" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3944" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3945" name="Constant_429627" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3946" name="ScatterUpdate_429630" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3947" name="Constant_429631" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3948" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3949" name="ScatterUpdate_429632" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3950" name="Constant_429635" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3951" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3025"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="3952" name="Constant_429707" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3953" name="Constant_429708" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3954" name="Constant_429704" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="3955" name="Constant_429703" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3956" name="ScatterUpdate_429709" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3957" name="Constant_429712" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3958" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3026,3027,causal_mask.97"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3959" name="__module.model.gpt_neox.layers.16.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="3960" name="__module.model.gpt_neox.layers.16.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="3014,mask_value.33" /> </output> </layer> <layer id="3961" name="__module.model.gpt_neox.layers.16.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="3028,causal_mask.99"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3962" name="Add_454455" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3963" name="ShapeOf_454456" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3964" name="Constant_402515" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3965" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="3966" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3012"> <dim>1</dim> </port> </output> </layer> <layer id="3967" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3968" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3969" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3970" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="3029"> <dim>4</dim> </port> </output> </layer> <layer id="3971" name="__module.model.gpt_neox.layers.16.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3972" name="__module.model.gpt_neox.layers.16.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="3973" name="__module.model.gpt_neox.layers.16.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3974" name="__module.model.gpt_neox.layers.16.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3975" name="Maximum_454457" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="3976" name="__module.model.gpt_neox.layers.16.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3031,attention_mask.41"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="3977" name="__module.model.gpt_neox.layers.16.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3032,3033,sdpa_result.33,tensor.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="3978" name="Constant_9958" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3034"> <dim>4</dim> </port> </output> </layer> <layer id="3979" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3035"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="3980" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="3981" name="__module.model.gpt_neox.layers.16.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="3040,input.263"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3982" name="Constant_9976" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="7987419672" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3983" name="Convert_549485" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3984" name="Constant_9976/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="8025168408" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3985" name="Convert_549488" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3986" name="Subtract_549489" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3987" name="Constant_549490" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="8025174552" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="3988" name="Constant_9976/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3989" name="__module.model.gpt_neox.layers.16.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3990" name="Constant_466955" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="7873933848" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3991" name="__module.model.gpt_neox.layers.16.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3043,3045,input.265"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3992" name="__module.model.gpt_neox.layers.16/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3071"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3993" name="__module.model.gpt_neox.layers.16/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3072,3074,input.275"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3994" name="__module.model.gpt_neox.layers.17.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="3995" name="__module.model.gpt_neox.layers.17.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3996" name="Constant_466956" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8025199128" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3997" name="__module.model.gpt_neox.layers.17.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3998" name="Constant_466957" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8025223704" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="3999" name="__module.model.gpt_neox.layers.17.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3209,input.283"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4000" name="Constant_10576" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="8025248280" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4001" name="Convert_549530" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4002" name="Constant_10576/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="8176243224" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4003" name="Convert_549533" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4004" name="Subtract_549534" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4005" name="Constant_549535" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="8176267800" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4006" name="Constant_10576/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4007" name="__module.model.gpt_neox.layers.17.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4008" name="Constant_466958" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="8176366104" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4009" name="__module.model.gpt_neox.layers.17.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="3214,input.285"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4010" name="__module.model.gpt_neox.layers.17.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="3223,input.287"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4011" name="Constant_10596" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="8176464408" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4012" name="Convert_549539" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4013" name="Constant_10596/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="8327459352" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4014" name="Convert_549542" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4015" name="Subtract_549543" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4016" name="Constant_549544" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="8327465496" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4017" name="Constant_10596/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4018" name="__module.model.gpt_neox.layers.17.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4019" name="Constant_466959" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8327490072" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4020" name="__module.model.gpt_neox.layers.17.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3226,input.289"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4021" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4022" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4023" name="Constant_466960" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8327514648" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4024" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4025" name="Constant_466961" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8327539224" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4026" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3084,input.277"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4027" name="Constant_10055" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="8327563800" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4028" name="Convert_549512" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4029" name="Constant_10055/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="8440810008" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4030" name="Convert_549515" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4031" name="Subtract_549516" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4032" name="Constant_549517" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="8440828440" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4033" name="Constant_10055/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4034" name="__module.model.gpt_neox.layers.17.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4035" name="Constant_466962" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="8440902168" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4036" name="__module.model.gpt_neox.layers.17.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="3091,qkv.69"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4037" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4038" name="__module.model.gpt_neox.layers.17.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3095,qkv.71"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="4039" name="Constant_429720" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4040" name="Constant_429723" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4041" name="Constant_429726" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4042" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3096"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4043" name="Constant_10083" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3097"> <dim>4</dim> </port> </output> </layer> <layer id="4044" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3098,query.137"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4045" name="Constant_429732" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4046" name="Constant_429735" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4047" name="Constant_429738" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4048" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3105,q.35"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4049" name="Constant_10159" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="3117,cos_cached.35"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4050" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4051" name="Constant_429744" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4052" name="Constant_429747" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4053" name="Constant_429750" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4054" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3099"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4055" name="Constant_10097" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3100"> <dim>4</dim> </port> </output> </layer> <layer id="4056" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3101,key.171"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4057" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4058" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4059" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3109" /> </output> </layer> <layer id="4060" name="Constant_535446" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4061" name="Constant_535439" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4062" name="Constant_535441" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4063" name="Constant_535443" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4064" name="Concat_535444" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4065" name="Broadcast_535447" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4066" name="ReadValue_522339" type="ReadValue" version="opset6"> <data variable_id="past_key_values.17.keypresent.17.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.17.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4067" name="Constant_509644" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4068" name="Gather_509645" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4069" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4070" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4071" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3111" /> </output> </layer> <layer id="4072" name="__module.model.gpt_neox.layers.17.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3110,3114,3115,seq_len.69" /> </output> </layer> <layer id="4073" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4074" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4075" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4076" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3118,3121,cos.69"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4077" name="Constant_52005" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4078" name="__module.model.gpt_neox.layers.17.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3124"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4079" name="__module.model.gpt_neox.layers.17.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3125,cos.71"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4080" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3129"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4081" name="Constant_429755" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4082" name="Constant_429754" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4083" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4084" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4085" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3130,3135" /> </output> </layer> <layer id="4086" name="__module.model.gpt_neox.layers.17.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="4087" name="__module.model.gpt_neox.layers.17.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3132,3133,3137,3138" /> </output> </layer> <layer id="4088" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4089" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4090" name="Constant_429753" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4091" name="ScatterUpdate_429756" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4092" name="Constant_429759" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4093" name="Constant_429762" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4094" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3139,x2.69"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4095" name="Constant_466963" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="4096" name="__module.model.gpt_neox.layers.17.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3140"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4097" name="Constant_429808" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4098" name="Constant_429813" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4099" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3134,x1.69"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4100" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3142"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4101" name="Constant_10157" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="3116,sin_cached.35"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4102" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4103" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4104" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3119,3122,sin.69"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4105" name="Constant_52008" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4106" name="__module.model.gpt_neox.layers.17.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3127"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4107" name="__module.model.gpt_neox.layers.17.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3128,sin.71"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4108" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3143"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4109" name="__module.model.gpt_neox.layers.17.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3144,query.139"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4110" name="Constant_429860" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4111" name="Constant_429863" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4112" name="Constant_429866" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4113" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3106,query_pass.35"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="4114" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3162,3172,query.141,query.143"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4115" name="Constant_429872" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4116" name="Constant_429875" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4117" name="Constant_429878" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4118" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3107,k.35"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4119" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3145"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4120" name="Constant_429882" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4121" name="Constant_429881" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4122" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4123" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4124" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3146,3151" /> </output> </layer> <layer id="4125" name="__module.model.gpt_neox.layers.17.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="4126" name="__module.model.gpt_neox.layers.17.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3148,3149,3153,3154" /> </output> </layer> <layer id="4127" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4128" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4129" name="Constant_429880" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4130" name="ScatterUpdate_429883" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4131" name="Constant_429886" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4132" name="Constant_429889" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4133" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3155,x2.71"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4134" name="Constant_466964" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="4135" name="__module.model.gpt_neox.layers.17.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3156"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4136" name="Constant_429935" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4137" name="Constant_429940" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4138" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3150,x1.71"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4139" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3158"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4140" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3159"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4141" name="__module.model.gpt_neox.layers.17.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3160,key.173"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4142" name="Constant_429986" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4143" name="Constant_429989" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4144" name="Constant_429992" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4145" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3108,key_pass.35"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="4146" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3164,key.175"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4147" name="__module.model.gpt_neox.layers.17.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.17.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4148" name="Constant_535459" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4149" name="Constant_535452" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4150" name="Constant_535454" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4151" name="Constant_535456" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4152" name="Concat_535457" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4153" name="Broadcast_535460" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4154" name="ReadValue_522341" type="ReadValue" version="opset6"> <data variable_id="past_key_values.17.valuepresent.17.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.17.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4155" name="Constant_509647" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4156" name="Gather_509648" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4157" name="Constant_429998" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4158" name="Constant_430001" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4159" name="Constant_430004" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4160" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3102"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4161" name="Constant_10111" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3103"> <dim>4</dim> </port> </output> </layer> <layer id="4162" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3104,value.69"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4163" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.17.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4164" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="3181"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="4165" name="Constant_430008" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4166" name="Constant_430007" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4167" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4168" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4169" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3176" /> </output> </layer> <layer id="4170" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4171" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4172" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3174" /> </output> </layer> <layer id="4173" name="__module.model.gpt_neox.layers.17.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3178,3179" /> </output> </layer> <layer id="4174" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4175" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4176" name="Constant_430006" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4177" name="ScatterUpdate_430009" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4178" name="Constant_430010" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4179" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4180" name="ScatterUpdate_430011" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4181" name="Constant_430014" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4182" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3182"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="4183" name="Constant_430086" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4184" name="Constant_430087" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4185" name="Constant_430083" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4186" name="Constant_430082" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4187" name="ScatterUpdate_430088" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4188" name="Constant_430091" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4189" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3183,3184,causal_mask.103"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4190" name="__module.model.gpt_neox.layers.17.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4191" name="__module.model.gpt_neox.layers.17.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="3171,mask_value.35" /> </output> </layer> <layer id="4192" name="__module.model.gpt_neox.layers.17.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="3185,causal_mask.105"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4193" name="Add_454531" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4194" name="ShapeOf_454532" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4195" name="Constant_402530" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4196" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4197" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3169"> <dim>1</dim> </port> </output> </layer> <layer id="4198" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4199" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4200" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4201" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="3186"> <dim>4</dim> </port> </output> </layer> <layer id="4202" name="__module.model.gpt_neox.layers.17.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4203" name="__module.model.gpt_neox.layers.17.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="4204" name="__module.model.gpt_neox.layers.17.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4205" name="__module.model.gpt_neox.layers.17.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4206" name="Maximum_454533" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4207" name="__module.model.gpt_neox.layers.17.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3188,attention_mask.43"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4208" name="__module.model.gpt_neox.layers.17.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3189,3190,sdpa_result.35,tensor.69"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4209" name="Constant_10536" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3191"> <dim>4</dim> </port> </output> </layer> <layer id="4210" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3192"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4211" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4212" name="__module.model.gpt_neox.layers.17.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="3197,input.279"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4213" name="Constant_10554" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="8440975896" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4214" name="Convert_549521" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4215" name="Constant_10554/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="8478724632" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4216" name="Convert_549524" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4217" name="Subtract_549525" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4218" name="Constant_549526" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="8478730776" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4219" name="Constant_10554/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4220" name="__module.model.gpt_neox.layers.17.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4221" name="Constant_466965" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8327490072" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4222" name="__module.model.gpt_neox.layers.17.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3200,3202,input.281"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4223" name="__module.model.gpt_neox.layers.17/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3228"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4224" name="__module.model.gpt_neox.layers.17/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3229,3231,input.291"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4225" name="__module.model.gpt_neox.layers.18.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4226" name="__module.model.gpt_neox.layers.18.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4227" name="Constant_466966" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8478755352" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4228" name="__module.model.gpt_neox.layers.18.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4229" name="Constant_466967" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8478779928" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4230" name="__module.model.gpt_neox.layers.18.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3366,input.299"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4231" name="Constant_11154" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="8478804504" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4232" name="Convert_549566" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4233" name="Constant_11154/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="8629799448" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4234" name="Convert_549569" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4235" name="Subtract_549570" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4236" name="Constant_549571" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="8629824024" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4237" name="Constant_11154/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4238" name="__module.model.gpt_neox.layers.18.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4239" name="Constant_466968" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="8629922328" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4240" name="__module.model.gpt_neox.layers.18.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="3371,input.301"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4241" name="__module.model.gpt_neox.layers.18.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="3380,input.303"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4242" name="Constant_11174" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="8630020632" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4243" name="Convert_549575" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4244" name="Constant_11174/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="8781015576" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4245" name="Convert_549578" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4246" name="Subtract_549579" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4247" name="Constant_549580" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="8781021720" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4248" name="Constant_11174/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4249" name="__module.model.gpt_neox.layers.18.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4250" name="Constant_466969" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8781046296" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4251" name="__module.model.gpt_neox.layers.18.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3383,input.305"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4252" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4253" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4254" name="Constant_466970" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8781070872" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4255" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4256" name="Constant_466971" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8781095448" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4257" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3241,input.293"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4258" name="Constant_10633" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="8781120024" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4259" name="Convert_549548" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4260" name="Constant_10633/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="8894366232" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4261" name="Convert_549551" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4262" name="Subtract_549552" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4263" name="Constant_549553" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="8894384664" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4264" name="Constant_10633/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4265" name="__module.model.gpt_neox.layers.18.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4266" name="Constant_466972" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="8894458392" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4267" name="__module.model.gpt_neox.layers.18.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="3248,qkv.73"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4268" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4269" name="__module.model.gpt_neox.layers.18.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3252,qkv.75"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="4270" name="Constant_430099" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4271" name="Constant_430102" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4272" name="Constant_430105" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4273" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3253"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4274" name="Constant_10661" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3254"> <dim>4</dim> </port> </output> </layer> <layer id="4275" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3255,query.145"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4276" name="Constant_430111" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4277" name="Constant_430114" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4278" name="Constant_430117" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4279" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3262,q.37"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4280" name="Constant_10737" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="3274,cos_cached.37"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4281" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4282" name="Constant_430123" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4283" name="Constant_430126" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4284" name="Constant_430129" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4285" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3256"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4286" name="Constant_10675" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3257"> <dim>4</dim> </port> </output> </layer> <layer id="4287" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3258,key.181"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4288" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4289" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4290" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3266" /> </output> </layer> <layer id="4291" name="Constant_535472" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4292" name="Constant_535465" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4293" name="Constant_535467" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4294" name="Constant_535469" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4295" name="Concat_535470" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4296" name="Broadcast_535473" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4297" name="ReadValue_522343" type="ReadValue" version="opset6"> <data variable_id="past_key_values.18.keypresent.18.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.18.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4298" name="Constant_509650" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4299" name="Gather_509651" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4300" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4301" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4302" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3268" /> </output> </layer> <layer id="4303" name="__module.model.gpt_neox.layers.18.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3267,3271,3272,seq_len.73" /> </output> </layer> <layer id="4304" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4305" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4306" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4307" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3275,3278,cos.73"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4308" name="Constant_52011" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4309" name="__module.model.gpt_neox.layers.18.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3281"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4310" name="__module.model.gpt_neox.layers.18.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3282,cos.75"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4311" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3286"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4312" name="Constant_430134" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4313" name="Constant_430133" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4314" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4315" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4316" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3287,3292" /> </output> </layer> <layer id="4317" name="__module.model.gpt_neox.layers.18.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="4318" name="__module.model.gpt_neox.layers.18.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3289,3290,3294,3295" /> </output> </layer> <layer id="4319" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4320" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4321" name="Constant_430132" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4322" name="ScatterUpdate_430135" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4323" name="Constant_430138" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4324" name="Constant_430141" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4325" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3296,x2.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4326" name="Constant_466973" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="4327" name="__module.model.gpt_neox.layers.18.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3297"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4328" name="Constant_430187" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4329" name="Constant_430192" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4330" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3291,x1.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4331" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3299"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4332" name="Constant_10735" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="3273,sin_cached.37"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4333" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4334" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4335" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3276,3279,sin.73"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4336" name="Constant_52014" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4337" name="__module.model.gpt_neox.layers.18.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3284"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4338" name="__module.model.gpt_neox.layers.18.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3285,sin.75"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4339" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3300"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4340" name="__module.model.gpt_neox.layers.18.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3301,query.147"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4341" name="Constant_430239" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4342" name="Constant_430242" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4343" name="Constant_430245" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4344" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3263,query_pass.37"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="4345" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3319,3329,query.149,query.151"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4346" name="Constant_430251" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4347" name="Constant_430254" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4348" name="Constant_430257" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4349" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3264,k.37"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4350" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3302"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4351" name="Constant_430261" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4352" name="Constant_430260" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4353" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4354" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4355" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3303,3308" /> </output> </layer> <layer id="4356" name="__module.model.gpt_neox.layers.18.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="4357" name="__module.model.gpt_neox.layers.18.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3305,3306,3310,3311" /> </output> </layer> <layer id="4358" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4359" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4360" name="Constant_430259" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4361" name="ScatterUpdate_430262" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4362" name="Constant_430265" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4363" name="Constant_430268" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4364" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3312,x2.75"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4365" name="Constant_466974" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="4366" name="__module.model.gpt_neox.layers.18.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3313"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4367" name="Constant_430314" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4368" name="Constant_430319" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4369" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3307,x1.75"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4370" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3315"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4371" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3316"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4372" name="__module.model.gpt_neox.layers.18.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3317,key.183"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4373" name="Constant_430365" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4374" name="Constant_430368" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4375" name="Constant_430371" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4376" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3265,key_pass.37"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="4377" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3321,key.185"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4378" name="__module.model.gpt_neox.layers.18.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.18.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4379" name="Constant_535485" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4380" name="Constant_535478" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4381" name="Constant_535480" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4382" name="Constant_535482" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4383" name="Concat_535483" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4384" name="Broadcast_535486" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4385" name="ReadValue_522345" type="ReadValue" version="opset6"> <data variable_id="past_key_values.18.valuepresent.18.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.18.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4386" name="Constant_509653" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4387" name="Gather_509654" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4388" name="Constant_430377" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4389" name="Constant_430380" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4390" name="Constant_430383" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4391" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3259"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4392" name="Constant_10689" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3260"> <dim>4</dim> </port> </output> </layer> <layer id="4393" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3261,value.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4394" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.18.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4395" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="3338"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="4396" name="Constant_430387" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4397" name="Constant_430386" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4398" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4399" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4400" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3333" /> </output> </layer> <layer id="4401" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4402" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4403" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3331" /> </output> </layer> <layer id="4404" name="__module.model.gpt_neox.layers.18.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3335,3336" /> </output> </layer> <layer id="4405" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4406" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4407" name="Constant_430385" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4408" name="ScatterUpdate_430388" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4409" name="Constant_430389" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4410" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4411" name="ScatterUpdate_430390" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4412" name="Constant_430393" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4413" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3339"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="4414" name="Constant_430465" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4415" name="Constant_430466" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4416" name="Constant_430462" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4417" name="Constant_430461" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4418" name="ScatterUpdate_430467" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4419" name="Constant_430470" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4420" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3340,3341,causal_mask.109"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4421" name="__module.model.gpt_neox.layers.18.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4422" name="__module.model.gpt_neox.layers.18.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="3328,mask_value.37" /> </output> </layer> <layer id="4423" name="__module.model.gpt_neox.layers.18.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="3342,causal_mask.111"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4424" name="Add_454607" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4425" name="ShapeOf_454608" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4426" name="Constant_402545" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4427" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4428" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3326"> <dim>1</dim> </port> </output> </layer> <layer id="4429" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4430" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4431" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4432" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="3343"> <dim>4</dim> </port> </output> </layer> <layer id="4433" name="__module.model.gpt_neox.layers.18.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4434" name="__module.model.gpt_neox.layers.18.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="4435" name="__module.model.gpt_neox.layers.18.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4436" name="__module.model.gpt_neox.layers.18.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4437" name="Maximum_454609" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4438" name="__module.model.gpt_neox.layers.18.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3345,attention_mask.45"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4439" name="__module.model.gpt_neox.layers.18.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3346,3347,sdpa_result.37,tensor.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4440" name="Constant_11114" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3348"> <dim>4</dim> </port> </output> </layer> <layer id="4441" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3349"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4442" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4443" name="__module.model.gpt_neox.layers.18.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="3354,input.295"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4444" name="Constant_11132" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="8894532120" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4445" name="Convert_549557" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4446" name="Constant_11132/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="8932280856" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4447" name="Convert_549560" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4448" name="Subtract_549561" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4449" name="Constant_549562" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="8932287000" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4450" name="Constant_11132/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4451" name="__module.model.gpt_neox.layers.18.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4452" name="Constant_466975" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8781046296" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4453" name="__module.model.gpt_neox.layers.18.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3357,3359,input.297"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4454" name="__module.model.gpt_neox.layers.18/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3385"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4455" name="__module.model.gpt_neox.layers.18/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3386,3388,input.307"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4456" name="__module.model.gpt_neox.layers.19.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4457" name="__module.model.gpt_neox.layers.19.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4458" name="Constant_466976" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8932311576" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4459" name="__module.model.gpt_neox.layers.19.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4460" name="Constant_466977" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="8932336152" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4461" name="__module.model.gpt_neox.layers.19.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3523,input.315"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4462" name="Constant_11732" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="8932360728" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4463" name="Convert_549602" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4464" name="Constant_11732/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="9083355672" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4465" name="Convert_549605" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4466" name="Subtract_549606" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4467" name="Constant_549607" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="9083380248" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4468" name="Constant_11732/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4469" name="__module.model.gpt_neox.layers.19.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4470" name="Constant_466978" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="9083478552" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4471" name="__module.model.gpt_neox.layers.19.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="3528,input.317"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4472" name="__module.model.gpt_neox.layers.19.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="3537,input.319"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4473" name="Constant_11752" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="9083576856" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4474" name="Convert_549611" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4475" name="Constant_11752/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="9234571800" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4476" name="Convert_549614" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4477" name="Subtract_549615" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4478" name="Constant_549616" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="9234577944" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4479" name="Constant_11752/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4480" name="__module.model.gpt_neox.layers.19.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4481" name="Constant_466979" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9234602520" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4482" name="__module.model.gpt_neox.layers.19.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3540,input.321"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4483" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4484" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4485" name="Constant_466980" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9234627096" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4486" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4487" name="Constant_466981" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9234651672" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4488" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3398,input.309"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4489" name="Constant_11211" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="9234676248" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4490" name="Convert_549584" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4491" name="Constant_11211/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="9347922456" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4492" name="Convert_549587" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4493" name="Subtract_549588" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4494" name="Constant_549589" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="9347940888" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4495" name="Constant_11211/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4496" name="__module.model.gpt_neox.layers.19.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4497" name="Constant_466982" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="9348014616" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4498" name="__module.model.gpt_neox.layers.19.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="3405,qkv.77"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4499" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4500" name="__module.model.gpt_neox.layers.19.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3409,qkv.79"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="4501" name="Constant_430478" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4502" name="Constant_430481" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4503" name="Constant_430484" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4504" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3410"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4505" name="Constant_11239" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3411"> <dim>4</dim> </port> </output> </layer> <layer id="4506" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3412,query.153"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4507" name="Constant_430490" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4508" name="Constant_430493" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4509" name="Constant_430496" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4510" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3419,q.39"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4511" name="Constant_11315" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="3431,cos_cached.39"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4512" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4513" name="Constant_430502" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4514" name="Constant_430505" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4515" name="Constant_430508" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4516" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3413"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4517" name="Constant_11253" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3414"> <dim>4</dim> </port> </output> </layer> <layer id="4518" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3415,key.191"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4519" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4520" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4521" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3423" /> </output> </layer> <layer id="4522" name="Constant_535498" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4523" name="Constant_535491" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4524" name="Constant_535493" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4525" name="Constant_535495" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4526" name="Concat_535496" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4527" name="Broadcast_535499" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4528" name="ReadValue_522347" type="ReadValue" version="opset6"> <data variable_id="past_key_values.19.keypresent.19.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.19.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4529" name="Constant_509656" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4530" name="Gather_509657" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4531" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4532" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4533" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3425" /> </output> </layer> <layer id="4534" name="__module.model.gpt_neox.layers.19.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3424,3428,3429,seq_len.77" /> </output> </layer> <layer id="4535" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4536" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4537" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4538" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3432,3435,cos.77"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4539" name="Constant_52017" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4540" name="__module.model.gpt_neox.layers.19.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3438"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4541" name="__module.model.gpt_neox.layers.19.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3439,cos.79"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4542" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3443"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4543" name="Constant_430513" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4544" name="Constant_430512" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4545" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4546" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4547" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3444,3449" /> </output> </layer> <layer id="4548" name="__module.model.gpt_neox.layers.19.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="4549" name="__module.model.gpt_neox.layers.19.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3446,3447,3451,3452" /> </output> </layer> <layer id="4550" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4551" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4552" name="Constant_430511" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4553" name="ScatterUpdate_430514" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4554" name="Constant_430517" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4555" name="Constant_430520" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4556" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3453,x2.77"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4557" name="Constant_466983" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="4558" name="__module.model.gpt_neox.layers.19.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3454"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4559" name="Constant_430566" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4560" name="Constant_430571" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4561" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3448,x1.77"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4562" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3456"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4563" name="Constant_11313" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="3430,sin_cached.39"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4564" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4565" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4566" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3433,3436,sin.77"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4567" name="Constant_52020" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4568" name="__module.model.gpt_neox.layers.19.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3441"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4569" name="__module.model.gpt_neox.layers.19.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3442,sin.79"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4570" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3457"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4571" name="__module.model.gpt_neox.layers.19.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3458,query.155"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4572" name="Constant_430618" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4573" name="Constant_430621" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4574" name="Constant_430624" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4575" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3420,query_pass.39"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="4576" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3476,3486,query.157,query.159"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4577" name="Constant_430630" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4578" name="Constant_430633" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4579" name="Constant_430636" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4580" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3421,k.39"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4581" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3459"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4582" name="Constant_430640" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4583" name="Constant_430639" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4584" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4585" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4586" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3460,3465" /> </output> </layer> <layer id="4587" name="__module.model.gpt_neox.layers.19.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="4588" name="__module.model.gpt_neox.layers.19.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3462,3463,3467,3468" /> </output> </layer> <layer id="4589" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4590" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4591" name="Constant_430638" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4592" name="ScatterUpdate_430641" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4593" name="Constant_430644" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4594" name="Constant_430647" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4595" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3469,x2.79"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4596" name="Constant_466984" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="4597" name="__module.model.gpt_neox.layers.19.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3470"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4598" name="Constant_430693" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4599" name="Constant_430698" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4600" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3464,x1.79"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4601" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3472"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4602" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3473"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4603" name="__module.model.gpt_neox.layers.19.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3474,key.193"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4604" name="Constant_430744" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4605" name="Constant_430747" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4606" name="Constant_430750" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4607" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3422,key_pass.39"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="4608" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3478,key.195"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4609" name="__module.model.gpt_neox.layers.19.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.19.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4610" name="Constant_535511" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4611" name="Constant_535504" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4612" name="Constant_535506" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4613" name="Constant_535508" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4614" name="Concat_535509" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4615" name="Broadcast_535512" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4616" name="ReadValue_522349" type="ReadValue" version="opset6"> <data variable_id="past_key_values.19.valuepresent.19.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.19.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4617" name="Constant_509659" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4618" name="Gather_509660" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4619" name="Constant_430756" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4620" name="Constant_430759" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4621" name="Constant_430762" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4622" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3416"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4623" name="Constant_11267" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3417"> <dim>4</dim> </port> </output> </layer> <layer id="4624" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3418,value.77"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4625" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.19.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4626" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="3495"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="4627" name="Constant_430766" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4628" name="Constant_430765" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4629" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4630" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4631" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3490" /> </output> </layer> <layer id="4632" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4633" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4634" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3488" /> </output> </layer> <layer id="4635" name="__module.model.gpt_neox.layers.19.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3492,3493" /> </output> </layer> <layer id="4636" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4637" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4638" name="Constant_430764" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4639" name="ScatterUpdate_430767" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4640" name="Constant_430768" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4641" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4642" name="ScatterUpdate_430769" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4643" name="Constant_430772" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4644" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3496"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="4645" name="Constant_430844" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4646" name="Constant_430845" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4647" name="Constant_430841" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4648" name="Constant_430840" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4649" name="ScatterUpdate_430846" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4650" name="Constant_430849" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4651" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3497,3498,causal_mask.115"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4652" name="__module.model.gpt_neox.layers.19.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4653" name="__module.model.gpt_neox.layers.19.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="3485,mask_value.39" /> </output> </layer> <layer id="4654" name="__module.model.gpt_neox.layers.19.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="3499,causal_mask.117"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4655" name="Add_454683" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4656" name="ShapeOf_454684" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4657" name="Constant_402560" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4658" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4659" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3483"> <dim>1</dim> </port> </output> </layer> <layer id="4660" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4661" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4662" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4663" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="3500"> <dim>4</dim> </port> </output> </layer> <layer id="4664" name="__module.model.gpt_neox.layers.19.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4665" name="__module.model.gpt_neox.layers.19.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="4666" name="__module.model.gpt_neox.layers.19.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4667" name="__module.model.gpt_neox.layers.19.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4668" name="Maximum_454685" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4669" name="__module.model.gpt_neox.layers.19.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3502,attention_mask.47"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4670" name="__module.model.gpt_neox.layers.19.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3503,3504,sdpa_result.39,tensor.77"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4671" name="Constant_11692" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3505"> <dim>4</dim> </port> </output> </layer> <layer id="4672" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3506"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4673" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4674" name="__module.model.gpt_neox.layers.19.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="3511,input.311"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4675" name="Constant_11710" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="9348088344" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4676" name="Convert_549593" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4677" name="Constant_11710/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="9385837080" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4678" name="Convert_549596" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4679" name="Subtract_549597" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4680" name="Constant_549598" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="9385843224" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4681" name="Constant_11710/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4682" name="__module.model.gpt_neox.layers.19.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4683" name="Constant_466985" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9234602520" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4684" name="__module.model.gpt_neox.layers.19.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3514,3516,input.313"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4685" name="__module.model.gpt_neox.layers.19/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3542"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4686" name="__module.model.gpt_neox.layers.19/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3543,3545,input.323"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4687" name="__module.model.gpt_neox.layers.20.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4688" name="__module.model.gpt_neox.layers.20.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4689" name="Constant_466986" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9385867800" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4690" name="__module.model.gpt_neox.layers.20.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4691" name="Constant_466987" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9385892376" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4692" name="__module.model.gpt_neox.layers.20.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3680,input.331"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4693" name="Constant_12310" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="9385916952" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4694" name="Convert_549638" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4695" name="Constant_12310/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="9536911896" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4696" name="Convert_549641" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4697" name="Subtract_549642" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4698" name="Constant_549643" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="9536936472" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4699" name="Constant_12310/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4700" name="__module.model.gpt_neox.layers.20.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4701" name="Constant_466988" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="9537034776" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4702" name="__module.model.gpt_neox.layers.20.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="3685,input.333"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4703" name="__module.model.gpt_neox.layers.20.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="3694,input.335"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4704" name="Constant_12330" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="9537133080" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4705" name="Convert_549647" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4706" name="Constant_12330/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="9688128024" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4707" name="Convert_549650" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4708" name="Subtract_549651" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4709" name="Constant_549652" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="9688134168" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4710" name="Constant_12330/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4711" name="__module.model.gpt_neox.layers.20.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4712" name="Constant_466989" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9688158744" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4713" name="__module.model.gpt_neox.layers.20.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3697,input.337"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4714" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4715" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4716" name="Constant_466990" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9688183320" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4717" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4718" name="Constant_466991" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9688207896" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4719" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3555,input.325"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4720" name="Constant_11789" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="9688232472" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4721" name="Convert_549620" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4722" name="Constant_11789/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="9801478680" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4723" name="Convert_549623" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4724" name="Subtract_549624" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4725" name="Constant_549625" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="9801497112" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4726" name="Constant_11789/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4727" name="__module.model.gpt_neox.layers.20.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4728" name="Constant_466992" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="9801570840" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4729" name="__module.model.gpt_neox.layers.20.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="3562,qkv.81"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4730" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4731" name="__module.model.gpt_neox.layers.20.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3566,qkv.83"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="4732" name="Constant_430857" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4733" name="Constant_430860" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4734" name="Constant_430863" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4735" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3567"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4736" name="Constant_11817" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3568"> <dim>4</dim> </port> </output> </layer> <layer id="4737" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3569,query.161"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4738" name="Constant_430869" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4739" name="Constant_430872" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4740" name="Constant_430875" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4741" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3576,q.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4742" name="Constant_11893" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="3588,cos_cached.41"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4743" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4744" name="Constant_430881" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4745" name="Constant_430884" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4746" name="Constant_430887" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4747" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3570"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4748" name="Constant_11831" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3571"> <dim>4</dim> </port> </output> </layer> <layer id="4749" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3572,key.201"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4750" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4751" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4752" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3580" /> </output> </layer> <layer id="4753" name="Constant_535524" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4754" name="Constant_535517" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4755" name="Constant_535519" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4756" name="Constant_535521" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4757" name="Concat_535522" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4758" name="Broadcast_535525" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4759" name="ReadValue_522355" type="ReadValue" version="opset6"> <data variable_id="past_key_values.20.keypresent.20.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.20.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4760" name="Constant_509662" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4761" name="Gather_509663" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4762" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4763" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4764" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3582" /> </output> </layer> <layer id="4765" name="__module.model.gpt_neox.layers.20.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3581,3585,3586,seq_len.81" /> </output> </layer> <layer id="4766" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4767" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4768" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4769" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3589,3592,cos.81"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4770" name="Constant_52023" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4771" name="__module.model.gpt_neox.layers.20.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3595"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4772" name="__module.model.gpt_neox.layers.20.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3596,cos.83"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4773" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3600"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4774" name="Constant_430892" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4775" name="Constant_430891" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4776" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4777" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4778" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3601,3606" /> </output> </layer> <layer id="4779" name="__module.model.gpt_neox.layers.20.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="4780" name="__module.model.gpt_neox.layers.20.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3603,3604,3608,3609" /> </output> </layer> <layer id="4781" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4782" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4783" name="Constant_430890" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4784" name="ScatterUpdate_430893" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4785" name="Constant_430896" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4786" name="Constant_430899" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4787" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3610,x2.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4788" name="Constant_466993" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="4789" name="__module.model.gpt_neox.layers.20.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3611"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4790" name="Constant_430945" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4791" name="Constant_430950" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4792" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3605,x1.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4793" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3613"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4794" name="Constant_11891" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="3587,sin_cached.41"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4795" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4796" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4797" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3590,3593,sin.81"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4798" name="Constant_52026" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4799" name="__module.model.gpt_neox.layers.20.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3598"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4800" name="__module.model.gpt_neox.layers.20.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3599,sin.83"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4801" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3614"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4802" name="__module.model.gpt_neox.layers.20.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3615,query.163"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4803" name="Constant_430997" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4804" name="Constant_431000" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4805" name="Constant_431003" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4806" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3577,query_pass.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="4807" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3633,3643,query.165,query.167"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4808" name="Constant_431009" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4809" name="Constant_431012" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4810" name="Constant_431015" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4811" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3578,k.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4812" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3616"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4813" name="Constant_431019" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4814" name="Constant_431018" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4815" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4816" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4817" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3617,3622" /> </output> </layer> <layer id="4818" name="__module.model.gpt_neox.layers.20.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="4819" name="__module.model.gpt_neox.layers.20.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3619,3620,3624,3625" /> </output> </layer> <layer id="4820" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4821" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4822" name="Constant_431017" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4823" name="ScatterUpdate_431020" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4824" name="Constant_431023" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4825" name="Constant_431026" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4826" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3626,x2.83"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4827" name="Constant_466994" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="4828" name="__module.model.gpt_neox.layers.20.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3627"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4829" name="Constant_431072" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4830" name="Constant_431077" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4831" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3621,x1.83"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="4832" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3629"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4833" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3630"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4834" name="__module.model.gpt_neox.layers.20.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3631,key.203"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4835" name="Constant_431123" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4836" name="Constant_431126" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4837" name="Constant_431129" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4838" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3579,key_pass.41"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="4839" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3635,key.205"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4840" name="__module.model.gpt_neox.layers.20.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.20.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4841" name="Constant_535537" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4842" name="Constant_535530" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4843" name="Constant_535532" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4844" name="Constant_535534" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4845" name="Concat_535535" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4846" name="Broadcast_535538" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4847" name="ReadValue_522357" type="ReadValue" version="opset6"> <data variable_id="past_key_values.20.valuepresent.20.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.20.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4848" name="Constant_509665" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4849" name="Gather_509666" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4850" name="Constant_431135" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4851" name="Constant_431138" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4852" name="Constant_431141" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4853" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3573"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4854" name="Constant_11845" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3574"> <dim>4</dim> </port> </output> </layer> <layer id="4855" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3575,value.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4856" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.20.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4857" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="3652"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="4858" name="Constant_431145" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4859" name="Constant_431144" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4860" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4861" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4862" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3647" /> </output> </layer> <layer id="4863" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4864" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4865" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3645" /> </output> </layer> <layer id="4866" name="__module.model.gpt_neox.layers.20.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3649,3650" /> </output> </layer> <layer id="4867" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4868" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4869" name="Constant_431143" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4870" name="ScatterUpdate_431146" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4871" name="Constant_431147" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4872" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4873" name="ScatterUpdate_431148" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4874" name="Constant_431151" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4875" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3653"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="4876" name="Constant_431223" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4877" name="Constant_431224" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4878" name="Constant_431220" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4879" name="Constant_431219" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4880" name="ScatterUpdate_431225" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4881" name="Constant_431228" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4882" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3654,3655,causal_mask.121"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4883" name="__module.model.gpt_neox.layers.20.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4884" name="__module.model.gpt_neox.layers.20.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="3642,mask_value.41" /> </output> </layer> <layer id="4885" name="__module.model.gpt_neox.layers.20.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="3656,causal_mask.123"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4886" name="Add_454759" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4887" name="ShapeOf_454760" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4888" name="Constant_402575" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4889" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4890" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3640"> <dim>1</dim> </port> </output> </layer> <layer id="4891" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4892" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4893" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4894" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="3657"> <dim>4</dim> </port> </output> </layer> <layer id="4895" name="__module.model.gpt_neox.layers.20.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4896" name="__module.model.gpt_neox.layers.20.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="4897" name="__module.model.gpt_neox.layers.20.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4898" name="__module.model.gpt_neox.layers.20.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4899" name="Maximum_454761" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4900" name="__module.model.gpt_neox.layers.20.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3659,attention_mask.49"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="4901" name="__module.model.gpt_neox.layers.20.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3660,3661,sdpa_result.41,tensor.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4902" name="Constant_12270" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3662"> <dim>4</dim> </port> </output> </layer> <layer id="4903" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3663"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4904" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="4905" name="__module.model.gpt_neox.layers.20.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="3668,input.327"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4906" name="Constant_12288" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="9801644568" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4907" name="Convert_549629" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4908" name="Constant_12288/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="9839393304" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4909" name="Convert_549632" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4910" name="Subtract_549633" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4911" name="Constant_549634" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="9839399448" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4912" name="Constant_12288/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4913" name="__module.model.gpt_neox.layers.20.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4914" name="Constant_466995" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9688158744" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4915" name="__module.model.gpt_neox.layers.20.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3671,3673,input.329"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4916" name="__module.model.gpt_neox.layers.20/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3699"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4917" name="__module.model.gpt_neox.layers.20/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3700,3702,input.339"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4918" name="__module.model.gpt_neox.layers.21.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4919" name="__module.model.gpt_neox.layers.21.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4920" name="Constant_466996" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9839424024" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4921" name="__module.model.gpt_neox.layers.21.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4922" name="Constant_466997" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="9839448600" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4923" name="__module.model.gpt_neox.layers.21.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3837,input.347"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4924" name="Constant_12888" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="9839473176" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4925" name="Convert_549674" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4926" name="Constant_12888/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="9990468120" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4927" name="Convert_549677" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4928" name="Subtract_549678" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4929" name="Constant_549679" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="9990492696" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="4930" name="Constant_12888/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4931" name="__module.model.gpt_neox.layers.21.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4932" name="Constant_466998" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="9990591000" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4933" name="__module.model.gpt_neox.layers.21.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="3842,input.349"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4934" name="__module.model.gpt_neox.layers.21.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="3851,input.351"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4935" name="Constant_12908" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="9990689304" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4936" name="Convert_549683" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4937" name="Constant_12908/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="10141684248" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4938" name="Convert_549686" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4939" name="Subtract_549687" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4940" name="Constant_549688" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="10141690392" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="4941" name="Constant_12908/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="4942" name="__module.model.gpt_neox.layers.21.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4943" name="Constant_466999" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10141714968" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4944" name="__module.model.gpt_neox.layers.21.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3854,input.353"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4945" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4946" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4947" name="Constant_467000" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10141739544" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4948" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4949" name="Constant_467001" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10141764120" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4950" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3712,input.341"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4951" name="Constant_12367" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="10141788696" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4952" name="Convert_549656" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4953" name="Constant_12367/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="10255034904" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4954" name="Convert_549659" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4955" name="Subtract_549660" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4956" name="Constant_549661" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="10255053336" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="4957" name="Constant_12367/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="4958" name="__module.model.gpt_neox.layers.21.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4959" name="Constant_467002" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="10255127064" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4960" name="__module.model.gpt_neox.layers.21.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="3719,qkv.85"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="4961" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4962" name="__module.model.gpt_neox.layers.21.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3723,qkv.87"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="4963" name="Constant_431236" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4964" name="Constant_431239" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4965" name="Constant_431242" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4966" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3724"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4967" name="Constant_12395" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3725"> <dim>4</dim> </port> </output> </layer> <layer id="4968" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3726,query.169"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4969" name="Constant_431248" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4970" name="Constant_431251" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4971" name="Constant_431254" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4972" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3733,q.43"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="4973" name="Constant_12471" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="3745,cos_cached.43"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="4974" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4975" name="Constant_431260" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4976" name="Constant_431263" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4977" name="Constant_431266" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4978" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3727"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="4979" name="Constant_12409" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3728"> <dim>4</dim> </port> </output> </layer> <layer id="4980" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3729,key.211"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4981" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4982" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4983" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3737" /> </output> </layer> <layer id="4984" name="Constant_535550" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="4985" name="Constant_535543" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4986" name="Constant_535545" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4987" name="Constant_535547" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="4988" name="Concat_535548" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="4989" name="Broadcast_535551" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="4990" name="ReadValue_522359" type="ReadValue" version="opset6"> <data variable_id="past_key_values.21.keypresent.21.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.21.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4991" name="Constant_509668" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="4992" name="Gather_509669" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="4993" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="4994" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4995" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3739" /> </output> </layer> <layer id="4996" name="__module.model.gpt_neox.layers.21.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3738,3742,3743,seq_len.85" /> </output> </layer> <layer id="4997" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="4998" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="4999" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5000" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3746,3749,cos.85"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5001" name="Constant_52029" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5002" name="__module.model.gpt_neox.layers.21.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3752"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5003" name="__module.model.gpt_neox.layers.21.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3753,cos.87"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5004" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3757"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5005" name="Constant_431271" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5006" name="Constant_431270" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5007" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5008" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5009" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3758,3763" /> </output> </layer> <layer id="5010" name="__module.model.gpt_neox.layers.21.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5011" name="__module.model.gpt_neox.layers.21.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3760,3761,3765,3766" /> </output> </layer> <layer id="5012" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5013" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5014" name="Constant_431269" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5015" name="ScatterUpdate_431272" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5016" name="Constant_431275" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5017" name="Constant_431278" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5018" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3767,x2.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5019" name="Constant_467003" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5020" name="__module.model.gpt_neox.layers.21.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3768"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5021" name="Constant_431324" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5022" name="Constant_431329" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5023" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3762,x1.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5024" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3770"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5025" name="Constant_12469" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="3744,sin_cached.43"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5026" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5027" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5028" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3747,3750,sin.85"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5029" name="Constant_52032" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5030" name="__module.model.gpt_neox.layers.21.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3755"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5031" name="__module.model.gpt_neox.layers.21.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3756,sin.87"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5032" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3771"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5033" name="__module.model.gpt_neox.layers.21.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3772,query.171"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5034" name="Constant_431376" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5035" name="Constant_431379" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5036" name="Constant_431382" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5037" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3734,query_pass.43"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5038" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3790,3800,query.173,query.175"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5039" name="Constant_431388" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5040" name="Constant_431391" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5041" name="Constant_431394" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5042" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3735,k.43"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5043" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3773"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5044" name="Constant_431398" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5045" name="Constant_431397" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5046" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5047" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5048" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3774,3779" /> </output> </layer> <layer id="5049" name="__module.model.gpt_neox.layers.21.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5050" name="__module.model.gpt_neox.layers.21.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3776,3777,3781,3782" /> </output> </layer> <layer id="5051" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5052" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5053" name="Constant_431396" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5054" name="ScatterUpdate_431399" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5055" name="Constant_431402" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5056" name="Constant_431405" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5057" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3783,x2.87"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5058" name="Constant_467004" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5059" name="__module.model.gpt_neox.layers.21.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3784"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5060" name="Constant_431451" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5061" name="Constant_431456" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5062" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3778,x1.87"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5063" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3786"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5064" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3787"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5065" name="__module.model.gpt_neox.layers.21.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3788,key.213"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5066" name="Constant_431502" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5067" name="Constant_431505" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5068" name="Constant_431508" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5069" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3736,key_pass.43"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5070" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3792,key.215"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5071" name="__module.model.gpt_neox.layers.21.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.21.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5072" name="Constant_535563" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5073" name="Constant_535556" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5074" name="Constant_535558" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5075" name="Constant_535560" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5076" name="Concat_535561" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5077" name="Broadcast_535564" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="5078" name="ReadValue_522361" type="ReadValue" version="opset6"> <data variable_id="past_key_values.21.valuepresent.21.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.21.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5079" name="Constant_509671" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="5080" name="Gather_509672" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5081" name="Constant_431514" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5082" name="Constant_431517" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5083" name="Constant_431520" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5084" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3730"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5085" name="Constant_12423" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3731"> <dim>4</dim> </port> </output> </layer> <layer id="5086" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3732,value.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5087" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.21.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5088" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="3809"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="5089" name="Constant_431524" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5090" name="Constant_431523" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5091" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5092" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5093" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3804" /> </output> </layer> <layer id="5094" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5095" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5096" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3802" /> </output> </layer> <layer id="5097" name="__module.model.gpt_neox.layers.21.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3806,3807" /> </output> </layer> <layer id="5098" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5099" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5100" name="Constant_431522" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5101" name="ScatterUpdate_431525" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5102" name="Constant_431526" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5103" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5104" name="ScatterUpdate_431527" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5105" name="Constant_431530" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5106" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3810"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="5107" name="Constant_431602" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5108" name="Constant_431603" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5109" name="Constant_431599" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5110" name="Constant_431598" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5111" name="ScatterUpdate_431604" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5112" name="Constant_431607" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5113" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3811,3812,causal_mask.127"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5114" name="__module.model.gpt_neox.layers.21.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5115" name="__module.model.gpt_neox.layers.21.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="3799,mask_value.43" /> </output> </layer> <layer id="5116" name="__module.model.gpt_neox.layers.21.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="3813,causal_mask.129"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5117" name="Add_454835" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5118" name="ShapeOf_454836" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5119" name="Constant_402590" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5120" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5121" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3797"> <dim>1</dim> </port> </output> </layer> <layer id="5122" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5123" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5124" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5125" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="3814"> <dim>4</dim> </port> </output> </layer> <layer id="5126" name="__module.model.gpt_neox.layers.21.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5127" name="__module.model.gpt_neox.layers.21.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="5128" name="__module.model.gpt_neox.layers.21.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5129" name="__module.model.gpt_neox.layers.21.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5130" name="Maximum_454837" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5131" name="__module.model.gpt_neox.layers.21.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3816,attention_mask.51"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5132" name="__module.model.gpt_neox.layers.21.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3817,3818,sdpa_result.43,tensor.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5133" name="Constant_12848" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3819"> <dim>4</dim> </port> </output> </layer> <layer id="5134" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3820"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5135" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5136" name="__module.model.gpt_neox.layers.21.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="3825,input.343"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5137" name="Constant_12866" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="10255200792" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5138" name="Convert_549665" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5139" name="Constant_12866/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="10292949528" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5140" name="Convert_549668" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5141" name="Subtract_549669" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5142" name="Constant_549670" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="10292955672" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5143" name="Constant_12866/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5144" name="__module.model.gpt_neox.layers.21.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5145" name="Constant_467005" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10141714968" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5146" name="__module.model.gpt_neox.layers.21.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3828,3830,input.345"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5147" name="__module.model.gpt_neox.layers.21/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3856"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5148" name="__module.model.gpt_neox.layers.21/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3857,3859,input.355"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5149" name="__module.model.gpt_neox.layers.22.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5150" name="__module.model.gpt_neox.layers.22.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5151" name="Constant_467006" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10292980248" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5152" name="__module.model.gpt_neox.layers.22.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5153" name="Constant_467007" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10293004824" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5154" name="__module.model.gpt_neox.layers.22.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3994,input.363"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5155" name="Constant_13466" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="10293029400" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5156" name="Convert_549710" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5157" name="Constant_13466/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="10444024344" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5158" name="Convert_549713" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5159" name="Subtract_549714" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5160" name="Constant_549715" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="10444048920" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5161" name="Constant_13466/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5162" name="__module.model.gpt_neox.layers.22.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5163" name="Constant_467008" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="10444147224" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5164" name="__module.model.gpt_neox.layers.22.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="3999,input.365"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5165" name="__module.model.gpt_neox.layers.22.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="4008,input.367"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5166" name="Constant_13486" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="10444245528" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5167" name="Convert_549719" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5168" name="Constant_13486/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="10595240472" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5169" name="Convert_549722" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5170" name="Subtract_549723" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5171" name="Constant_549724" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="10595246616" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5172" name="Constant_13486/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5173" name="__module.model.gpt_neox.layers.22.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5174" name="Constant_467009" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10595271192" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5175" name="__module.model.gpt_neox.layers.22.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4011,input.369"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5176" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5177" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5178" name="Constant_467010" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10595295768" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5179" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5180" name="Constant_467011" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10595320344" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5181" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3869,input.357"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5182" name="Constant_12945" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="10595344920" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5183" name="Convert_549692" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5184" name="Constant_12945/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="10708591128" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5185" name="Convert_549695" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5186" name="Subtract_549696" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5187" name="Constant_549697" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="10708609560" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5188" name="Constant_12945/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5189" name="__module.model.gpt_neox.layers.22.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5190" name="Constant_467012" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="10708683288" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5191" name="__module.model.gpt_neox.layers.22.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="3876,qkv.89"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5192" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5193" name="__module.model.gpt_neox.layers.22.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3880,qkv.91"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="5194" name="Constant_431615" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5195" name="Constant_431618" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5196" name="Constant_431621" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5197" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3881"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5198" name="Constant_12973" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3882"> <dim>4</dim> </port> </output> </layer> <layer id="5199" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3883,query.177"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5200" name="Constant_431627" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5201" name="Constant_431630" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5202" name="Constant_431633" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5203" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3890,q.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5204" name="Constant_13049" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="3902,cos_cached.45"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5205" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5206" name="Constant_431639" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5207" name="Constant_431642" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5208" name="Constant_431645" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5209" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3884"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5210" name="Constant_12987" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3885"> <dim>4</dim> </port> </output> </layer> <layer id="5211" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3886,key.221"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5212" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5213" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5214" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3894" /> </output> </layer> <layer id="5215" name="Constant_535576" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5216" name="Constant_535569" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5217" name="Constant_535571" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5218" name="Constant_535573" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5219" name="Concat_535574" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5220" name="Broadcast_535577" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="5221" name="ReadValue_522363" type="ReadValue" version="opset6"> <data variable_id="past_key_values.22.keypresent.22.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.22.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5222" name="Constant_509674" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="5223" name="Gather_509675" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5224" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5225" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5226" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3896" /> </output> </layer> <layer id="5227" name="__module.model.gpt_neox.layers.22.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3895,3899,3900,seq_len.89" /> </output> </layer> <layer id="5228" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5229" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5230" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5231" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3903,3906,cos.89"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5232" name="Constant_52035" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5233" name="__module.model.gpt_neox.layers.22.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3909"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5234" name="__module.model.gpt_neox.layers.22.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3910,cos.91"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5235" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3914"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5236" name="Constant_431650" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5237" name="Constant_431649" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5238" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5239" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5240" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3915,3920" /> </output> </layer> <layer id="5241" name="__module.model.gpt_neox.layers.22.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5242" name="__module.model.gpt_neox.layers.22.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3917,3918,3922,3923" /> </output> </layer> <layer id="5243" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5244" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5245" name="Constant_431648" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5246" name="ScatterUpdate_431651" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5247" name="Constant_431654" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5248" name="Constant_431657" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5249" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3924,x2.89"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5250" name="Constant_467013" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5251" name="__module.model.gpt_neox.layers.22.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3925"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5252" name="Constant_431703" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5253" name="Constant_431708" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5254" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3919,x1.89"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5255" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3927"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5256" name="Constant_13047" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="3901,sin_cached.45"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5257" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5258" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5259" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3904,3907,sin.89"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5260" name="Constant_52038" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5261" name="__module.model.gpt_neox.layers.22.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="3912"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5262" name="__module.model.gpt_neox.layers.22.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="3913,sin.91"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5263" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3928"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5264" name="__module.model.gpt_neox.layers.22.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3929,query.179"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5265" name="Constant_431755" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5266" name="Constant_431758" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5267" name="Constant_431761" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5268" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3891,query_pass.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5269" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3947,3957,query.181,query.183"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5270" name="Constant_431767" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5271" name="Constant_431770" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5272" name="Constant_431773" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5273" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3892,k.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5274" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3930"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5275" name="Constant_431777" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5276" name="Constant_431776" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5277" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5278" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5279" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3931,3936" /> </output> </layer> <layer id="5280" name="__module.model.gpt_neox.layers.22.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5281" name="__module.model.gpt_neox.layers.22.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="3933,3934,3938,3939" /> </output> </layer> <layer id="5282" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5283" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5284" name="Constant_431775" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5285" name="ScatterUpdate_431778" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5286" name="Constant_431781" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5287" name="Constant_431784" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5288" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3940,x2.91"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5289" name="Constant_467014" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5290" name="__module.model.gpt_neox.layers.22.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="3941"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5291" name="Constant_431830" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5292" name="Constant_431835" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5293" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3935,x1.91"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5294" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="3943"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5295" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3944"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5296" name="__module.model.gpt_neox.layers.22.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="3945,key.223"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5297" name="Constant_431881" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5298" name="Constant_431884" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5299" name="Constant_431887" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5300" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3893,key_pass.45"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5301" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="3949,key.225"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5302" name="__module.model.gpt_neox.layers.22.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.22.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5303" name="Constant_535589" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5304" name="Constant_535582" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5305" name="Constant_535584" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5306" name="Constant_535586" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5307" name="Concat_535587" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5308" name="Broadcast_535590" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="5309" name="ReadValue_522365" type="ReadValue" version="opset6"> <data variable_id="past_key_values.22.valuepresent.22.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.22.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5310" name="Constant_509677" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="5311" name="Gather_509678" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5312" name="Constant_431893" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5313" name="Constant_431896" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5314" name="Constant_431899" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5315" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="3887"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5316" name="Constant_13001" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3888"> <dim>4</dim> </port> </output> </layer> <layer id="5317" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3889,value.89"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5318" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.22.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5319" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="3966"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="5320" name="Constant_431903" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5321" name="Constant_431902" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5322" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5323" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5324" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3961" /> </output> </layer> <layer id="5325" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5326" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5327" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3959" /> </output> </layer> <layer id="5328" name="__module.model.gpt_neox.layers.22.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="3963,3964" /> </output> </layer> <layer id="5329" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5330" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5331" name="Constant_431901" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5332" name="ScatterUpdate_431904" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5333" name="Constant_431905" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5334" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5335" name="ScatterUpdate_431906" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5336" name="Constant_431909" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5337" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3967"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="5338" name="Constant_431981" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5339" name="Constant_431982" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5340" name="Constant_431978" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5341" name="Constant_431977" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5342" name="ScatterUpdate_431983" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5343" name="Constant_431986" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5344" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="3968,3969,causal_mask.133"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5345" name="__module.model.gpt_neox.layers.22.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5346" name="__module.model.gpt_neox.layers.22.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="3956,mask_value.45" /> </output> </layer> <layer id="5347" name="__module.model.gpt_neox.layers.22.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="3970,causal_mask.135"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5348" name="Add_454911" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5349" name="ShapeOf_454912" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5350" name="Constant_402605" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5351" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5352" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="3954"> <dim>1</dim> </port> </output> </layer> <layer id="5353" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5354" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5355" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5356" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="3971"> <dim>4</dim> </port> </output> </layer> <layer id="5357" name="__module.model.gpt_neox.layers.22.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5358" name="__module.model.gpt_neox.layers.22.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="5359" name="__module.model.gpt_neox.layers.22.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5360" name="__module.model.gpt_neox.layers.22.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5361" name="Maximum_454913" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5362" name="__module.model.gpt_neox.layers.22.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3973,attention_mask.53"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5363" name="__module.model.gpt_neox.layers.22.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="3974,3975,sdpa_result.45,tensor.89"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5364" name="Constant_13426" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="3976"> <dim>4</dim> </port> </output> </layer> <layer id="5365" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="3977"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5366" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5367" name="__module.model.gpt_neox.layers.22.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="3982,input.359"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5368" name="Constant_13444" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="10708757016" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5369" name="Convert_549701" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5370" name="Constant_13444/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="10746505752" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5371" name="Convert_549704" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5372" name="Subtract_549705" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5373" name="Constant_549706" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="10746511896" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5374" name="Constant_13444/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5375" name="__module.model.gpt_neox.layers.22.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5376" name="Constant_467015" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10595271192" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5377" name="__module.model.gpt_neox.layers.22.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="3985,3987,input.361"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5378" name="__module.model.gpt_neox.layers.22/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4013"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5379" name="__module.model.gpt_neox.layers.22/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4014,4016,input.371"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5380" name="__module.model.gpt_neox.layers.23.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5381" name="__module.model.gpt_neox.layers.23.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5382" name="Constant_467016" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10746536472" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5383" name="__module.model.gpt_neox.layers.23.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5384" name="Constant_467017" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="10746561048" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5385" name="__module.model.gpt_neox.layers.23.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4151,input.379"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5386" name="Constant_14044" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="10746585624" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5387" name="Convert_549746" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5388" name="Constant_14044/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="10897580568" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5389" name="Convert_549749" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5390" name="Subtract_549750" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5391" name="Constant_549751" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="10897605144" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5392" name="Constant_14044/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5393" name="__module.model.gpt_neox.layers.23.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5394" name="Constant_467018" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="10897703448" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5395" name="__module.model.gpt_neox.layers.23.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="4156,input.381"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5396" name="__module.model.gpt_neox.layers.23.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="4165,input.383"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5397" name="Constant_14064" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="10897801752" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5398" name="Convert_549755" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5399" name="Constant_14064/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="11048796696" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5400" name="Convert_549758" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5401" name="Subtract_549759" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5402" name="Constant_549760" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="11048802840" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5403" name="Constant_14064/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5404" name="__module.model.gpt_neox.layers.23.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5405" name="Constant_467019" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11048827416" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5406" name="__module.model.gpt_neox.layers.23.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4168,input.385"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5407" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5408" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5409" name="Constant_467020" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11048851992" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5410" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5411" name="Constant_467021" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11048876568" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5412" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4026,input.373"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5413" name="Constant_13523" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="11048901144" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5414" name="Convert_549728" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5415" name="Constant_13523/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="11162147352" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5416" name="Convert_549731" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5417" name="Subtract_549732" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5418" name="Constant_549733" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="11162165784" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5419" name="Constant_13523/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5420" name="__module.model.gpt_neox.layers.23.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5421" name="Constant_467022" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="11162239512" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5422" name="__module.model.gpt_neox.layers.23.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="4033,qkv.93"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5423" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5424" name="__module.model.gpt_neox.layers.23.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4037,qkv.95"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="5425" name="Constant_431994" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5426" name="Constant_431997" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5427" name="Constant_432000" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5428" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4038"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5429" name="Constant_13551" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4039"> <dim>4</dim> </port> </output> </layer> <layer id="5430" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4040,query.185"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5431" name="Constant_432006" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5432" name="Constant_432009" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5433" name="Constant_432012" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5434" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4047,q.47"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5435" name="Constant_13627" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="4059,cos_cached.47"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5436" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5437" name="Constant_432018" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5438" name="Constant_432021" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5439" name="Constant_432024" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5440" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4041"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5441" name="Constant_13565" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4042"> <dim>4</dim> </port> </output> </layer> <layer id="5442" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4043,key.231"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5443" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5444" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5445" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4051" /> </output> </layer> <layer id="5446" name="Constant_535602" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5447" name="Constant_535595" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5448" name="Constant_535597" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5449" name="Constant_535599" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5450" name="Concat_535600" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5451" name="Broadcast_535603" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="5452" name="ReadValue_522367" type="ReadValue" version="opset6"> <data variable_id="past_key_values.23.keypresent.23.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.23.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5453" name="Constant_509680" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="5454" name="Gather_509681" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5455" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5456" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5457" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4053" /> </output> </layer> <layer id="5458" name="__module.model.gpt_neox.layers.23.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4052,4056,4057,seq_len.93" /> </output> </layer> <layer id="5459" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5460" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5461" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5462" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4060,4063,cos.93"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5463" name="Constant_52041" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5464" name="__module.model.gpt_neox.layers.23.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4066"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5465" name="__module.model.gpt_neox.layers.23.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4067,cos.95"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5466" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4071"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5467" name="Constant_432029" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5468" name="Constant_432028" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5469" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5470" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5471" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4072,4077" /> </output> </layer> <layer id="5472" name="__module.model.gpt_neox.layers.23.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5473" name="__module.model.gpt_neox.layers.23.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4074,4075,4079,4080" /> </output> </layer> <layer id="5474" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5475" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5476" name="Constant_432027" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5477" name="ScatterUpdate_432030" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5478" name="Constant_432033" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5479" name="Constant_432036" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5480" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4081,x2.93"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5481" name="Constant_467023" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5482" name="__module.model.gpt_neox.layers.23.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4082"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5483" name="Constant_432082" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5484" name="Constant_432087" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5485" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4076,x1.93"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5486" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4084"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5487" name="Constant_13625" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="4058,sin_cached.47"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5488" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5489" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5490" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4061,4064,sin.93"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5491" name="Constant_52044" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5492" name="__module.model.gpt_neox.layers.23.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4069"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5493" name="__module.model.gpt_neox.layers.23.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4070,sin.95"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5494" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4085"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5495" name="__module.model.gpt_neox.layers.23.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4086,query.187"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5496" name="Constant_432134" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5497" name="Constant_432137" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5498" name="Constant_432140" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5499" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4048,query_pass.47"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5500" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4104,4114,query.189,query.191"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5501" name="Constant_432146" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5502" name="Constant_432149" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5503" name="Constant_432152" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5504" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4049,k.47"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5505" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4087"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5506" name="Constant_432156" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5507" name="Constant_432155" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5508" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5509" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5510" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4088,4093" /> </output> </layer> <layer id="5511" name="__module.model.gpt_neox.layers.23.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5512" name="__module.model.gpt_neox.layers.23.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4090,4091,4095,4096" /> </output> </layer> <layer id="5513" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5514" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5515" name="Constant_432154" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5516" name="ScatterUpdate_432157" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5517" name="Constant_432160" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5518" name="Constant_432163" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5519" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4097,x2.95"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5520" name="Constant_467024" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5521" name="__module.model.gpt_neox.layers.23.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4098"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5522" name="Constant_432209" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5523" name="Constant_432214" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5524" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4092,x1.95"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5525" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4100"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5526" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4101"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5527" name="__module.model.gpt_neox.layers.23.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4102,key.233"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5528" name="Constant_432260" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5529" name="Constant_432263" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5530" name="Constant_432266" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5531" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4050,key_pass.47"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5532" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4106,key.235"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5533" name="__module.model.gpt_neox.layers.23.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.23.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5534" name="Constant_535615" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5535" name="Constant_535608" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5536" name="Constant_535610" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5537" name="Constant_535612" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5538" name="Concat_535613" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5539" name="Broadcast_535616" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="5540" name="ReadValue_522369" type="ReadValue" version="opset6"> <data variable_id="past_key_values.23.valuepresent.23.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.23.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5541" name="Constant_509683" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="5542" name="Gather_509684" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5543" name="Constant_432272" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5544" name="Constant_432275" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5545" name="Constant_432278" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5546" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4044"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5547" name="Constant_13579" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4045"> <dim>4</dim> </port> </output> </layer> <layer id="5548" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4046,value.93"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5549" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.23.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5550" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="4123"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="5551" name="Constant_432282" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5552" name="Constant_432281" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5553" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5554" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5555" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4118" /> </output> </layer> <layer id="5556" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5557" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5558" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4116" /> </output> </layer> <layer id="5559" name="__module.model.gpt_neox.layers.23.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4120,4121" /> </output> </layer> <layer id="5560" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5561" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5562" name="Constant_432280" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5563" name="ScatterUpdate_432283" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5564" name="Constant_432284" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5565" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5566" name="ScatterUpdate_432285" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5567" name="Constant_432288" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5568" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4124"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="5569" name="Constant_432360" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5570" name="Constant_432361" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5571" name="Constant_432357" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5572" name="Constant_432356" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5573" name="ScatterUpdate_432362" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5574" name="Constant_432365" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5575" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4125,4126,causal_mask.139"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5576" name="__module.model.gpt_neox.layers.23.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5577" name="__module.model.gpt_neox.layers.23.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="4113,mask_value.47" /> </output> </layer> <layer id="5578" name="__module.model.gpt_neox.layers.23.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="4127,causal_mask.141"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5579" name="Add_454987" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5580" name="ShapeOf_454988" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5581" name="Constant_402620" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5582" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5583" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4111"> <dim>1</dim> </port> </output> </layer> <layer id="5584" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5585" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5586" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5587" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="4128"> <dim>4</dim> </port> </output> </layer> <layer id="5588" name="__module.model.gpt_neox.layers.23.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5589" name="__module.model.gpt_neox.layers.23.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="5590" name="__module.model.gpt_neox.layers.23.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5591" name="__module.model.gpt_neox.layers.23.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5592" name="Maximum_454989" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5593" name="__module.model.gpt_neox.layers.23.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4130,attention_mask.55"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5594" name="__module.model.gpt_neox.layers.23.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4131,4132,sdpa_result.47,tensor.93"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5595" name="Constant_14004" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4133"> <dim>4</dim> </port> </output> </layer> <layer id="5596" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4134"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5597" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5598" name="__module.model.gpt_neox.layers.23.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="4139,input.375"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5599" name="Constant_14022" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="11162313240" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5600" name="Convert_549737" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5601" name="Constant_14022/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="11200061976" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5602" name="Convert_549740" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5603" name="Subtract_549741" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5604" name="Constant_549742" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="11200068120" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5605" name="Constant_14022/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5606" name="__module.model.gpt_neox.layers.23.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5607" name="Constant_467025" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11048827416" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5608" name="__module.model.gpt_neox.layers.23.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4142,4144,input.377"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5609" name="__module.model.gpt_neox.layers.23/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4170"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5610" name="__module.model.gpt_neox.layers.23/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4171,4173,input.387"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5611" name="__module.model.gpt_neox.layers.24.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5612" name="__module.model.gpt_neox.layers.24.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5613" name="Constant_467026" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11200092696" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5614" name="__module.model.gpt_neox.layers.24.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5615" name="Constant_467027" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11200117272" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5616" name="__module.model.gpt_neox.layers.24.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4308,input.395"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5617" name="Constant_14622" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="11200141848" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5618" name="Convert_549782" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5619" name="Constant_14622/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="11351136792" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5620" name="Convert_549785" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5621" name="Subtract_549786" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5622" name="Constant_549787" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="11351161368" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5623" name="Constant_14622/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5624" name="__module.model.gpt_neox.layers.24.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5625" name="Constant_467028" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="11351259672" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5626" name="__module.model.gpt_neox.layers.24.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="4313,input.397"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5627" name="__module.model.gpt_neox.layers.24.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="4322,input.399"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5628" name="Constant_14642" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="11351357976" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5629" name="Convert_549791" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5630" name="Constant_14642/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="11502352920" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5631" name="Convert_549794" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5632" name="Subtract_549795" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5633" name="Constant_549796" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="11502359064" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5634" name="Constant_14642/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5635" name="__module.model.gpt_neox.layers.24.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5636" name="Constant_467029" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11502383640" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5637" name="__module.model.gpt_neox.layers.24.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4325,input.401"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5638" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5639" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5640" name="Constant_467030" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11502408216" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5641" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5642" name="Constant_467031" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11502432792" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5643" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4183,input.389"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5644" name="Constant_14101" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="11502457368" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5645" name="Convert_549764" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5646" name="Constant_14101/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="11615703576" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5647" name="Convert_549767" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5648" name="Subtract_549768" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5649" name="Constant_549769" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="11615722008" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5650" name="Constant_14101/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5651" name="__module.model.gpt_neox.layers.24.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5652" name="Constant_467032" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="11615795736" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5653" name="__module.model.gpt_neox.layers.24.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="4190,qkv.97"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5654" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5655" name="__module.model.gpt_neox.layers.24.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4194,qkv.99"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="5656" name="Constant_432373" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5657" name="Constant_432376" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5658" name="Constant_432379" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5659" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4195"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5660" name="Constant_14129" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4196"> <dim>4</dim> </port> </output> </layer> <layer id="5661" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4197,query.193"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5662" name="Constant_432385" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5663" name="Constant_432388" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5664" name="Constant_432391" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5665" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4204,q.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5666" name="Constant_14205" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="4216,cos_cached.49"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5667" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5668" name="Constant_432397" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5669" name="Constant_432400" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5670" name="Constant_432403" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5671" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4198"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5672" name="Constant_14143" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4199"> <dim>4</dim> </port> </output> </layer> <layer id="5673" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4200,key.241"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5674" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5675" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5676" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4208" /> </output> </layer> <layer id="5677" name="Constant_535628" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5678" name="Constant_535621" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5679" name="Constant_535623" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5680" name="Constant_535625" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5681" name="Concat_535626" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5682" name="Broadcast_535629" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="5683" name="ReadValue_522371" type="ReadValue" version="opset6"> <data variable_id="past_key_values.24.keypresent.24.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.24.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5684" name="Constant_509686" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="5685" name="Gather_509687" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5686" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5687" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5688" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4210" /> </output> </layer> <layer id="5689" name="__module.model.gpt_neox.layers.24.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4209,4213,4214,seq_len.97" /> </output> </layer> <layer id="5690" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5691" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5692" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5693" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4217,4220,cos.97"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5694" name="Constant_52047" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5695" name="__module.model.gpt_neox.layers.24.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4223"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5696" name="__module.model.gpt_neox.layers.24.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4224,cos.99"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5697" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4228"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5698" name="Constant_432408" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5699" name="Constant_432407" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5700" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5701" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5702" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4229,4234" /> </output> </layer> <layer id="5703" name="__module.model.gpt_neox.layers.24.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5704" name="__module.model.gpt_neox.layers.24.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4231,4232,4236,4237" /> </output> </layer> <layer id="5705" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5706" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5707" name="Constant_432406" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5708" name="ScatterUpdate_432409" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5709" name="Constant_432412" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5710" name="Constant_432415" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5711" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4238,x2.97"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5712" name="Constant_467033" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5713" name="__module.model.gpt_neox.layers.24.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4239"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5714" name="Constant_432461" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5715" name="Constant_432466" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5716" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4233,x1.97"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5717" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4241"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5718" name="Constant_14203" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="4215,sin_cached.49"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5719" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5720" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5721" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4218,4221,sin.97"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5722" name="Constant_52050" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5723" name="__module.model.gpt_neox.layers.24.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4226"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5724" name="__module.model.gpt_neox.layers.24.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4227,sin.99"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5725" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4242"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5726" name="__module.model.gpt_neox.layers.24.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4243,query.195"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5727" name="Constant_432513" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5728" name="Constant_432516" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5729" name="Constant_432519" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5730" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4205,query_pass.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5731" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4261,4271,query.197,query.199"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5732" name="Constant_432525" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5733" name="Constant_432528" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5734" name="Constant_432531" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5735" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4206,k.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5736" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4244"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5737" name="Constant_432535" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5738" name="Constant_432534" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5739" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5740" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5741" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4245,4250" /> </output> </layer> <layer id="5742" name="__module.model.gpt_neox.layers.24.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5743" name="__module.model.gpt_neox.layers.24.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4247,4248,4252,4253" /> </output> </layer> <layer id="5744" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5745" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5746" name="Constant_432533" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5747" name="ScatterUpdate_432536" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5748" name="Constant_432539" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5749" name="Constant_432542" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5750" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4254,x2.99"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5751" name="Constant_467034" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5752" name="__module.model.gpt_neox.layers.24.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4255"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5753" name="Constant_432588" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5754" name="Constant_432593" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5755" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4249,x1.99"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5756" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4257"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5757" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4258"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5758" name="__module.model.gpt_neox.layers.24.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4259,key.243"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5759" name="Constant_432639" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5760" name="Constant_432642" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5761" name="Constant_432645" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5762" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4207,key_pass.49"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5763" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4263,key.245"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5764" name="__module.model.gpt_neox.layers.24.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.24.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5765" name="Constant_535641" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5766" name="Constant_535634" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5767" name="Constant_535636" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5768" name="Constant_535638" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5769" name="Concat_535639" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5770" name="Broadcast_535642" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="5771" name="ReadValue_522373" type="ReadValue" version="opset6"> <data variable_id="past_key_values.24.valuepresent.24.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.24.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5772" name="Constant_509689" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="5773" name="Gather_509690" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5774" name="Constant_432651" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5775" name="Constant_432654" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5776" name="Constant_432657" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5777" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4201"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5778" name="Constant_14157" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4202"> <dim>4</dim> </port> </output> </layer> <layer id="5779" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4203,value.97"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5780" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.24.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5781" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="4280"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="5782" name="Constant_432661" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5783" name="Constant_432660" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5784" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5785" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5786" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4275" /> </output> </layer> <layer id="5787" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5788" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5789" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4273" /> </output> </layer> <layer id="5790" name="__module.model.gpt_neox.layers.24.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4277,4278" /> </output> </layer> <layer id="5791" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5792" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5793" name="Constant_432659" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5794" name="ScatterUpdate_432662" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5795" name="Constant_432663" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5796" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5797" name="ScatterUpdate_432664" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5798" name="Constant_432667" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5799" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4281"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="5800" name="Constant_432739" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5801" name="Constant_432740" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5802" name="Constant_432736" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5803" name="Constant_432735" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5804" name="ScatterUpdate_432741" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5805" name="Constant_432744" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5806" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4282,4283,causal_mask.145"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5807" name="__module.model.gpt_neox.layers.24.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5808" name="__module.model.gpt_neox.layers.24.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="4270,mask_value.49" /> </output> </layer> <layer id="5809" name="__module.model.gpt_neox.layers.24.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="4284,causal_mask.147"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5810" name="Add_455063" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5811" name="ShapeOf_455064" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5812" name="Constant_402635" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5813" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5814" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4268"> <dim>1</dim> </port> </output> </layer> <layer id="5815" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5816" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5817" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5818" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="4285"> <dim>4</dim> </port> </output> </layer> <layer id="5819" name="__module.model.gpt_neox.layers.24.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5820" name="__module.model.gpt_neox.layers.24.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="5821" name="__module.model.gpt_neox.layers.24.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5822" name="__module.model.gpt_neox.layers.24.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5823" name="Maximum_455065" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5824" name="__module.model.gpt_neox.layers.24.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4287,attention_mask.57"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="5825" name="__module.model.gpt_neox.layers.24.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4288,4289,sdpa_result.49,tensor.97"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5826" name="Constant_14582" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4290"> <dim>4</dim> </port> </output> </layer> <layer id="5827" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4291"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5828" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="5829" name="__module.model.gpt_neox.layers.24.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="4296,input.391"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5830" name="Constant_14600" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="11615869464" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5831" name="Convert_549773" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5832" name="Constant_14600/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="11653618200" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5833" name="Convert_549776" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5834" name="Subtract_549777" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5835" name="Constant_549778" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="11653624344" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5836" name="Constant_14600/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5837" name="__module.model.gpt_neox.layers.24.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5838" name="Constant_467035" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11502383640" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5839" name="__module.model.gpt_neox.layers.24.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4299,4301,input.393"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5840" name="__module.model.gpt_neox.layers.24/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4327"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5841" name="__module.model.gpt_neox.layers.24/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4328,4330,input.403"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5842" name="__module.model.gpt_neox.layers.25.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5843" name="__module.model.gpt_neox.layers.25.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5844" name="Constant_467036" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11653648920" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5845" name="__module.model.gpt_neox.layers.25.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5846" name="Constant_467037" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11653673496" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5847" name="__module.model.gpt_neox.layers.25.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4465,input.411"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5848" name="Constant_15200" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="11653698072" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5849" name="Convert_549818" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5850" name="Constant_15200/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="11804693016" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5851" name="Convert_549821" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5852" name="Subtract_549822" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5853" name="Constant_549823" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="11804717592" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="5854" name="Constant_15200/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5855" name="__module.model.gpt_neox.layers.25.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5856" name="Constant_467038" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="11804815896" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5857" name="__module.model.gpt_neox.layers.25.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="4470,input.413"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5858" name="__module.model.gpt_neox.layers.25.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="4479,input.415"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5859" name="Constant_15220" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="11804914200" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5860" name="Convert_549827" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5861" name="Constant_15220/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="11955909144" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5862" name="Convert_549830" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5863" name="Subtract_549831" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5864" name="Constant_549832" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="11955915288" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="5865" name="Constant_15220/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="5866" name="__module.model.gpt_neox.layers.25.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5867" name="Constant_467039" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11955939864" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5868" name="__module.model.gpt_neox.layers.25.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4482,input.417"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5869" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5870" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5871" name="Constant_467040" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11955964440" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5872" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5873" name="Constant_467041" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11955989016" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5874" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4340,input.405"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5875" name="Constant_14679" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="11956013592" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5876" name="Convert_549800" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5877" name="Constant_14679/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="12069259800" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5878" name="Convert_549803" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5879" name="Subtract_549804" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5880" name="Constant_549805" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="12069278232" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="5881" name="Constant_14679/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="5882" name="__module.model.gpt_neox.layers.25.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5883" name="Constant_467042" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="12069351960" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5884" name="__module.model.gpt_neox.layers.25.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="4347,qkv.101"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="5885" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5886" name="__module.model.gpt_neox.layers.25.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4351,qkv.103"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="5887" name="Constant_432752" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5888" name="Constant_432755" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5889" name="Constant_432758" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5890" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4352"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5891" name="Constant_14707" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4353"> <dim>4</dim> </port> </output> </layer> <layer id="5892" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4354,query.201"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5893" name="Constant_432764" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5894" name="Constant_432767" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5895" name="Constant_432770" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5896" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4361,q.51"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5897" name="Constant_14783" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="4373,cos_cached.51"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5898" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5899" name="Constant_432776" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5900" name="Constant_432779" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5901" name="Constant_432782" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5902" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4355"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="5903" name="Constant_14721" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4356"> <dim>4</dim> </port> </output> </layer> <layer id="5904" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4357,key.251"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5905" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5906" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5907" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4365" /> </output> </layer> <layer id="5908" name="Constant_535654" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5909" name="Constant_535647" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5910" name="Constant_535649" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5911" name="Constant_535651" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5912" name="Concat_535652" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5913" name="Broadcast_535655" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="5914" name="ReadValue_522375" type="ReadValue" version="opset6"> <data variable_id="past_key_values.25.keypresent.25.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.25.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5915" name="Constant_509692" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="5916" name="Gather_509693" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5917" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5918" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5919" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4367" /> </output> </layer> <layer id="5920" name="__module.model.gpt_neox.layers.25.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4366,4370,4371,seq_len.101" /> </output> </layer> <layer id="5921" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5922" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5923" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5924" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4374,4377,cos.101"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5925" name="Constant_52053" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5926" name="__module.model.gpt_neox.layers.25.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4380"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5927" name="__module.model.gpt_neox.layers.25.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4381,cos.103"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5928" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4385"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5929" name="Constant_432787" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5930" name="Constant_432786" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5931" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5932" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5933" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4386,4391" /> </output> </layer> <layer id="5934" name="__module.model.gpt_neox.layers.25.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5935" name="__module.model.gpt_neox.layers.25.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4388,4389,4393,4394" /> </output> </layer> <layer id="5936" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5937" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5938" name="Constant_432785" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5939" name="ScatterUpdate_432788" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5940" name="Constant_432791" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5941" name="Constant_432794" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5942" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4395,x2.101"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5943" name="Constant_467043" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5944" name="__module.model.gpt_neox.layers.25.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4396"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5945" name="Constant_432840" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5946" name="Constant_432845" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5947" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4390,x1.101"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5948" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4398"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5949" name="Constant_14781" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="4372,sin_cached.51"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="5950" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5951" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5952" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4375,4378,sin.101"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5953" name="Constant_52056" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5954" name="__module.model.gpt_neox.layers.25.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4383"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5955" name="__module.model.gpt_neox.layers.25.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4384,sin.103"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5956" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4399"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5957" name="__module.model.gpt_neox.layers.25.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4400,query.203"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5958" name="Constant_432892" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5959" name="Constant_432895" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5960" name="Constant_432898" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5961" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4362,query_pass.51"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5962" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4418,4428,query.205,query.207"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5963" name="Constant_432904" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5964" name="Constant_432907" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5965" name="Constant_432910" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5966" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4363,k.51"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5967" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4401"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5968" name="Constant_432914" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5969" name="Constant_432913" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5970" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5971" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5972" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4402,4407" /> </output> </layer> <layer id="5973" name="__module.model.gpt_neox.layers.25.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="5974" name="__module.model.gpt_neox.layers.25.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4404,4405,4409,4410" /> </output> </layer> <layer id="5975" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="5976" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5977" name="Constant_432912" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="5978" name="ScatterUpdate_432915" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5979" name="Constant_432918" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5980" name="Constant_432921" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5981" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4411,x2.103"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5982" name="Constant_467044" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="5983" name="__module.model.gpt_neox.layers.25.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4412"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5984" name="Constant_432967" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5985" name="Constant_432972" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5986" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4406,x1.103"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="5987" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4414"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5988" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4415"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5989" name="__module.model.gpt_neox.layers.25.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4416,key.253"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="5990" name="Constant_433018" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5991" name="Constant_433021" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="5992" name="Constant_433024" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="5993" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4364,key_pass.51"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="5994" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4420,key.255"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5995" name="__module.model.gpt_neox.layers.25.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.25.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="5996" name="Constant_535667" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="5997" name="Constant_535660" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5998" name="Constant_535662" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="5999" name="Constant_535664" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6000" name="Concat_535665" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6001" name="Broadcast_535668" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6002" name="ReadValue_522377" type="ReadValue" version="opset6"> <data variable_id="past_key_values.25.valuepresent.25.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.25.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6003" name="Constant_509695" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6004" name="Gather_509696" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6005" name="Constant_433030" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6006" name="Constant_433033" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6007" name="Constant_433036" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6008" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4358"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6009" name="Constant_14735" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4359"> <dim>4</dim> </port> </output> </layer> <layer id="6010" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4360,value.101"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6011" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.25.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6012" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="4437"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6013" name="Constant_433040" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6014" name="Constant_433039" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6015" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6016" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6017" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4432" /> </output> </layer> <layer id="6018" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6019" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6020" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4430" /> </output> </layer> <layer id="6021" name="__module.model.gpt_neox.layers.25.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4434,4435" /> </output> </layer> <layer id="6022" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6023" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6024" name="Constant_433038" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6025" name="ScatterUpdate_433041" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6026" name="Constant_433042" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6027" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6028" name="ScatterUpdate_433043" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6029" name="Constant_433046" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6030" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4438"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6031" name="Constant_433118" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6032" name="Constant_433119" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6033" name="Constant_433115" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6034" name="Constant_433114" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6035" name="ScatterUpdate_433120" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6036" name="Constant_433123" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6037" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4439,4440,causal_mask.151"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6038" name="__module.model.gpt_neox.layers.25.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6039" name="__module.model.gpt_neox.layers.25.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="4427,mask_value.51" /> </output> </layer> <layer id="6040" name="__module.model.gpt_neox.layers.25.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="4441,causal_mask.153"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6041" name="Add_455139" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6042" name="ShapeOf_455140" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6043" name="Constant_402650" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6044" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6045" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4425"> <dim>1</dim> </port> </output> </layer> <layer id="6046" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6047" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6048" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6049" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="4442"> <dim>4</dim> </port> </output> </layer> <layer id="6050" name="__module.model.gpt_neox.layers.25.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6051" name="__module.model.gpt_neox.layers.25.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="6052" name="__module.model.gpt_neox.layers.25.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6053" name="__module.model.gpt_neox.layers.25.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6054" name="Maximum_455141" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6055" name="__module.model.gpt_neox.layers.25.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4444,attention_mask.59"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6056" name="__module.model.gpt_neox.layers.25.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4445,4446,sdpa_result.51,tensor.101"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6057" name="Constant_15160" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4447"> <dim>4</dim> </port> </output> </layer> <layer id="6058" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4448"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6059" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6060" name="__module.model.gpt_neox.layers.25.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="4453,input.407"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6061" name="Constant_15178" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="12069425688" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6062" name="Convert_549809" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6063" name="Constant_15178/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="12107174424" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6064" name="Convert_549812" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6065" name="Subtract_549813" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6066" name="Constant_549814" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="12107180568" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6067" name="Constant_15178/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6068" name="__module.model.gpt_neox.layers.25.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6069" name="Constant_467045" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="11955939864" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6070" name="__module.model.gpt_neox.layers.25.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4456,4458,input.409"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6071" name="__module.model.gpt_neox.layers.25/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4484"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6072" name="__module.model.gpt_neox.layers.25/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4485,4487,input.419"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6073" name="__module.model.gpt_neox.layers.26.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6074" name="__module.model.gpt_neox.layers.26.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6075" name="Constant_467046" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12107205144" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6076" name="__module.model.gpt_neox.layers.26.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6077" name="Constant_467047" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12107229720" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6078" name="__module.model.gpt_neox.layers.26.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4622,input.427"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6079" name="Constant_15778" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="12107254296" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6080" name="Convert_549854" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6081" name="Constant_15778/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="12258249240" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6082" name="Convert_549857" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6083" name="Subtract_549858" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6084" name="Constant_549859" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="12258273816" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6085" name="Constant_15778/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6086" name="__module.model.gpt_neox.layers.26.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6087" name="Constant_467048" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="12258372120" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6088" name="__module.model.gpt_neox.layers.26.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="4627,input.429"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6089" name="__module.model.gpt_neox.layers.26.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="4636,input.431"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6090" name="Constant_15798" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="12258470424" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6091" name="Convert_549863" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6092" name="Constant_15798/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="12409465368" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6093" name="Convert_549866" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6094" name="Subtract_549867" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6095" name="Constant_549868" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="12409471512" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6096" name="Constant_15798/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6097" name="__module.model.gpt_neox.layers.26.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6098" name="Constant_467049" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12409496088" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6099" name="__module.model.gpt_neox.layers.26.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4639,input.433"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6100" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6101" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6102" name="Constant_467050" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12409520664" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6103" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6104" name="Constant_467051" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12409545240" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6105" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4497,input.421"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6106" name="Constant_15257" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="12409569816" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6107" name="Convert_549836" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6108" name="Constant_15257/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="12522816024" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6109" name="Convert_549839" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6110" name="Subtract_549840" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6111" name="Constant_549841" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="12522834456" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6112" name="Constant_15257/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6113" name="__module.model.gpt_neox.layers.26.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6114" name="Constant_467052" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="12522908184" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6115" name="__module.model.gpt_neox.layers.26.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="4504,qkv.105"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6116" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6117" name="__module.model.gpt_neox.layers.26.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4508,qkv.107"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="6118" name="Constant_433131" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6119" name="Constant_433134" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6120" name="Constant_433137" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6121" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4509"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6122" name="Constant_15285" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4510"> <dim>4</dim> </port> </output> </layer> <layer id="6123" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4511,query.209"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6124" name="Constant_433143" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6125" name="Constant_433146" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6126" name="Constant_433149" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6127" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4518,q.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6128" name="Constant_15361" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="4530,cos_cached.53"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="6129" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6130" name="Constant_433155" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6131" name="Constant_433158" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6132" name="Constant_433161" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6133" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4512"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6134" name="Constant_15299" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4513"> <dim>4</dim> </port> </output> </layer> <layer id="6135" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4514,key.261"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6136" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6137" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6138" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4522" /> </output> </layer> <layer id="6139" name="Constant_535680" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6140" name="Constant_535673" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6141" name="Constant_535675" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6142" name="Constant_535677" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6143" name="Concat_535678" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6144" name="Broadcast_535681" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6145" name="ReadValue_522379" type="ReadValue" version="opset6"> <data variable_id="past_key_values.26.keypresent.26.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.26.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6146" name="Constant_509698" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6147" name="Gather_509699" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6148" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6149" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6150" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4524" /> </output> </layer> <layer id="6151" name="__module.model.gpt_neox.layers.26.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4523,4527,4528,seq_len.105" /> </output> </layer> <layer id="6152" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6153" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6154" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6155" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4531,4534,cos.105"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6156" name="Constant_52059" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6157" name="__module.model.gpt_neox.layers.26.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4537"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6158" name="__module.model.gpt_neox.layers.26.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4538,cos.107"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6159" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4542"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6160" name="Constant_433166" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6161" name="Constant_433165" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6162" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6163" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6164" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4543,4548" /> </output> </layer> <layer id="6165" name="__module.model.gpt_neox.layers.26.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="6166" name="__module.model.gpt_neox.layers.26.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4545,4546,4550,4551" /> </output> </layer> <layer id="6167" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6168" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6169" name="Constant_433164" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6170" name="ScatterUpdate_433167" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6171" name="Constant_433170" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6172" name="Constant_433173" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6173" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4552,x2.105"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6174" name="Constant_467053" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="6175" name="__module.model.gpt_neox.layers.26.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4553"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6176" name="Constant_433219" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6177" name="Constant_433224" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6178" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4547,x1.105"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6179" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4555"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6180" name="Constant_15359" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="4529,sin_cached.53"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="6181" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6182" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6183" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4532,4535,sin.105"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6184" name="Constant_52062" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6185" name="__module.model.gpt_neox.layers.26.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4540"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6186" name="__module.model.gpt_neox.layers.26.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4541,sin.107"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6187" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4556"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6188" name="__module.model.gpt_neox.layers.26.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4557,query.211"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6189" name="Constant_433271" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6190" name="Constant_433274" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6191" name="Constant_433277" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6192" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4519,query_pass.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="6193" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4575,4585,query.213,query.215"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6194" name="Constant_433283" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6195" name="Constant_433286" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6196" name="Constant_433289" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6197" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4520,k.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6198" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4558"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6199" name="Constant_433293" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6200" name="Constant_433292" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6201" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6202" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6203" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4559,4564" /> </output> </layer> <layer id="6204" name="__module.model.gpt_neox.layers.26.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="6205" name="__module.model.gpt_neox.layers.26.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4561,4562,4566,4567" /> </output> </layer> <layer id="6206" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6207" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6208" name="Constant_433291" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6209" name="ScatterUpdate_433294" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6210" name="Constant_433297" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6211" name="Constant_433300" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6212" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4568,x2.107"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6213" name="Constant_467054" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="6214" name="__module.model.gpt_neox.layers.26.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4569"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6215" name="Constant_433346" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6216" name="Constant_433351" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6217" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4563,x1.107"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6218" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4571"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6219" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4572"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6220" name="__module.model.gpt_neox.layers.26.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4573,key.263"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6221" name="Constant_433397" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6222" name="Constant_433400" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6223" name="Constant_433403" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6224" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4521,key_pass.53"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="6225" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4577,key.265"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6226" name="__module.model.gpt_neox.layers.26.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.26.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6227" name="Constant_535693" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6228" name="Constant_535686" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6229" name="Constant_535688" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6230" name="Constant_535690" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6231" name="Concat_535691" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6232" name="Broadcast_535694" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6233" name="ReadValue_522381" type="ReadValue" version="opset6"> <data variable_id="past_key_values.26.valuepresent.26.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.26.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6234" name="Constant_509701" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6235" name="Gather_509702" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6236" name="Constant_433409" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6237" name="Constant_433412" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6238" name="Constant_433415" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6239" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4515"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6240" name="Constant_15313" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4516"> <dim>4</dim> </port> </output> </layer> <layer id="6241" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4517,value.105"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6242" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.26.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6243" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="4594"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6244" name="Constant_433419" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6245" name="Constant_433418" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6246" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6247" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6248" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4589" /> </output> </layer> <layer id="6249" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6250" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6251" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4587" /> </output> </layer> <layer id="6252" name="__module.model.gpt_neox.layers.26.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4591,4592" /> </output> </layer> <layer id="6253" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6254" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6255" name="Constant_433417" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6256" name="ScatterUpdate_433420" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6257" name="Constant_433421" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6258" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6259" name="ScatterUpdate_433422" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6260" name="Constant_433425" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6261" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4595"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6262" name="Constant_433497" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6263" name="Constant_433498" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6264" name="Constant_433494" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6265" name="Constant_433493" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6266" name="ScatterUpdate_433499" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6267" name="Constant_433502" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6268" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4596,4597,causal_mask.157"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6269" name="__module.model.gpt_neox.layers.26.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6270" name="__module.model.gpt_neox.layers.26.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="4584,mask_value.53" /> </output> </layer> <layer id="6271" name="__module.model.gpt_neox.layers.26.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="4598,causal_mask.159"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6272" name="Add_455215" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6273" name="ShapeOf_455216" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6274" name="Constant_402665" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6275" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6276" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4582"> <dim>1</dim> </port> </output> </layer> <layer id="6277" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6278" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6279" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6280" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="4599"> <dim>4</dim> </port> </output> </layer> <layer id="6281" name="__module.model.gpt_neox.layers.26.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6282" name="__module.model.gpt_neox.layers.26.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="6283" name="__module.model.gpt_neox.layers.26.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6284" name="__module.model.gpt_neox.layers.26.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6285" name="Maximum_455217" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6286" name="__module.model.gpt_neox.layers.26.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4601,attention_mask.61"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6287" name="__module.model.gpt_neox.layers.26.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4602,4603,sdpa_result.53,tensor.105"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6288" name="Constant_15738" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4604"> <dim>4</dim> </port> </output> </layer> <layer id="6289" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4605"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6290" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6291" name="__module.model.gpt_neox.layers.26.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="4610,input.423"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6292" name="Constant_15756" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="12522981912" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6293" name="Convert_549845" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6294" name="Constant_15756/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="12560730648" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6295" name="Convert_549848" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6296" name="Subtract_549849" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6297" name="Constant_549850" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="12560736792" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6298" name="Constant_15756/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6299" name="__module.model.gpt_neox.layers.26.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6300" name="Constant_467055" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12409496088" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6301" name="__module.model.gpt_neox.layers.26.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4613,4615,input.425"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6302" name="__module.model.gpt_neox.layers.26/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4641"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6303" name="__module.model.gpt_neox.layers.26/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4642,4644,input.435"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6304" name="__module.model.gpt_neox.layers.27.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6305" name="__module.model.gpt_neox.layers.27.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6306" name="Constant_467056" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12560761368" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6307" name="__module.model.gpt_neox.layers.27.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6308" name="Constant_467057" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12560785944" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6309" name="__module.model.gpt_neox.layers.27.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4779,input.443"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6310" name="Constant_16356" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="12560810520" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6311" name="Convert_549890" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6312" name="Constant_16356/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="12711805464" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6313" name="Convert_549893" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6314" name="Subtract_549894" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6315" name="Constant_549895" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="12711830040" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6316" name="Constant_16356/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6317" name="__module.model.gpt_neox.layers.27.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6318" name="Constant_467058" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="12711928344" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6319" name="__module.model.gpt_neox.layers.27.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="4784,input.445"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6320" name="__module.model.gpt_neox.layers.27.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="4793,input.447"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6321" name="Constant_16376" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="12712026648" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6322" name="Convert_549899" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6323" name="Constant_16376/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="12863021592" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6324" name="Convert_549902" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6325" name="Subtract_549903" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6326" name="Constant_549904" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="12863027736" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6327" name="Constant_16376/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6328" name="__module.model.gpt_neox.layers.27.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6329" name="Constant_467059" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12863052312" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6330" name="__module.model.gpt_neox.layers.27.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4796,input.449"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6331" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6332" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6333" name="Constant_467060" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12863076888" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6334" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6335" name="Constant_467061" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12863101464" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6336" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4654,input.437"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6337" name="Constant_15835" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="12863126040" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6338" name="Convert_549872" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6339" name="Constant_15835/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="12976372248" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6340" name="Convert_549875" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6341" name="Subtract_549876" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6342" name="Constant_549877" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="12976390680" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6343" name="Constant_15835/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6344" name="__module.model.gpt_neox.layers.27.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6345" name="Constant_467062" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="12976464408" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6346" name="__module.model.gpt_neox.layers.27.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="4661,qkv.109"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6347" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6348" name="__module.model.gpt_neox.layers.27.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4665,qkv.111"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="6349" name="Constant_433510" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6350" name="Constant_433513" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6351" name="Constant_433516" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6352" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4666"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6353" name="Constant_15863" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4667"> <dim>4</dim> </port> </output> </layer> <layer id="6354" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4668,query.217"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6355" name="Constant_433522" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6356" name="Constant_433525" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6357" name="Constant_433528" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6358" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4675,q.55"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6359" name="Constant_15939" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="4687,cos_cached.55"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="6360" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6361" name="Constant_433534" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6362" name="Constant_433537" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6363" name="Constant_433540" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6364" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4669"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6365" name="Constant_15877" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4670"> <dim>4</dim> </port> </output> </layer> <layer id="6366" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4671,key.271"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6367" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6368" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6369" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4679" /> </output> </layer> <layer id="6370" name="Constant_535706" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6371" name="Constant_535699" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6372" name="Constant_535701" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6373" name="Constant_535703" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6374" name="Concat_535704" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6375" name="Broadcast_535707" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6376" name="ReadValue_522383" type="ReadValue" version="opset6"> <data variable_id="past_key_values.27.keypresent.27.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.27.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6377" name="Constant_509704" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6378" name="Gather_509705" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6379" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6380" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6381" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4681" /> </output> </layer> <layer id="6382" name="__module.model.gpt_neox.layers.27.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4680,4684,4685,seq_len.109" /> </output> </layer> <layer id="6383" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6384" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6385" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6386" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4688,4691,cos.109"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6387" name="Constant_52065" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6388" name="__module.model.gpt_neox.layers.27.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4694"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6389" name="__module.model.gpt_neox.layers.27.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4695,cos.111"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6390" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4699"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6391" name="Constant_433545" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6392" name="Constant_433544" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6393" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6394" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6395" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4700,4705" /> </output> </layer> <layer id="6396" name="__module.model.gpt_neox.layers.27.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="6397" name="__module.model.gpt_neox.layers.27.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4702,4703,4707,4708" /> </output> </layer> <layer id="6398" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6399" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6400" name="Constant_433543" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6401" name="ScatterUpdate_433546" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6402" name="Constant_433549" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6403" name="Constant_433552" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6404" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4709,x2.109"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6405" name="Constant_467063" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="6406" name="__module.model.gpt_neox.layers.27.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4710"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6407" name="Constant_433598" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6408" name="Constant_433603" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6409" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4704,x1.109"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6410" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4712"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6411" name="Constant_15937" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="4686,sin_cached.55"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="6412" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6413" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6414" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4689,4692,sin.109"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6415" name="Constant_52068" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6416" name="__module.model.gpt_neox.layers.27.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4697"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6417" name="__module.model.gpt_neox.layers.27.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4698,sin.111"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6418" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4713"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6419" name="__module.model.gpt_neox.layers.27.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4714,query.219"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6420" name="Constant_433650" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6421" name="Constant_433653" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6422" name="Constant_433656" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6423" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4676,query_pass.55"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="6424" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4732,4742,query.221,query.223"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6425" name="Constant_433662" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6426" name="Constant_433665" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6427" name="Constant_433668" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6428" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4677,k.55"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6429" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4715"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6430" name="Constant_433672" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6431" name="Constant_433671" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6432" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6433" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6434" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4716,4721" /> </output> </layer> <layer id="6435" name="__module.model.gpt_neox.layers.27.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="6436" name="__module.model.gpt_neox.layers.27.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4718,4719,4723,4724" /> </output> </layer> <layer id="6437" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6438" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6439" name="Constant_433670" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6440" name="ScatterUpdate_433673" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6441" name="Constant_433676" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6442" name="Constant_433679" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6443" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4725,x2.111"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6444" name="Constant_467064" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="6445" name="__module.model.gpt_neox.layers.27.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4726"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6446" name="Constant_433725" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6447" name="Constant_433730" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6448" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4720,x1.111"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6449" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4728"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6450" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4729"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6451" name="__module.model.gpt_neox.layers.27.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4730,key.273"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6452" name="Constant_433776" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6453" name="Constant_433779" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6454" name="Constant_433782" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6455" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4678,key_pass.55"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="6456" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4734,key.275"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6457" name="__module.model.gpt_neox.layers.27.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.27.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6458" name="Constant_535719" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6459" name="Constant_535712" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6460" name="Constant_535714" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6461" name="Constant_535716" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6462" name="Concat_535717" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6463" name="Broadcast_535720" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6464" name="ReadValue_522385" type="ReadValue" version="opset6"> <data variable_id="past_key_values.27.valuepresent.27.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.27.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6465" name="Constant_509707" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6466" name="Gather_509708" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6467" name="Constant_433788" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6468" name="Constant_433791" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6469" name="Constant_433794" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6470" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4672"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6471" name="Constant_15891" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4673"> <dim>4</dim> </port> </output> </layer> <layer id="6472" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4674,value.109"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6473" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.27.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6474" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="4751"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6475" name="Constant_433798" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6476" name="Constant_433797" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6477" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6478" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6479" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4746" /> </output> </layer> <layer id="6480" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6481" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6482" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4744" /> </output> </layer> <layer id="6483" name="__module.model.gpt_neox.layers.27.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4748,4749" /> </output> </layer> <layer id="6484" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6485" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6486" name="Constant_433796" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6487" name="ScatterUpdate_433799" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6488" name="Constant_433800" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6489" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6490" name="ScatterUpdate_433801" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6491" name="Constant_433804" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6492" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4752"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6493" name="Constant_433876" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6494" name="Constant_433877" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6495" name="Constant_433873" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6496" name="Constant_433872" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6497" name="ScatterUpdate_433878" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6498" name="Constant_433881" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6499" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4753,4754,causal_mask.163"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6500" name="__module.model.gpt_neox.layers.27.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6501" name="__module.model.gpt_neox.layers.27.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="4741,mask_value.55" /> </output> </layer> <layer id="6502" name="__module.model.gpt_neox.layers.27.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="4755,causal_mask.165"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6503" name="Add_455291" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6504" name="ShapeOf_455292" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6505" name="Constant_402680" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6506" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6507" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4739"> <dim>1</dim> </port> </output> </layer> <layer id="6508" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6509" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6510" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6511" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="4756"> <dim>4</dim> </port> </output> </layer> <layer id="6512" name="__module.model.gpt_neox.layers.27.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6513" name="__module.model.gpt_neox.layers.27.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="6514" name="__module.model.gpt_neox.layers.27.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6515" name="__module.model.gpt_neox.layers.27.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6516" name="Maximum_455293" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6517" name="__module.model.gpt_neox.layers.27.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4758,attention_mask.63"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6518" name="__module.model.gpt_neox.layers.27.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4759,4760,sdpa_result.55,tensor.109"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6519" name="Constant_16316" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4761"> <dim>4</dim> </port> </output> </layer> <layer id="6520" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4762"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6521" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6522" name="__module.model.gpt_neox.layers.27.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="4767,input.439"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6523" name="Constant_16334" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="12976538136" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6524" name="Convert_549881" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6525" name="Constant_16334/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="13014286872" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6526" name="Convert_549884" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6527" name="Subtract_549885" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6528" name="Constant_549886" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="13014293016" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6529" name="Constant_16334/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6530" name="__module.model.gpt_neox.layers.27.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6531" name="Constant_467065" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="12863052312" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6532" name="__module.model.gpt_neox.layers.27.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4770,4772,input.441"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6533" name="__module.model.gpt_neox.layers.27/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4798"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6534" name="__module.model.gpt_neox.layers.27/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4799,4801,input.451"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6535" name="__module.model.gpt_neox.layers.28.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6536" name="__module.model.gpt_neox.layers.28.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6537" name="Constant_467066" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13014317592" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6538" name="__module.model.gpt_neox.layers.28.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6539" name="Constant_467067" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13014342168" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6540" name="__module.model.gpt_neox.layers.28.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4936,input.459"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6541" name="Constant_16934" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="13014366744" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6542" name="Convert_549926" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6543" name="Constant_16934/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="13165361688" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6544" name="Convert_549929" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6545" name="Subtract_549930" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6546" name="Constant_549931" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="13165386264" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6547" name="Constant_16934/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6548" name="__module.model.gpt_neox.layers.28.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6549" name="Constant_467068" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="13165484568" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6550" name="__module.model.gpt_neox.layers.28.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="4941,input.461"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6551" name="__module.model.gpt_neox.layers.28.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="4950,input.463"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6552" name="Constant_16954" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="13165582872" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6553" name="Convert_549935" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6554" name="Constant_16954/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="13316577816" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6555" name="Convert_549938" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6556" name="Subtract_549939" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6557" name="Constant_549940" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="13316583960" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6558" name="Constant_16954/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6559" name="__module.model.gpt_neox.layers.28.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6560" name="Constant_467069" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13316608536" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6561" name="__module.model.gpt_neox.layers.28.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4953,input.465"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6562" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6563" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6564" name="Constant_467070" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13316633112" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6565" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6566" name="Constant_467071" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13316657688" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6567" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4811,input.453"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6568" name="Constant_16413" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="13316682264" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6569" name="Convert_549908" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6570" name="Constant_16413/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="13429928472" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6571" name="Convert_549911" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6572" name="Subtract_549912" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6573" name="Constant_549913" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="13429946904" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6574" name="Constant_16413/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6575" name="__module.model.gpt_neox.layers.28.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6576" name="Constant_467072" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="13430020632" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6577" name="__module.model.gpt_neox.layers.28.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="4818,qkv.113"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6578" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6579" name="__module.model.gpt_neox.layers.28.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4822,qkv.115"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="6580" name="Constant_433889" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6581" name="Constant_433892" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6582" name="Constant_433895" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6583" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4823"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6584" name="Constant_16441" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4824"> <dim>4</dim> </port> </output> </layer> <layer id="6585" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4825,query.225"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6586" name="Constant_433901" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6587" name="Constant_433904" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6588" name="Constant_433907" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6589" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4832,q.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6590" name="Constant_16517" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="4844,cos_cached.57"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="6591" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6592" name="Constant_433913" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6593" name="Constant_433916" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6594" name="Constant_433919" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6595" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4826"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6596" name="Constant_16455" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4827"> <dim>4</dim> </port> </output> </layer> <layer id="6597" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4828,key.281"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6598" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6599" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6600" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4836" /> </output> </layer> <layer id="6601" name="Constant_535732" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6602" name="Constant_535725" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6603" name="Constant_535727" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6604" name="Constant_535729" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6605" name="Concat_535730" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6606" name="Broadcast_535733" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6607" name="ReadValue_522387" type="ReadValue" version="opset6"> <data variable_id="past_key_values.28.keypresent.28.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.28.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6608" name="Constant_509710" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6609" name="Gather_509711" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6610" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6611" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6612" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4838" /> </output> </layer> <layer id="6613" name="__module.model.gpt_neox.layers.28.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4837,4841,4842,seq_len.113" /> </output> </layer> <layer id="6614" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6615" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6616" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6617" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4845,4848,cos.113"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6618" name="Constant_52071" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6619" name="__module.model.gpt_neox.layers.28.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4851"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6620" name="__module.model.gpt_neox.layers.28.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4852,cos.115"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6621" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4856"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6622" name="Constant_433924" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6623" name="Constant_433923" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6624" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6625" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6626" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4857,4862" /> </output> </layer> <layer id="6627" name="__module.model.gpt_neox.layers.28.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="6628" name="__module.model.gpt_neox.layers.28.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4859,4860,4864,4865" /> </output> </layer> <layer id="6629" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6630" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6631" name="Constant_433922" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6632" name="ScatterUpdate_433925" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6633" name="Constant_433928" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6634" name="Constant_433931" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6635" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4866,x2.113"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6636" name="Constant_467073" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="6637" name="__module.model.gpt_neox.layers.28.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4867"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6638" name="Constant_433977" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6639" name="Constant_433982" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6640" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4861,x1.113"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6641" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4869"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6642" name="Constant_16515" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="4843,sin_cached.57"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="6643" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6644" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6645" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4846,4849,sin.113"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6646" name="Constant_52074" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6647" name="__module.model.gpt_neox.layers.28.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="4854"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6648" name="__module.model.gpt_neox.layers.28.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="4855,sin.115"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6649" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4870"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6650" name="__module.model.gpt_neox.layers.28.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4871,query.227"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6651" name="Constant_434029" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6652" name="Constant_434032" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6653" name="Constant_434035" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6654" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4833,query_pass.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="6655" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4889,4899,query.229,query.231"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6656" name="Constant_434041" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6657" name="Constant_434044" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6658" name="Constant_434047" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6659" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4834,k.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6660" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4872"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6661" name="Constant_434051" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6662" name="Constant_434050" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6663" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6664" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6665" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4873,4878" /> </output> </layer> <layer id="6666" name="__module.model.gpt_neox.layers.28.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="6667" name="__module.model.gpt_neox.layers.28.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="4875,4876,4880,4881" /> </output> </layer> <layer id="6668" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6669" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6670" name="Constant_434049" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6671" name="ScatterUpdate_434052" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6672" name="Constant_434055" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6673" name="Constant_434058" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6674" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4882,x2.115"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6675" name="Constant_467074" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="6676" name="__module.model.gpt_neox.layers.28.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="4883"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6677" name="Constant_434104" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6678" name="Constant_434109" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6679" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4877,x1.115"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6680" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="4885"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6681" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4886"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6682" name="__module.model.gpt_neox.layers.28.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="4887,key.283"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6683" name="Constant_434155" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6684" name="Constant_434158" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6685" name="Constant_434161" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6686" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4835,key_pass.57"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="6687" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="4891,key.285"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6688" name="__module.model.gpt_neox.layers.28.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.28.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6689" name="Constant_535745" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6690" name="Constant_535738" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6691" name="Constant_535740" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6692" name="Constant_535742" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6693" name="Concat_535743" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6694" name="Broadcast_535746" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6695" name="ReadValue_522389" type="ReadValue" version="opset6"> <data variable_id="past_key_values.28.valuepresent.28.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.28.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6696" name="Constant_509713" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6697" name="Gather_509714" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6698" name="Constant_434167" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6699" name="Constant_434170" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6700" name="Constant_434173" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6701" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4829"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6702" name="Constant_16469" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4830"> <dim>4</dim> </port> </output> </layer> <layer id="6703" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4831,value.113"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6704" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.28.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6705" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="4908"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6706" name="Constant_434177" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6707" name="Constant_434176" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6708" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6709" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6710" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4903" /> </output> </layer> <layer id="6711" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6712" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6713" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4901" /> </output> </layer> <layer id="6714" name="__module.model.gpt_neox.layers.28.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4905,4906" /> </output> </layer> <layer id="6715" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6716" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6717" name="Constant_434175" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6718" name="ScatterUpdate_434178" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6719" name="Constant_434179" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6720" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6721" name="ScatterUpdate_434180" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6722" name="Constant_434183" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6723" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4909"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6724" name="Constant_434255" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6725" name="Constant_434256" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6726" name="Constant_434252" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6727" name="Constant_434251" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6728" name="ScatterUpdate_434257" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6729" name="Constant_434260" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6730" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="4910,4911,causal_mask.169"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6731" name="__module.model.gpt_neox.layers.28.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6732" name="__module.model.gpt_neox.layers.28.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="4898,mask_value.57" /> </output> </layer> <layer id="6733" name="__module.model.gpt_neox.layers.28.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="4912,causal_mask.171"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6734" name="Add_455367" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6735" name="ShapeOf_455368" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6736" name="Constant_402695" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6737" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6738" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4896"> <dim>1</dim> </port> </output> </layer> <layer id="6739" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6740" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6741" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6742" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="4913"> <dim>4</dim> </port> </output> </layer> <layer id="6743" name="__module.model.gpt_neox.layers.28.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6744" name="__module.model.gpt_neox.layers.28.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="6745" name="__module.model.gpt_neox.layers.28.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6746" name="__module.model.gpt_neox.layers.28.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6747" name="Maximum_455369" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6748" name="__module.model.gpt_neox.layers.28.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4915,attention_mask.65"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6749" name="__module.model.gpt_neox.layers.28.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="4916,4917,sdpa_result.57,tensor.113"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6750" name="Constant_16894" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4918"> <dim>4</dim> </port> </output> </layer> <layer id="6751" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4919"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6752" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6753" name="__module.model.gpt_neox.layers.28.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="4924,input.455"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6754" name="Constant_16912" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="13430094360" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6755" name="Convert_549917" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6756" name="Constant_16912/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="13467843096" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6757" name="Convert_549920" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6758" name="Subtract_549921" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6759" name="Constant_549922" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="13467849240" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6760" name="Constant_16912/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6761" name="__module.model.gpt_neox.layers.28.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6762" name="Constant_467075" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13316608536" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6763" name="__module.model.gpt_neox.layers.28.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4927,4929,input.457"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6764" name="__module.model.gpt_neox.layers.28/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4955"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6765" name="__module.model.gpt_neox.layers.28/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4956,4958,input.467"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6766" name="__module.model.gpt_neox.layers.29.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6767" name="__module.model.gpt_neox.layers.29.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6768" name="Constant_467076" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13467873816" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6769" name="__module.model.gpt_neox.layers.29.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6770" name="Constant_467077" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13467898392" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6771" name="__module.model.gpt_neox.layers.29.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5093,input.475"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6772" name="Constant_17512" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="13467922968" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6773" name="Convert_549962" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6774" name="Constant_17512/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="13618917912" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6775" name="Convert_549965" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6776" name="Subtract_549966" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6777" name="Constant_549967" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="13618942488" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="6778" name="Constant_17512/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6779" name="__module.model.gpt_neox.layers.29.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6780" name="Constant_467078" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="13619040792" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6781" name="__module.model.gpt_neox.layers.29.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="5098,input.477"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6782" name="__module.model.gpt_neox.layers.29.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="5107,input.479"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6783" name="Constant_17532" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="13619139096" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6784" name="Convert_549971" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6785" name="Constant_17532/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="13770134040" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6786" name="Convert_549974" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6787" name="Subtract_549975" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6788" name="Constant_549976" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="13770140184" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6789" name="Constant_17532/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="6790" name="__module.model.gpt_neox.layers.29.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6791" name="Constant_467079" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13770164760" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6792" name="__module.model.gpt_neox.layers.29.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5110,input.481"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6793" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6794" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6795" name="Constant_467080" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13770189336" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6796" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6797" name="Constant_467081" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13770213912" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6798" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="4968,input.469"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6799" name="Constant_16991" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="13770238488" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6800" name="Convert_549944" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6801" name="Constant_16991/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="13883484696" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6802" name="Convert_549947" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6803" name="Subtract_549948" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6804" name="Constant_549949" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="13883503128" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="6805" name="Constant_16991/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6806" name="__module.model.gpt_neox.layers.29.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6807" name="Constant_467082" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="13883576856" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6808" name="__module.model.gpt_neox.layers.29.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="4975,qkv.117"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="6809" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6810" name="__module.model.gpt_neox.layers.29.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4979,qkv.119"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="6811" name="Constant_434268" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6812" name="Constant_434271" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6813" name="Constant_434274" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6814" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4980"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6815" name="Constant_17019" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4981"> <dim>4</dim> </port> </output> </layer> <layer id="6816" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4982,query.233"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6817" name="Constant_434280" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6818" name="Constant_434283" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6819" name="Constant_434286" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6820" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4989,q.59"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6821" name="Constant_17095" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="5001,cos_cached.59"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="6822" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6823" name="Constant_434292" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6824" name="Constant_434295" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6825" name="Constant_434298" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6826" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4983"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6827" name="Constant_17033" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4984"> <dim>4</dim> </port> </output> </layer> <layer id="6828" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4985,key.291"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6829" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6830" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6831" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4993" /> </output> </layer> <layer id="6832" name="Constant_535758" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6833" name="Constant_535751" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6834" name="Constant_535753" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6835" name="Constant_535755" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6836" name="Concat_535756" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6837" name="Broadcast_535759" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6838" name="ReadValue_522391" type="ReadValue" version="opset6"> <data variable_id="past_key_values.29.keypresent.29.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.29.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6839" name="Constant_509716" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6840" name="Gather_509717" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6841" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6842" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6843" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="4995" /> </output> </layer> <layer id="6844" name="__module.model.gpt_neox.layers.29.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="4994,4998,4999,seq_len.117" /> </output> </layer> <layer id="6845" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6846" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6847" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6848" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5002,5005,cos.117"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6849" name="Constant_52077" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6850" name="__module.model.gpt_neox.layers.29.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5008"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6851" name="__module.model.gpt_neox.layers.29.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5009,cos.119"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6852" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5013"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6853" name="Constant_434303" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6854" name="Constant_434302" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6855" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6856" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6857" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5014,5019" /> </output> </layer> <layer id="6858" name="__module.model.gpt_neox.layers.29.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="6859" name="__module.model.gpt_neox.layers.29.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5016,5017,5021,5022" /> </output> </layer> <layer id="6860" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6861" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6862" name="Constant_434301" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6863" name="ScatterUpdate_434304" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6864" name="Constant_434307" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6865" name="Constant_434310" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6866" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5023,x2.117"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6867" name="Constant_467083" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="6868" name="__module.model.gpt_neox.layers.29.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5024"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6869" name="Constant_434356" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6870" name="Constant_434361" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6871" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5018,x1.117"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6872" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5026"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6873" name="Constant_17093" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="5000,sin_cached.59"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="6874" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6875" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6876" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5003,5006,sin.117"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6877" name="Constant_52080" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6878" name="__module.model.gpt_neox.layers.29.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5011"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6879" name="__module.model.gpt_neox.layers.29.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5012,sin.119"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6880" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5027"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6881" name="__module.model.gpt_neox.layers.29.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5028,query.235"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6882" name="Constant_434408" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6883" name="Constant_434411" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6884" name="Constant_434414" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6885" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4990,query_pass.59"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="6886" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5046,5056,query.237,query.239"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6887" name="Constant_434420" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6888" name="Constant_434423" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6889" name="Constant_434426" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6890" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4991,k.59"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6891" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5029"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6892" name="Constant_434430" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6893" name="Constant_434429" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6894" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6895" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6896" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5030,5035" /> </output> </layer> <layer id="6897" name="__module.model.gpt_neox.layers.29.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="6898" name="__module.model.gpt_neox.layers.29.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5032,5033,5037,5038" /> </output> </layer> <layer id="6899" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6900" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6901" name="Constant_434428" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6902" name="ScatterUpdate_434431" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6903" name="Constant_434434" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6904" name="Constant_434437" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6905" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5039,x2.119"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6906" name="Constant_467084" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="6907" name="__module.model.gpt_neox.layers.29.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5040"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6908" name="Constant_434483" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6909" name="Constant_434488" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6910" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5034,x1.119"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="6911" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5042"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6912" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5043"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6913" name="__module.model.gpt_neox.layers.29.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5044,key.293"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="6914" name="Constant_434534" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6915" name="Constant_434537" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6916" name="Constant_434540" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6917" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4992,key_pass.59"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="6918" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5048,key.295"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6919" name="__module.model.gpt_neox.layers.29.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.29.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6920" name="Constant_535771" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6921" name="Constant_535764" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6922" name="Constant_535766" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6923" name="Constant_535768" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6924" name="Concat_535769" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6925" name="Broadcast_535772" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="6926" name="ReadValue_522393" type="ReadValue" version="opset6"> <data variable_id="past_key_values.29.valuepresent.29.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.29.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6927" name="Constant_509719" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="6928" name="Gather_509720" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6929" name="Constant_434546" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6930" name="Constant_434549" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="6931" name="Constant_434552" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6932" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="4986"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6933" name="Constant_17047" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="4987"> <dim>4</dim> </port> </output> </layer> <layer id="6934" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="4988,value.117"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6935" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.29.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6936" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="5065"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6937" name="Constant_434556" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6938" name="Constant_434555" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6939" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6940" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6941" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5060" /> </output> </layer> <layer id="6942" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6943" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6944" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5058" /> </output> </layer> <layer id="6945" name="__module.model.gpt_neox.layers.29.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5062,5063" /> </output> </layer> <layer id="6946" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6947" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6948" name="Constant_434554" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6949" name="ScatterUpdate_434557" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6950" name="Constant_434558" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6951" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6952" name="ScatterUpdate_434559" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6953" name="Constant_434562" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6954" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5066"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="6955" name="Constant_434634" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6956" name="Constant_434635" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6957" name="Constant_434631" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="6958" name="Constant_434630" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6959" name="ScatterUpdate_434636" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6960" name="Constant_434639" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6961" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5067,5068,causal_mask.175"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6962" name="__module.model.gpt_neox.layers.29.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="6963" name="__module.model.gpt_neox.layers.29.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="5055,mask_value.59" /> </output> </layer> <layer id="6964" name="__module.model.gpt_neox.layers.29.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="5069,causal_mask.177"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6965" name="Add_455443" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6966" name="ShapeOf_455444" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6967" name="Constant_402710" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6968" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="6969" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5053"> <dim>1</dim> </port> </output> </layer> <layer id="6970" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6971" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6972" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6973" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="5070"> <dim>4</dim> </port> </output> </layer> <layer id="6974" name="__module.model.gpt_neox.layers.29.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6975" name="__module.model.gpt_neox.layers.29.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="6976" name="__module.model.gpt_neox.layers.29.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6977" name="__module.model.gpt_neox.layers.29.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6978" name="Maximum_455445" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="6979" name="__module.model.gpt_neox.layers.29.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5072,attention_mask.67"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="6980" name="__module.model.gpt_neox.layers.29.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5073,5074,sdpa_result.59,tensor.117"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="6981" name="Constant_17472" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5075"> <dim>4</dim> </port> </output> </layer> <layer id="6982" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5076"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="6983" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="6984" name="__module.model.gpt_neox.layers.29.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="5081,input.471"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6985" name="Constant_17490" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="13883650584" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6986" name="Convert_549953" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6987" name="Constant_17490/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="13921399320" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6988" name="Convert_549956" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6989" name="Subtract_549957" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6990" name="Constant_549958" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="13921405464" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="6991" name="Constant_17490/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6992" name="__module.model.gpt_neox.layers.29.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6993" name="Constant_467085" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13770164760" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6994" name="__module.model.gpt_neox.layers.29.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5084,5086,input.473"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6995" name="__module.model.gpt_neox.layers.29/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5112"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6996" name="__module.model.gpt_neox.layers.29/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5113,5115,input.483"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6997" name="__module.model.gpt_neox.layers.30.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="6998" name="__module.model.gpt_neox.layers.30.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="6999" name="Constant_467086" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13921430040" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7000" name="__module.model.gpt_neox.layers.30.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7001" name="Constant_467087" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="13921454616" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7002" name="__module.model.gpt_neox.layers.30.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5250,input.491"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7003" name="Constant_18090" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="13921479192" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7004" name="Convert_549998" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7005" name="Constant_18090/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="14072474136" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7006" name="Convert_550001" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7007" name="Subtract_550002" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7008" name="Constant_550003" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="14072498712" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7009" name="Constant_18090/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7010" name="__module.model.gpt_neox.layers.30.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7011" name="Constant_467088" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="14072597016" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7012" name="__module.model.gpt_neox.layers.30.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="5255,input.493"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7013" name="__module.model.gpt_neox.layers.30.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="5264,input.495"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7014" name="Constant_18110" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="14072695320" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7015" name="Convert_550007" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7016" name="Constant_18110/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="14223690264" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7017" name="Convert_550010" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7018" name="Subtract_550011" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7019" name="Constant_550012" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="14223696408" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7020" name="Constant_18110/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7021" name="__module.model.gpt_neox.layers.30.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7022" name="Constant_467089" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14223720984" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7023" name="__module.model.gpt_neox.layers.30.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5267,input.497"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7024" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7025" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7026" name="Constant_467090" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14223745560" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7027" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7028" name="Constant_467091" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14223770136" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7029" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5125,input.485"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7030" name="Constant_17569" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="14223794712" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7031" name="Convert_549980" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7032" name="Constant_17569/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="14337040920" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7033" name="Convert_549983" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7034" name="Subtract_549984" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7035" name="Constant_549985" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="14337059352" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7036" name="Constant_17569/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7037" name="__module.model.gpt_neox.layers.30.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7038" name="Constant_467092" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="14337133080" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7039" name="__module.model.gpt_neox.layers.30.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="5132,qkv.121"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7040" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7041" name="__module.model.gpt_neox.layers.30.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5136,qkv.123"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="7042" name="Constant_434647" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7043" name="Constant_434650" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7044" name="Constant_434653" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7045" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5137"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7046" name="Constant_17597" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5138"> <dim>4</dim> </port> </output> </layer> <layer id="7047" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5139,query.241"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7048" name="Constant_434659" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7049" name="Constant_434662" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7050" name="Constant_434665" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7051" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5146,q.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7052" name="Constant_17673" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="5158,cos_cached.61"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7053" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7054" name="Constant_434671" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7055" name="Constant_434674" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7056" name="Constant_434677" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7057" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5140"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7058" name="Constant_17611" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5141"> <dim>4</dim> </port> </output> </layer> <layer id="7059" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5142,key.301"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7060" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7061" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7062" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5150" /> </output> </layer> <layer id="7063" name="Constant_535784" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7064" name="Constant_535777" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7065" name="Constant_535779" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7066" name="Constant_535781" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7067" name="Concat_535782" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7068" name="Broadcast_535785" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7069" name="ReadValue_522399" type="ReadValue" version="opset6"> <data variable_id="past_key_values.30.keypresent.30.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.30.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7070" name="Constant_509722" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7071" name="Gather_509723" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7072" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7073" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7074" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5152" /> </output> </layer> <layer id="7075" name="__module.model.gpt_neox.layers.30.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5151,5155,5156,seq_len.121" /> </output> </layer> <layer id="7076" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7077" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7078" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7079" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5159,5162,cos.121"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7080" name="Constant_52083" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7081" name="__module.model.gpt_neox.layers.30.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5165"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7082" name="__module.model.gpt_neox.layers.30.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5166,cos.123"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7083" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5170"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7084" name="Constant_434682" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7085" name="Constant_434681" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7086" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7087" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7088" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5171,5176" /> </output> </layer> <layer id="7089" name="__module.model.gpt_neox.layers.30.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="7090" name="__module.model.gpt_neox.layers.30.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5173,5174,5178,5179" /> </output> </layer> <layer id="7091" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7092" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7093" name="Constant_434680" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7094" name="ScatterUpdate_434683" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7095" name="Constant_434686" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7096" name="Constant_434689" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7097" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5180,x2.121"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7098" name="Constant_467093" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="7099" name="__module.model.gpt_neox.layers.30.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5181"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7100" name="Constant_434735" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7101" name="Constant_434740" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7102" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5175,x1.121"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7103" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5183"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7104" name="Constant_17671" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="5157,sin_cached.61"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7105" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7106" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7107" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5160,5163,sin.121"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7108" name="Constant_52086" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7109" name="__module.model.gpt_neox.layers.30.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5168"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7110" name="__module.model.gpt_neox.layers.30.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5169,sin.123"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7111" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5184"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7112" name="__module.model.gpt_neox.layers.30.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5185,query.243"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7113" name="Constant_434787" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7114" name="Constant_434790" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7115" name="Constant_434793" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7116" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5147,query_pass.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="7117" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5203,5213,query.245,query.247"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7118" name="Constant_434799" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7119" name="Constant_434802" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7120" name="Constant_434805" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7121" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5148,k.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7122" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5186"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7123" name="Constant_434809" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7124" name="Constant_434808" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7125" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7126" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7127" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5187,5192" /> </output> </layer> <layer id="7128" name="__module.model.gpt_neox.layers.30.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="7129" name="__module.model.gpt_neox.layers.30.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5189,5190,5194,5195" /> </output> </layer> <layer id="7130" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7131" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7132" name="Constant_434807" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7133" name="ScatterUpdate_434810" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7134" name="Constant_434813" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7135" name="Constant_434816" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7136" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5196,x2.123"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7137" name="Constant_467094" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="7138" name="__module.model.gpt_neox.layers.30.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5197"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7139" name="Constant_434862" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7140" name="Constant_434867" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7141" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5191,x1.123"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7142" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5199"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7143" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5200"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7144" name="__module.model.gpt_neox.layers.30.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5201,key.303"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7145" name="Constant_434913" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7146" name="Constant_434916" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7147" name="Constant_434919" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7148" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5149,key_pass.61"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="7149" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5205,key.305"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7150" name="__module.model.gpt_neox.layers.30.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.30.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7151" name="Constant_535797" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7152" name="Constant_535790" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7153" name="Constant_535792" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7154" name="Constant_535794" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7155" name="Concat_535795" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7156" name="Broadcast_535798" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7157" name="ReadValue_522401" type="ReadValue" version="opset6"> <data variable_id="past_key_values.30.valuepresent.30.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.30.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7158" name="Constant_509725" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7159" name="Gather_509726" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7160" name="Constant_434925" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7161" name="Constant_434928" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7162" name="Constant_434931" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7163" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5143"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7164" name="Constant_17625" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5144"> <dim>4</dim> </port> </output> </layer> <layer id="7165" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5145,value.121"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7166" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.30.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7167" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="5222"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="7168" name="Constant_434935" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7169" name="Constant_434934" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7170" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7171" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7172" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5217" /> </output> </layer> <layer id="7173" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7174" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7175" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5215" /> </output> </layer> <layer id="7176" name="__module.model.gpt_neox.layers.30.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5219,5220" /> </output> </layer> <layer id="7177" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7178" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7179" name="Constant_434933" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7180" name="ScatterUpdate_434936" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7181" name="Constant_434937" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7182" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7183" name="ScatterUpdate_434938" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7184" name="Constant_434941" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7185" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5223"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="7186" name="Constant_435013" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7187" name="Constant_435014" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7188" name="Constant_435010" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7189" name="Constant_435009" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7190" name="ScatterUpdate_435015" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7191" name="Constant_435018" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7192" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5224,5225,causal_mask.181"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7193" name="__module.model.gpt_neox.layers.30.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7194" name="__module.model.gpt_neox.layers.30.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="5212,mask_value.61" /> </output> </layer> <layer id="7195" name="__module.model.gpt_neox.layers.30.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="5226,causal_mask.183"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7196" name="Add_455519" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7197" name="ShapeOf_455520" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7198" name="Constant_402725" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7199" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7200" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5210"> <dim>1</dim> </port> </output> </layer> <layer id="7201" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7202" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7203" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7204" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="5227"> <dim>4</dim> </port> </output> </layer> <layer id="7205" name="__module.model.gpt_neox.layers.30.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7206" name="__module.model.gpt_neox.layers.30.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="7207" name="__module.model.gpt_neox.layers.30.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7208" name="__module.model.gpt_neox.layers.30.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7209" name="Maximum_455521" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7210" name="__module.model.gpt_neox.layers.30.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5229,attention_mask.69"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7211" name="__module.model.gpt_neox.layers.30.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5230,5231,sdpa_result.61,tensor.121"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7212" name="Constant_18050" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5232"> <dim>4</dim> </port> </output> </layer> <layer id="7213" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5233"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7214" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7215" name="__module.model.gpt_neox.layers.30.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="5238,input.487"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7216" name="Constant_18068" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="14337206808" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7217" name="Convert_549989" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7218" name="Constant_18068/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="14374955544" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7219" name="Convert_549992" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7220" name="Subtract_549993" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7221" name="Constant_549994" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="14374961688" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7222" name="Constant_18068/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7223" name="__module.model.gpt_neox.layers.30.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7224" name="Constant_467095" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14223720984" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7225" name="__module.model.gpt_neox.layers.30.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5241,5243,input.489"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7226" name="__module.model.gpt_neox.layers.30/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5269"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7227" name="__module.model.gpt_neox.layers.30/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5270,5272,input.499"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7228" name="__module.model.gpt_neox.layers.31.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7229" name="__module.model.gpt_neox.layers.31.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7230" name="Constant_467096" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14374986264" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7231" name="__module.model.gpt_neox.layers.31.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7232" name="Constant_467097" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14375010840" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7233" name="__module.model.gpt_neox.layers.31.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5407,input.507"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7234" name="Constant_18668" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="14375035416" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7235" name="Convert_550034" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7236" name="Constant_18668/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="14526030360" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7237" name="Convert_550037" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7238" name="Subtract_550038" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7239" name="Constant_550039" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="14526054936" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7240" name="Constant_18668/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7241" name="__module.model.gpt_neox.layers.31.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7242" name="Constant_467098" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="14526153240" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7243" name="__module.model.gpt_neox.layers.31.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="5412,input.509"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7244" name="__module.model.gpt_neox.layers.31.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="5421,input.511"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7245" name="Constant_18688" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="14526251544" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7246" name="Convert_550043" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7247" name="Constant_18688/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="14677246488" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7248" name="Convert_550046" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7249" name="Subtract_550047" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7250" name="Constant_550048" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="14677252632" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7251" name="Constant_18688/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7252" name="__module.model.gpt_neox.layers.31.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7253" name="Constant_467099" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14677277208" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7254" name="__module.model.gpt_neox.layers.31.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5424,input.513"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7255" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7256" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7257" name="Constant_467100" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14677301784" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7258" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7259" name="Constant_467101" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14677326360" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7260" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5282,input.501"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7261" name="Constant_18147" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="14677350936" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7262" name="Convert_550016" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7263" name="Constant_18147/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="14790597144" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7264" name="Convert_550019" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7265" name="Subtract_550020" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7266" name="Constant_550021" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="14790615576" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7267" name="Constant_18147/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7268" name="__module.model.gpt_neox.layers.31.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7269" name="Constant_467102" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="14790689304" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7270" name="__module.model.gpt_neox.layers.31.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="5289,qkv.125"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7271" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7272" name="__module.model.gpt_neox.layers.31.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5293,qkv.127"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="7273" name="Constant_435026" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7274" name="Constant_435029" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7275" name="Constant_435032" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7276" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5294"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7277" name="Constant_18175" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5295"> <dim>4</dim> </port> </output> </layer> <layer id="7278" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5296,query.249"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7279" name="Constant_435038" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7280" name="Constant_435041" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7281" name="Constant_435044" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7282" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5303,q.63"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7283" name="Constant_18251" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="5315,cos_cached.63"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7284" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7285" name="Constant_435050" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7286" name="Constant_435053" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7287" name="Constant_435056" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7288" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5297"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7289" name="Constant_18189" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5298"> <dim>4</dim> </port> </output> </layer> <layer id="7290" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5299,key.311"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7291" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7292" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7293" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5307" /> </output> </layer> <layer id="7294" name="Constant_535810" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7295" name="Constant_535803" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7296" name="Constant_535805" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7297" name="Constant_535807" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7298" name="Concat_535808" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7299" name="Broadcast_535811" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7300" name="ReadValue_522403" type="ReadValue" version="opset6"> <data variable_id="past_key_values.31.keypresent.31.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.31.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7301" name="Constant_509728" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7302" name="Gather_509729" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7303" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7304" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7305" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5309" /> </output> </layer> <layer id="7306" name="__module.model.gpt_neox.layers.31.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5308,5312,5313,seq_len.125" /> </output> </layer> <layer id="7307" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7308" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7309" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7310" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5316,5319,cos.125"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7311" name="Constant_52089" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7312" name="__module.model.gpt_neox.layers.31.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5322"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7313" name="__module.model.gpt_neox.layers.31.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5323,cos.127"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7314" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5327"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7315" name="Constant_435061" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7316" name="Constant_435060" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7317" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7318" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7319" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5328,5333" /> </output> </layer> <layer id="7320" name="__module.model.gpt_neox.layers.31.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="7321" name="__module.model.gpt_neox.layers.31.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5330,5331,5335,5336" /> </output> </layer> <layer id="7322" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7323" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7324" name="Constant_435059" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7325" name="ScatterUpdate_435062" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7326" name="Constant_435065" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7327" name="Constant_435068" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7328" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5337,x2.125"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7329" name="Constant_467103" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="7330" name="__module.model.gpt_neox.layers.31.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5338"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7331" name="Constant_435114" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7332" name="Constant_435119" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7333" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5332,x1.125"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7334" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5340"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7335" name="Constant_18249" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="5314,sin_cached.63"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7336" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7337" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7338" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5317,5320,sin.125"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7339" name="Constant_52092" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7340" name="__module.model.gpt_neox.layers.31.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5325"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7341" name="__module.model.gpt_neox.layers.31.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5326,sin.127"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7342" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5341"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7343" name="__module.model.gpt_neox.layers.31.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5342,query.251"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7344" name="Constant_435166" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7345" name="Constant_435169" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7346" name="Constant_435172" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7347" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5304,query_pass.63"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="7348" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5360,5370,query.253,query.255"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7349" name="Constant_435178" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7350" name="Constant_435181" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7351" name="Constant_435184" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7352" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5305,k.63"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7353" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5343"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7354" name="Constant_435188" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7355" name="Constant_435187" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7356" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7357" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7358" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5344,5349" /> </output> </layer> <layer id="7359" name="__module.model.gpt_neox.layers.31.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="7360" name="__module.model.gpt_neox.layers.31.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5346,5347,5351,5352" /> </output> </layer> <layer id="7361" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7362" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7363" name="Constant_435186" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7364" name="ScatterUpdate_435189" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7365" name="Constant_435192" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7366" name="Constant_435195" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7367" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5353,x2.127"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7368" name="Constant_467104" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="7369" name="__module.model.gpt_neox.layers.31.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5354"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7370" name="Constant_435241" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7371" name="Constant_435246" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7372" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5348,x1.127"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7373" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5356"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7374" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5357"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7375" name="__module.model.gpt_neox.layers.31.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5358,key.313"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7376" name="Constant_435292" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7377" name="Constant_435295" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7378" name="Constant_435298" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7379" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5306,key_pass.63"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="7380" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5362,key.315"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7381" name="__module.model.gpt_neox.layers.31.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.31.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7382" name="Constant_535823" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7383" name="Constant_535816" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7384" name="Constant_535818" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7385" name="Constant_535820" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7386" name="Concat_535821" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7387" name="Broadcast_535824" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7388" name="ReadValue_522405" type="ReadValue" version="opset6"> <data variable_id="past_key_values.31.valuepresent.31.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.31.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7389" name="Constant_509731" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7390" name="Gather_509732" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7391" name="Constant_435304" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7392" name="Constant_435307" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7393" name="Constant_435310" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7394" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5300"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7395" name="Constant_18203" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5301"> <dim>4</dim> </port> </output> </layer> <layer id="7396" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5302,value.125"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7397" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.31.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7398" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="5379"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="7399" name="Constant_435314" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7400" name="Constant_435313" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7401" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7402" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7403" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5374" /> </output> </layer> <layer id="7404" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7405" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7406" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5372" /> </output> </layer> <layer id="7407" name="__module.model.gpt_neox.layers.31.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5376,5377" /> </output> </layer> <layer id="7408" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7409" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7410" name="Constant_435312" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7411" name="ScatterUpdate_435315" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7412" name="Constant_435316" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7413" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7414" name="ScatterUpdate_435317" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7415" name="Constant_435320" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7416" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5380"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="7417" name="Constant_435392" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7418" name="Constant_435393" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7419" name="Constant_435389" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7420" name="Constant_435388" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7421" name="ScatterUpdate_435394" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7422" name="Constant_435397" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7423" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5381,5382,causal_mask.187"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7424" name="__module.model.gpt_neox.layers.31.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7425" name="__module.model.gpt_neox.layers.31.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="5369,mask_value.63" /> </output> </layer> <layer id="7426" name="__module.model.gpt_neox.layers.31.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="5383,causal_mask.189"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7427" name="Add_455595" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7428" name="ShapeOf_455596" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7429" name="Constant_402740" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7430" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7431" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5367"> <dim>1</dim> </port> </output> </layer> <layer id="7432" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7433" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7434" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7435" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="5384"> <dim>4</dim> </port> </output> </layer> <layer id="7436" name="__module.model.gpt_neox.layers.31.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7437" name="__module.model.gpt_neox.layers.31.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="7438" name="__module.model.gpt_neox.layers.31.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7439" name="__module.model.gpt_neox.layers.31.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7440" name="Maximum_455597" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7441" name="__module.model.gpt_neox.layers.31.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5386,attention_mask.71"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7442" name="__module.model.gpt_neox.layers.31.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5387,5388,sdpa_result.63,tensor.125"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7443" name="Constant_18628" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5389"> <dim>4</dim> </port> </output> </layer> <layer id="7444" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5390"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7445" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7446" name="__module.model.gpt_neox.layers.31.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="5395,input.503"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7447" name="Constant_18646" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="14790763032" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7448" name="Convert_550025" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7449" name="Constant_18646/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="14828511768" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7450" name="Convert_550028" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7451" name="Subtract_550029" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7452" name="Constant_550030" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="14828517912" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7453" name="Constant_18646/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7454" name="__module.model.gpt_neox.layers.31.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7455" name="Constant_467105" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14677277208" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7456" name="__module.model.gpt_neox.layers.31.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5398,5400,input.505"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7457" name="__module.model.gpt_neox.layers.31/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5426"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7458" name="__module.model.gpt_neox.layers.31/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5427,5429,input.515"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7459" name="__module.model.gpt_neox.layers.32.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7460" name="__module.model.gpt_neox.layers.32.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7461" name="Constant_467106" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14828542488" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7462" name="__module.model.gpt_neox.layers.32.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7463" name="Constant_467107" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="14828567064" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7464" name="__module.model.gpt_neox.layers.32.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5564,input.523"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7465" name="Constant_19246" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="14828591640" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7466" name="Convert_550070" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7467" name="Constant_19246/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="14979586584" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7468" name="Convert_550073" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7469" name="Subtract_550074" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7470" name="Constant_550075" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="14979611160" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7471" name="Constant_19246/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7472" name="__module.model.gpt_neox.layers.32.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7473" name="Constant_467108" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="14979709464" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7474" name="__module.model.gpt_neox.layers.32.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="5569,input.525"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7475" name="__module.model.gpt_neox.layers.32.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="5578,input.527"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7476" name="Constant_19266" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="14979807768" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7477" name="Convert_550079" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7478" name="Constant_19266/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="15130802712" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7479" name="Convert_550082" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7480" name="Subtract_550083" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7481" name="Constant_550084" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="15130808856" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7482" name="Constant_19266/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7483" name="__module.model.gpt_neox.layers.32.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7484" name="Constant_467109" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15130833432" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7485" name="__module.model.gpt_neox.layers.32.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5581,input.529"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7486" name="__module.model.gpt_neox.layers.32.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7487" name="__module.model.gpt_neox.layers.32.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7488" name="Constant_467110" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15130858008" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7489" name="__module.model.gpt_neox.layers.32.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7490" name="Constant_467111" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15130882584" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7491" name="__module.model.gpt_neox.layers.32.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5439,input.517"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7492" name="Constant_18725" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="15130907160" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7493" name="Convert_550052" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7494" name="Constant_18725/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="15244153368" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7495" name="Convert_550055" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7496" name="Subtract_550056" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7497" name="Constant_550057" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="15244171800" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7498" name="Constant_18725/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7499" name="__module.model.gpt_neox.layers.32.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7500" name="Constant_467112" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="15244245528" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7501" name="__module.model.gpt_neox.layers.32.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="5446,qkv.129"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7502" name="__module.model.gpt_neox.layers.32.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7503" name="__module.model.gpt_neox.layers.32.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5450,qkv.131"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="7504" name="Constant_435405" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7505" name="Constant_435408" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7506" name="Constant_435411" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7507" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5451"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7508" name="Constant_18753" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5452"> <dim>4</dim> </port> </output> </layer> <layer id="7509" name="__module.model.gpt_neox.layers.32.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5453,query.257"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7510" name="Constant_435417" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7511" name="Constant_435420" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7512" name="Constant_435423" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7513" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5460,q.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7514" name="Constant_18829" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="5472,cos_cached.65"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7515" name="__module.model.gpt_neox.layers.32.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7516" name="Constant_435429" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7517" name="Constant_435432" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7518" name="Constant_435435" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7519" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5454"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7520" name="Constant_18767" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5455"> <dim>4</dim> </port> </output> </layer> <layer id="7521" name="__module.model.gpt_neox.layers.32.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5456,key.321"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7522" name="__module.model.gpt_neox.layers.32.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7523" name="__module.model.gpt_neox.layers.32.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7524" name="__module.model.gpt_neox.layers.32.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5464" /> </output> </layer> <layer id="7525" name="Constant_535836" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7526" name="Constant_535829" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7527" name="Constant_535831" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7528" name="Constant_535833" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7529" name="Concat_535834" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7530" name="Broadcast_535837" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7531" name="ReadValue_522407" type="ReadValue" version="opset6"> <data variable_id="past_key_values.32.keypresent.32.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.32.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7532" name="Constant_509734" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7533" name="Gather_509735" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7534" name="__module.model.gpt_neox.layers.32.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7535" name="__module.model.gpt_neox.layers.32.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7536" name="__module.model.gpt_neox.layers.32.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5466" /> </output> </layer> <layer id="7537" name="__module.model.gpt_neox.layers.32.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5465,5469,5470,seq_len.129" /> </output> </layer> <layer id="7538" name="__module.model.gpt_neox.layers.32.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7539" name="__module.model.gpt_neox.layers.32.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7540" name="__module.model.gpt_neox.layers.32.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7541" name="__module.model.gpt_neox.layers.32.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5473,5476,cos.129"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7542" name="Constant_52095" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7543" name="__module.model.gpt_neox.layers.32.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5479"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7544" name="__module.model.gpt_neox.layers.32.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5480,cos.131"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7545" name="__module.model.gpt_neox.layers.32.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5484"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7546" name="Constant_435440" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7547" name="Constant_435439" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7548" name="__module.model.gpt_neox.layers.32.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7549" name="__module.model.gpt_neox.layers.32.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7550" name="__module.model.gpt_neox.layers.32.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5485,5490" /> </output> </layer> <layer id="7551" name="__module.model.gpt_neox.layers.32.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="7552" name="__module.model.gpt_neox.layers.32.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5487,5488,5492,5493" /> </output> </layer> <layer id="7553" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7554" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7555" name="Constant_435438" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7556" name="ScatterUpdate_435441" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7557" name="Constant_435444" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7558" name="Constant_435447" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7559" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5494,x2.129"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7560" name="Constant_467113" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="7561" name="__module.model.gpt_neox.layers.32.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5495"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7562" name="Constant_435493" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7563" name="Constant_435498" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7564" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5489,x1.129"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7565" name="__module.model.gpt_neox.layers.32.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5497"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7566" name="Constant_18827" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="5471,sin_cached.65"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7567" name="__module.model.gpt_neox.layers.32.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7568" name="__module.model.gpt_neox.layers.32.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7569" name="__module.model.gpt_neox.layers.32.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5474,5477,sin.129"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7570" name="Constant_52098" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7571" name="__module.model.gpt_neox.layers.32.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5482"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7572" name="__module.model.gpt_neox.layers.32.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5483,sin.131"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7573" name="__module.model.gpt_neox.layers.32.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5498"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7574" name="__module.model.gpt_neox.layers.32.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5499,query.259"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7575" name="Constant_435545" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7576" name="Constant_435548" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7577" name="Constant_435551" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7578" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5461,query_pass.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="7579" name="__module.model.gpt_neox.layers.32.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5517,5527,query.261,query.263"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7580" name="Constant_435557" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7581" name="Constant_435560" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7582" name="Constant_435563" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7583" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5462,k.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7584" name="__module.model.gpt_neox.layers.32.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5500"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7585" name="Constant_435567" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7586" name="Constant_435566" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7587" name="__module.model.gpt_neox.layers.32.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7588" name="__module.model.gpt_neox.layers.32.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7589" name="__module.model.gpt_neox.layers.32.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5501,5506" /> </output> </layer> <layer id="7590" name="__module.model.gpt_neox.layers.32.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="7591" name="__module.model.gpt_neox.layers.32.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5503,5504,5508,5509" /> </output> </layer> <layer id="7592" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7593" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7594" name="Constant_435565" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7595" name="ScatterUpdate_435568" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7596" name="Constant_435571" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7597" name="Constant_435574" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7598" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5510,x2.131"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7599" name="Constant_467114" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="7600" name="__module.model.gpt_neox.layers.32.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5511"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7601" name="Constant_435620" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7602" name="Constant_435625" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7603" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5505,x1.131"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7604" name="__module.model.gpt_neox.layers.32.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5513"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7605" name="__module.model.gpt_neox.layers.32.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5514"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7606" name="__module.model.gpt_neox.layers.32.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5515,key.323"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7607" name="Constant_435671" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7608" name="Constant_435674" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7609" name="Constant_435677" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7610" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5463,key_pass.65"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="7611" name="__module.model.gpt_neox.layers.32.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5519,key.325"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7612" name="__module.model.gpt_neox.layers.32.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.32.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7613" name="Constant_535849" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7614" name="Constant_535842" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7615" name="Constant_535844" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7616" name="Constant_535846" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7617" name="Concat_535847" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7618" name="Broadcast_535850" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7619" name="ReadValue_522409" type="ReadValue" version="opset6"> <data variable_id="past_key_values.32.valuepresent.32.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.32.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7620" name="Constant_509737" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7621" name="Gather_509738" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7622" name="Constant_435683" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7623" name="Constant_435686" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7624" name="Constant_435689" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7625" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5457"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7626" name="Constant_18781" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5458"> <dim>4</dim> </port> </output> </layer> <layer id="7627" name="__module.model.gpt_neox.layers.32.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5459,value.129"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7628" name="__module.model.gpt_neox.layers.32.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.32.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7629" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="5536"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="7630" name="Constant_435693" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7631" name="Constant_435692" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7632" name="__module.model.gpt_neox.layers.32.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7633" name="__module.model.gpt_neox.layers.32.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7634" name="__module.model.gpt_neox.layers.32.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5531" /> </output> </layer> <layer id="7635" name="__module.model.gpt_neox.layers.32.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7636" name="__module.model.gpt_neox.layers.32.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7637" name="__module.model.gpt_neox.layers.32.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5529" /> </output> </layer> <layer id="7638" name="__module.model.gpt_neox.layers.32.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5533,5534" /> </output> </layer> <layer id="7639" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7640" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7641" name="Constant_435691" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7642" name="ScatterUpdate_435694" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7643" name="Constant_435695" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7644" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7645" name="ScatterUpdate_435696" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7646" name="Constant_435699" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7647" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5537"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="7648" name="Constant_435771" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7649" name="Constant_435772" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7650" name="Constant_435768" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7651" name="Constant_435767" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7652" name="ScatterUpdate_435773" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7653" name="Constant_435776" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7654" name="__module.model.gpt_neox.layers.32.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5538,5539,causal_mask.193"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7655" name="__module.model.gpt_neox.layers.32.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7656" name="__module.model.gpt_neox.layers.32.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="5526,mask_value.65" /> </output> </layer> <layer id="7657" name="__module.model.gpt_neox.layers.32.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="5540,causal_mask.195"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7658" name="Add_455671" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7659" name="ShapeOf_455672" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7660" name="Constant_402755" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7661" name="__module.model.gpt_neox.layers.32.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7662" name="__module.model.gpt_neox.layers.32.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5524"> <dim>1</dim> </port> </output> </layer> <layer id="7663" name="__module.model.gpt_neox.layers.32.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7664" name="__module.model.gpt_neox.layers.32.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7665" name="__module.model.gpt_neox.layers.32.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7666" name="__module.model.gpt_neox.layers.32.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="5541"> <dim>4</dim> </port> </output> </layer> <layer id="7667" name="__module.model.gpt_neox.layers.32.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7668" name="__module.model.gpt_neox.layers.32.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="7669" name="__module.model.gpt_neox.layers.32.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7670" name="__module.model.gpt_neox.layers.32.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7671" name="Maximum_455673" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7672" name="__module.model.gpt_neox.layers.32.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5543,attention_mask.73"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7673" name="__module.model.gpt_neox.layers.32.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5544,5545,sdpa_result.65,tensor.129"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7674" name="Constant_19206" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5546"> <dim>4</dim> </port> </output> </layer> <layer id="7675" name="__module.model.gpt_neox.layers.32.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5547"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7676" name="__module.model.gpt_neox.layers.32.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7677" name="__module.model.gpt_neox.layers.32.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="5552,input.519"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7678" name="Constant_19224" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="15244319256" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7679" name="Convert_550061" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7680" name="Constant_19224/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="15282067992" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7681" name="Convert_550064" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7682" name="Subtract_550065" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7683" name="Constant_550066" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="15282074136" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7684" name="Constant_19224/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7685" name="__module.model.gpt_neox.layers.32.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7686" name="Constant_467115" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15130833432" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7687" name="__module.model.gpt_neox.layers.32.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5555,5557,input.521"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7688" name="__module.model.gpt_neox.layers.32/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5583"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7689" name="__module.model.gpt_neox.layers.32/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5584,5586,input.531"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7690" name="__module.model.gpt_neox.layers.33.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7691" name="__module.model.gpt_neox.layers.33.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7692" name="Constant_467116" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15282098712" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7693" name="__module.model.gpt_neox.layers.33.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7694" name="Constant_467117" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15282123288" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7695" name="__module.model.gpt_neox.layers.33.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5721,input.539"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7696" name="Constant_19824" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="15282147864" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7697" name="Convert_550106" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7698" name="Constant_19824/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="15433142808" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7699" name="Convert_550109" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7700" name="Subtract_550110" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7701" name="Constant_550111" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="15433167384" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7702" name="Constant_19824/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7703" name="__module.model.gpt_neox.layers.33.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7704" name="Constant_467118" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="15433265688" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7705" name="__module.model.gpt_neox.layers.33.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="5726,input.541"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7706" name="__module.model.gpt_neox.layers.33.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="5735,input.543"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7707" name="Constant_19844" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="15433363992" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7708" name="Convert_550115" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7709" name="Constant_19844/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="15584358936" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7710" name="Convert_550118" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7711" name="Subtract_550119" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7712" name="Constant_550120" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="15584365080" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7713" name="Constant_19844/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7714" name="__module.model.gpt_neox.layers.33.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7715" name="Constant_467119" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15584389656" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7716" name="__module.model.gpt_neox.layers.33.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5738,input.545"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7717" name="__module.model.gpt_neox.layers.33.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7718" name="__module.model.gpt_neox.layers.33.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7719" name="Constant_467120" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15584414232" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7720" name="__module.model.gpt_neox.layers.33.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7721" name="Constant_467121" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15584438808" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7722" name="__module.model.gpt_neox.layers.33.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5596,input.533"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7723" name="Constant_19303" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="15584463384" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7724" name="Convert_550088" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7725" name="Constant_19303/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="15697709592" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7726" name="Convert_550091" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7727" name="Subtract_550092" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7728" name="Constant_550093" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="15697728024" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7729" name="Constant_19303/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7730" name="__module.model.gpt_neox.layers.33.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7731" name="Constant_467122" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="15697801752" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7732" name="__module.model.gpt_neox.layers.33.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="5603,qkv.133"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7733" name="__module.model.gpt_neox.layers.33.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7734" name="__module.model.gpt_neox.layers.33.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5607,qkv.135"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="7735" name="Constant_435784" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7736" name="Constant_435787" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7737" name="Constant_435790" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7738" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5608"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7739" name="Constant_19331" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5609"> <dim>4</dim> </port> </output> </layer> <layer id="7740" name="__module.model.gpt_neox.layers.33.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5610,query.265"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7741" name="Constant_435796" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7742" name="Constant_435799" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7743" name="Constant_435802" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7744" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5617,q.67"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7745" name="Constant_19407" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="5629,cos_cached.67"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7746" name="__module.model.gpt_neox.layers.33.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7747" name="Constant_435808" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7748" name="Constant_435811" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7749" name="Constant_435814" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7750" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5611"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7751" name="Constant_19345" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5612"> <dim>4</dim> </port> </output> </layer> <layer id="7752" name="__module.model.gpt_neox.layers.33.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5613,key.331"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7753" name="__module.model.gpt_neox.layers.33.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7754" name="__module.model.gpt_neox.layers.33.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7755" name="__module.model.gpt_neox.layers.33.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5621" /> </output> </layer> <layer id="7756" name="Constant_535862" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7757" name="Constant_535855" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7758" name="Constant_535857" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7759" name="Constant_535859" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7760" name="Concat_535860" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7761" name="Broadcast_535863" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7762" name="ReadValue_522411" type="ReadValue" version="opset6"> <data variable_id="past_key_values.33.keypresent.33.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.33.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7763" name="Constant_509740" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7764" name="Gather_509741" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7765" name="__module.model.gpt_neox.layers.33.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7766" name="__module.model.gpt_neox.layers.33.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7767" name="__module.model.gpt_neox.layers.33.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5623" /> </output> </layer> <layer id="7768" name="__module.model.gpt_neox.layers.33.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5622,5626,5627,seq_len.133" /> </output> </layer> <layer id="7769" name="__module.model.gpt_neox.layers.33.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7770" name="__module.model.gpt_neox.layers.33.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7771" name="__module.model.gpt_neox.layers.33.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7772" name="__module.model.gpt_neox.layers.33.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5630,5633,cos.133"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7773" name="Constant_52101" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7774" name="__module.model.gpt_neox.layers.33.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5636"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7775" name="__module.model.gpt_neox.layers.33.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5637,cos.135"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7776" name="__module.model.gpt_neox.layers.33.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5641"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7777" name="Constant_435819" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7778" name="Constant_435818" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7779" name="__module.model.gpt_neox.layers.33.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7780" name="__module.model.gpt_neox.layers.33.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7781" name="__module.model.gpt_neox.layers.33.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5642,5647" /> </output> </layer> <layer id="7782" name="__module.model.gpt_neox.layers.33.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="7783" name="__module.model.gpt_neox.layers.33.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5644,5645,5649,5650" /> </output> </layer> <layer id="7784" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7785" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7786" name="Constant_435817" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7787" name="ScatterUpdate_435820" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7788" name="Constant_435823" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7789" name="Constant_435826" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7790" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5651,x2.133"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7791" name="Constant_467123" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="7792" name="__module.model.gpt_neox.layers.33.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5652"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7793" name="Constant_435872" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7794" name="Constant_435877" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7795" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5646,x1.133"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7796" name="__module.model.gpt_neox.layers.33.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5654"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7797" name="Constant_19405" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="5628,sin_cached.67"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7798" name="__module.model.gpt_neox.layers.33.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7799" name="__module.model.gpt_neox.layers.33.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7800" name="__module.model.gpt_neox.layers.33.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5631,5634,sin.133"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7801" name="Constant_52104" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7802" name="__module.model.gpt_neox.layers.33.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5639"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7803" name="__module.model.gpt_neox.layers.33.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5640,sin.135"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7804" name="__module.model.gpt_neox.layers.33.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5655"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7805" name="__module.model.gpt_neox.layers.33.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5656,query.267"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7806" name="Constant_435924" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7807" name="Constant_435927" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7808" name="Constant_435930" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7809" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5618,query_pass.67"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="7810" name="__module.model.gpt_neox.layers.33.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5674,5684,query.269,query.271"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7811" name="Constant_435936" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7812" name="Constant_435939" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7813" name="Constant_435942" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7814" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5619,k.67"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7815" name="__module.model.gpt_neox.layers.33.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5657"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7816" name="Constant_435946" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7817" name="Constant_435945" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7818" name="__module.model.gpt_neox.layers.33.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7819" name="__module.model.gpt_neox.layers.33.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7820" name="__module.model.gpt_neox.layers.33.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5658,5663" /> </output> </layer> <layer id="7821" name="__module.model.gpt_neox.layers.33.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="7822" name="__module.model.gpt_neox.layers.33.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5660,5661,5665,5666" /> </output> </layer> <layer id="7823" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7824" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7825" name="Constant_435944" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7826" name="ScatterUpdate_435947" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7827" name="Constant_435950" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7828" name="Constant_435953" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7829" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5667,x2.135"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7830" name="Constant_467124" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="7831" name="__module.model.gpt_neox.layers.33.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5668"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7832" name="Constant_435999" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7833" name="Constant_436004" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7834" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5662,x1.135"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="7835" name="__module.model.gpt_neox.layers.33.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5670"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7836" name="__module.model.gpt_neox.layers.33.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5671"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7837" name="__module.model.gpt_neox.layers.33.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5672,key.333"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7838" name="Constant_436050" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7839" name="Constant_436053" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7840" name="Constant_436056" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7841" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5620,key_pass.67"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="7842" name="__module.model.gpt_neox.layers.33.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5676,key.335"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7843" name="__module.model.gpt_neox.layers.33.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.33.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7844" name="Constant_535875" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7845" name="Constant_535868" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7846" name="Constant_535870" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7847" name="Constant_535872" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7848" name="Concat_535873" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7849" name="Broadcast_535876" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7850" name="ReadValue_522413" type="ReadValue" version="opset6"> <data variable_id="past_key_values.33.valuepresent.33.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.33.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7851" name="Constant_509743" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7852" name="Gather_509744" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7853" name="Constant_436062" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7854" name="Constant_436065" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7855" name="Constant_436068" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7856" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5614"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7857" name="Constant_19359" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5615"> <dim>4</dim> </port> </output> </layer> <layer id="7858" name="__module.model.gpt_neox.layers.33.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5616,value.133"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7859" name="__module.model.gpt_neox.layers.33.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.33.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7860" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="5693"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="7861" name="Constant_436072" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7862" name="Constant_436071" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7863" name="__module.model.gpt_neox.layers.33.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7864" name="__module.model.gpt_neox.layers.33.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7865" name="__module.model.gpt_neox.layers.33.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5688" /> </output> </layer> <layer id="7866" name="__module.model.gpt_neox.layers.33.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7867" name="__module.model.gpt_neox.layers.33.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7868" name="__module.model.gpt_neox.layers.33.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5686" /> </output> </layer> <layer id="7869" name="__module.model.gpt_neox.layers.33.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5690,5691" /> </output> </layer> <layer id="7870" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7871" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7872" name="Constant_436070" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7873" name="ScatterUpdate_436073" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7874" name="Constant_436074" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7875" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7876" name="ScatterUpdate_436075" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7877" name="Constant_436078" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7878" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5694"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="7879" name="Constant_436150" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7880" name="Constant_436151" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7881" name="Constant_436147" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7882" name="Constant_436146" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7883" name="ScatterUpdate_436152" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7884" name="Constant_436155" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7885" name="__module.model.gpt_neox.layers.33.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5695,5696,causal_mask.199"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7886" name="__module.model.gpt_neox.layers.33.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7887" name="__module.model.gpt_neox.layers.33.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="5683,mask_value.67" /> </output> </layer> <layer id="7888" name="__module.model.gpt_neox.layers.33.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="5697,causal_mask.201"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7889" name="Add_455747" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7890" name="ShapeOf_455748" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7891" name="Constant_402770" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7892" name="__module.model.gpt_neox.layers.33.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7893" name="__module.model.gpt_neox.layers.33.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5681"> <dim>1</dim> </port> </output> </layer> <layer id="7894" name="__module.model.gpt_neox.layers.33.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7895" name="__module.model.gpt_neox.layers.33.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7896" name="__module.model.gpt_neox.layers.33.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7897" name="__module.model.gpt_neox.layers.33.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="5698"> <dim>4</dim> </port> </output> </layer> <layer id="7898" name="__module.model.gpt_neox.layers.33.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7899" name="__module.model.gpt_neox.layers.33.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="7900" name="__module.model.gpt_neox.layers.33.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7901" name="__module.model.gpt_neox.layers.33.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7902" name="Maximum_455749" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7903" name="__module.model.gpt_neox.layers.33.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5700,attention_mask.75"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="7904" name="__module.model.gpt_neox.layers.33.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5701,5702,sdpa_result.67,tensor.133"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7905" name="Constant_19784" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5703"> <dim>4</dim> </port> </output> </layer> <layer id="7906" name="__module.model.gpt_neox.layers.33.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5704"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7907" name="__module.model.gpt_neox.layers.33.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="7908" name="__module.model.gpt_neox.layers.33.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="5709,input.535"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7909" name="Constant_19802" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="15697875480" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7910" name="Convert_550097" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7911" name="Constant_19802/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="15735624216" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7912" name="Convert_550100" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7913" name="Subtract_550101" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7914" name="Constant_550102" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="15735630360" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7915" name="Constant_19802/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7916" name="__module.model.gpt_neox.layers.33.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7917" name="Constant_467125" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15584389656" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7918" name="__module.model.gpt_neox.layers.33.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5712,5714,input.537"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7919" name="__module.model.gpt_neox.layers.33/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5740"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7920" name="__module.model.gpt_neox.layers.33/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5741,5743,input.547"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7921" name="__module.model.gpt_neox.layers.34.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7922" name="__module.model.gpt_neox.layers.34.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7923" name="Constant_467126" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15735654936" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7924" name="__module.model.gpt_neox.layers.34.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7925" name="Constant_467127" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="15735679512" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7926" name="__module.model.gpt_neox.layers.34.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5878,input.555"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7927" name="Constant_20402" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="15735704088" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7928" name="Convert_550142" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7929" name="Constant_20402/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="15886699032" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7930" name="Convert_550145" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7931" name="Subtract_550146" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7932" name="Constant_550147" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="15886723608" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="7933" name="Constant_20402/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7934" name="__module.model.gpt_neox.layers.34.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7935" name="Constant_467128" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="15886821912" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7936" name="__module.model.gpt_neox.layers.34.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="5883,input.557"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7937" name="__module.model.gpt_neox.layers.34.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="5892,input.559"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7938" name="Constant_20422" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="15886920216" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7939" name="Convert_550151" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7940" name="Constant_20422/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="16037915160" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7941" name="Convert_550154" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7942" name="Subtract_550155" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7943" name="Constant_550156" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="16037921304" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="7944" name="Constant_20422/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="7945" name="__module.model.gpt_neox.layers.34.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7946" name="Constant_467129" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16037945880" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7947" name="__module.model.gpt_neox.layers.34.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5895,input.561"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7948" name="__module.model.gpt_neox.layers.34.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7949" name="__module.model.gpt_neox.layers.34.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7950" name="Constant_467130" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16037970456" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7951" name="__module.model.gpt_neox.layers.34.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7952" name="Constant_467131" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16037995032" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7953" name="__module.model.gpt_neox.layers.34.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5753,input.549"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7954" name="Constant_19881" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="16038019608" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7955" name="Convert_550124" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7956" name="Constant_19881/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="16151265816" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7957" name="Convert_550127" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7958" name="Subtract_550128" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7959" name="Constant_550129" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="16151284248" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="7960" name="Constant_19881/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="7961" name="__module.model.gpt_neox.layers.34.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7962" name="Constant_467132" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="16151357976" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7963" name="__module.model.gpt_neox.layers.34.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="5760,qkv.137"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="7964" name="__module.model.gpt_neox.layers.34.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7965" name="__module.model.gpt_neox.layers.34.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5764,qkv.139"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="7966" name="Constant_436163" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7967" name="Constant_436166" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7968" name="Constant_436169" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7969" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5765"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7970" name="Constant_19909" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5766"> <dim>4</dim> </port> </output> </layer> <layer id="7971" name="__module.model.gpt_neox.layers.34.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5767,query.273"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7972" name="Constant_436175" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7973" name="Constant_436178" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7974" name="Constant_436181" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7975" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5774,q.69"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="7976" name="Constant_19985" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="5786,cos_cached.69"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="7977" name="__module.model.gpt_neox.layers.34.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="7978" name="Constant_436187" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7979" name="Constant_436190" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7980" name="Constant_436193" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7981" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5768"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="7982" name="Constant_19923" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5769"> <dim>4</dim> </port> </output> </layer> <layer id="7983" name="__module.model.gpt_neox.layers.34.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5770,key.341"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7984" name="__module.model.gpt_neox.layers.34.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7985" name="__module.model.gpt_neox.layers.34.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7986" name="__module.model.gpt_neox.layers.34.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5778" /> </output> </layer> <layer id="7987" name="Constant_535888" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="7988" name="Constant_535881" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7989" name="Constant_535883" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7990" name="Constant_535885" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="7991" name="Concat_535886" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="7992" name="Broadcast_535889" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="7993" name="ReadValue_522415" type="ReadValue" version="opset6"> <data variable_id="past_key_values.34.keypresent.34.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.34.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7994" name="Constant_509746" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7995" name="Gather_509747" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="7996" name="__module.model.gpt_neox.layers.34.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="7997" name="__module.model.gpt_neox.layers.34.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="7998" name="__module.model.gpt_neox.layers.34.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5780" /> </output> </layer> <layer id="7999" name="__module.model.gpt_neox.layers.34.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5779,5783,5784,seq_len.137" /> </output> </layer> <layer id="8000" name="__module.model.gpt_neox.layers.34.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8001" name="__module.model.gpt_neox.layers.34.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8002" name="__module.model.gpt_neox.layers.34.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8003" name="__module.model.gpt_neox.layers.34.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5787,5790,cos.137"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8004" name="Constant_52107" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8005" name="__module.model.gpt_neox.layers.34.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5793"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8006" name="__module.model.gpt_neox.layers.34.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5794,cos.139"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8007" name="__module.model.gpt_neox.layers.34.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5798"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8008" name="Constant_436198" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8009" name="Constant_436197" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8010" name="__module.model.gpt_neox.layers.34.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8011" name="__module.model.gpt_neox.layers.34.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8012" name="__module.model.gpt_neox.layers.34.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5799,5804" /> </output> </layer> <layer id="8013" name="__module.model.gpt_neox.layers.34.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8014" name="__module.model.gpt_neox.layers.34.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5801,5802,5806,5807" /> </output> </layer> <layer id="8015" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8016" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8017" name="Constant_436196" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8018" name="ScatterUpdate_436199" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8019" name="Constant_436202" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8020" name="Constant_436205" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8021" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5808,x2.137"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8022" name="Constant_467133" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8023" name="__module.model.gpt_neox.layers.34.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5809"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8024" name="Constant_436251" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8025" name="Constant_436256" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8026" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5803,x1.137"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8027" name="__module.model.gpt_neox.layers.34.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5811"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8028" name="Constant_19983" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="5785,sin_cached.69"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8029" name="__module.model.gpt_neox.layers.34.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8030" name="__module.model.gpt_neox.layers.34.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8031" name="__module.model.gpt_neox.layers.34.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5788,5791,sin.137"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8032" name="Constant_52110" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8033" name="__module.model.gpt_neox.layers.34.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5796"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8034" name="__module.model.gpt_neox.layers.34.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5797,sin.139"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8035" name="__module.model.gpt_neox.layers.34.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5812"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8036" name="__module.model.gpt_neox.layers.34.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5813,query.275"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8037" name="Constant_436303" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8038" name="Constant_436306" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8039" name="Constant_436309" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8040" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5775,query_pass.69"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8041" name="__module.model.gpt_neox.layers.34.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5831,5841,query.277,query.279"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8042" name="Constant_436315" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8043" name="Constant_436318" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8044" name="Constant_436321" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8045" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5776,k.69"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8046" name="__module.model.gpt_neox.layers.34.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5814"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8047" name="Constant_436325" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8048" name="Constant_436324" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8049" name="__module.model.gpt_neox.layers.34.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8050" name="__module.model.gpt_neox.layers.34.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8051" name="__module.model.gpt_neox.layers.34.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5815,5820" /> </output> </layer> <layer id="8052" name="__module.model.gpt_neox.layers.34.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8053" name="__module.model.gpt_neox.layers.34.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5817,5818,5822,5823" /> </output> </layer> <layer id="8054" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8055" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8056" name="Constant_436323" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8057" name="ScatterUpdate_436326" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8058" name="Constant_436329" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8059" name="Constant_436332" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8060" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5824,x2.139"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8061" name="Constant_467134" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8062" name="__module.model.gpt_neox.layers.34.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5825"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8063" name="Constant_436378" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8064" name="Constant_436383" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8065" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5819,x1.139"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8066" name="__module.model.gpt_neox.layers.34.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5827"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8067" name="__module.model.gpt_neox.layers.34.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5828"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8068" name="__module.model.gpt_neox.layers.34.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5829,key.343"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8069" name="Constant_436429" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8070" name="Constant_436432" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8071" name="Constant_436435" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8072" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5777,key_pass.69"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8073" name="__module.model.gpt_neox.layers.34.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5833,key.345"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8074" name="__module.model.gpt_neox.layers.34.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.34.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8075" name="Constant_535901" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8076" name="Constant_535894" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8077" name="Constant_535896" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8078" name="Constant_535898" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8079" name="Concat_535899" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8080" name="Broadcast_535902" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="8081" name="ReadValue_522417" type="ReadValue" version="opset6"> <data variable_id="past_key_values.34.valuepresent.34.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.34.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8082" name="Constant_509749" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8083" name="Gather_509750" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8084" name="Constant_436441" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8085" name="Constant_436444" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8086" name="Constant_436447" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8087" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5771"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8088" name="Constant_19937" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5772"> <dim>4</dim> </port> </output> </layer> <layer id="8089" name="__module.model.gpt_neox.layers.34.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5773,value.137"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8090" name="__module.model.gpt_neox.layers.34.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.34.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8091" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="5850"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="8092" name="Constant_436451" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8093" name="Constant_436450" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8094" name="__module.model.gpt_neox.layers.34.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8095" name="__module.model.gpt_neox.layers.34.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8096" name="__module.model.gpt_neox.layers.34.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5845" /> </output> </layer> <layer id="8097" name="__module.model.gpt_neox.layers.34.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8098" name="__module.model.gpt_neox.layers.34.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8099" name="__module.model.gpt_neox.layers.34.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5843" /> </output> </layer> <layer id="8100" name="__module.model.gpt_neox.layers.34.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5847,5848" /> </output> </layer> <layer id="8101" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8102" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8103" name="Constant_436449" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8104" name="ScatterUpdate_436452" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8105" name="Constant_436453" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8106" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8107" name="ScatterUpdate_436454" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8108" name="Constant_436457" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8109" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5851"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="8110" name="Constant_436529" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8111" name="Constant_436530" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8112" name="Constant_436526" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8113" name="Constant_436525" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8114" name="ScatterUpdate_436531" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8115" name="Constant_436534" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8116" name="__module.model.gpt_neox.layers.34.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="5852,5853,causal_mask.205"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8117" name="__module.model.gpt_neox.layers.34.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8118" name="__module.model.gpt_neox.layers.34.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="5840,mask_value.69" /> </output> </layer> <layer id="8119" name="__module.model.gpt_neox.layers.34.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="5854,causal_mask.207"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8120" name="Add_455823" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8121" name="ShapeOf_455824" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8122" name="Constant_402785" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8123" name="__module.model.gpt_neox.layers.34.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8124" name="__module.model.gpt_neox.layers.34.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5838"> <dim>1</dim> </port> </output> </layer> <layer id="8125" name="__module.model.gpt_neox.layers.34.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8126" name="__module.model.gpt_neox.layers.34.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8127" name="__module.model.gpt_neox.layers.34.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8128" name="__module.model.gpt_neox.layers.34.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="5855"> <dim>4</dim> </port> </output> </layer> <layer id="8129" name="__module.model.gpt_neox.layers.34.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8130" name="__module.model.gpt_neox.layers.34.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="8131" name="__module.model.gpt_neox.layers.34.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8132" name="__module.model.gpt_neox.layers.34.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8133" name="Maximum_455825" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8134" name="__module.model.gpt_neox.layers.34.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5857,attention_mask.77"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8135" name="__module.model.gpt_neox.layers.34.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5858,5859,sdpa_result.69,tensor.137"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8136" name="Constant_20362" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5860"> <dim>4</dim> </port> </output> </layer> <layer id="8137" name="__module.model.gpt_neox.layers.34.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5861"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8138" name="__module.model.gpt_neox.layers.34.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8139" name="__module.model.gpt_neox.layers.34.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="5866,input.551"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8140" name="Constant_20380" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="16151431704" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8141" name="Convert_550133" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8142" name="Constant_20380/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="16189180440" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8143" name="Convert_550136" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8144" name="Subtract_550137" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8145" name="Constant_550138" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="16189186584" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8146" name="Constant_20380/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8147" name="__module.model.gpt_neox.layers.34.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8148" name="Constant_467135" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16037945880" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8149" name="__module.model.gpt_neox.layers.34.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5869,5871,input.553"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8150" name="__module.model.gpt_neox.layers.34/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5897"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8151" name="__module.model.gpt_neox.layers.34/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5898,5900,input.563"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8152" name="__module.model.gpt_neox.layers.35.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8153" name="__module.model.gpt_neox.layers.35.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8154" name="Constant_467136" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16189211160" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8155" name="__module.model.gpt_neox.layers.35.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8156" name="Constant_467137" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16189235736" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8157" name="__module.model.gpt_neox.layers.35.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6035,input.571"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8158" name="Constant_20980" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="16189260312" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8159" name="Convert_550178" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8160" name="Constant_20980/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="16340255256" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8161" name="Convert_550181" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8162" name="Subtract_550182" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8163" name="Constant_550183" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="16340279832" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8164" name="Constant_20980/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8165" name="__module.model.gpt_neox.layers.35.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8166" name="Constant_467138" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="16340378136" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8167" name="__module.model.gpt_neox.layers.35.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="6040,input.573"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8168" name="__module.model.gpt_neox.layers.35.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="6049,input.575"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8169" name="Constant_21000" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="16340476440" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8170" name="Convert_550187" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8171" name="Constant_21000/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="16491471384" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8172" name="Convert_550190" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8173" name="Subtract_550191" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8174" name="Constant_550192" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="16491477528" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8175" name="Constant_21000/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8176" name="__module.model.gpt_neox.layers.35.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8177" name="Constant_467139" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16491502104" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8178" name="__module.model.gpt_neox.layers.35.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6052,input.577"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8179" name="__module.model.gpt_neox.layers.35.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8180" name="__module.model.gpt_neox.layers.35.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8181" name="Constant_467140" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16491526680" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8182" name="__module.model.gpt_neox.layers.35.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8183" name="Constant_467141" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16491551256" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8184" name="__module.model.gpt_neox.layers.35.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="5910,input.565"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8185" name="Constant_20459" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="16491575832" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8186" name="Convert_550160" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8187" name="Constant_20459/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="16604822040" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8188" name="Convert_550163" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8189" name="Subtract_550164" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8190" name="Constant_550165" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="16604840472" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8191" name="Constant_20459/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8192" name="__module.model.gpt_neox.layers.35.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8193" name="Constant_467142" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="16604914200" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8194" name="__module.model.gpt_neox.layers.35.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="5917,qkv.141"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8195" name="__module.model.gpt_neox.layers.35.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8196" name="__module.model.gpt_neox.layers.35.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5921,qkv.143"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="8197" name="Constant_436542" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8198" name="Constant_436545" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8199" name="Constant_436548" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8200" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5922"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8201" name="Constant_20487" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5923"> <dim>4</dim> </port> </output> </layer> <layer id="8202" name="__module.model.gpt_neox.layers.35.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5924,query.281"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8203" name="Constant_436554" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8204" name="Constant_436557" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8205" name="Constant_436560" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8206" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5931,q.71"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8207" name="Constant_20563" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="5943,cos_cached.71"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8208" name="__module.model.gpt_neox.layers.35.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8209" name="Constant_436566" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8210" name="Constant_436569" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8211" name="Constant_436572" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8212" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5925"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8213" name="Constant_20501" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5926"> <dim>4</dim> </port> </output> </layer> <layer id="8214" name="__module.model.gpt_neox.layers.35.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5927,key.351"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8215" name="__module.model.gpt_neox.layers.35.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8216" name="__module.model.gpt_neox.layers.35.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8217" name="__module.model.gpt_neox.layers.35.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5935" /> </output> </layer> <layer id="8218" name="Constant_535914" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8219" name="Constant_535907" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8220" name="Constant_535909" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8221" name="Constant_535911" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8222" name="Concat_535912" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8223" name="Broadcast_535915" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="8224" name="ReadValue_522419" type="ReadValue" version="opset6"> <data variable_id="past_key_values.35.keypresent.35.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.35.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8225" name="Constant_509752" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8226" name="Gather_509753" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8227" name="__module.model.gpt_neox.layers.35.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8228" name="__module.model.gpt_neox.layers.35.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8229" name="__module.model.gpt_neox.layers.35.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5937" /> </output> </layer> <layer id="8230" name="__module.model.gpt_neox.layers.35.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="5936,5940,5941,seq_len.141" /> </output> </layer> <layer id="8231" name="__module.model.gpt_neox.layers.35.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8232" name="__module.model.gpt_neox.layers.35.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8233" name="__module.model.gpt_neox.layers.35.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8234" name="__module.model.gpt_neox.layers.35.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5944,5947,cos.141"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8235" name="Constant_52113" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8236" name="__module.model.gpt_neox.layers.35.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5950"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8237" name="__module.model.gpt_neox.layers.35.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5951,cos.143"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8238" name="__module.model.gpt_neox.layers.35.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5955"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8239" name="Constant_436577" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8240" name="Constant_436576" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8241" name="__module.model.gpt_neox.layers.35.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8242" name="__module.model.gpt_neox.layers.35.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8243" name="__module.model.gpt_neox.layers.35.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5956,5961" /> </output> </layer> <layer id="8244" name="__module.model.gpt_neox.layers.35.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8245" name="__module.model.gpt_neox.layers.35.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5958,5959,5963,5964" /> </output> </layer> <layer id="8246" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8247" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8248" name="Constant_436575" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8249" name="ScatterUpdate_436578" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8250" name="Constant_436581" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8251" name="Constant_436584" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8252" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5965,x2.141"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8253" name="Constant_467143" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8254" name="__module.model.gpt_neox.layers.35.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5966"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8255" name="Constant_436630" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8256" name="Constant_436635" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8257" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5960,x1.141"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8258" name="__module.model.gpt_neox.layers.35.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5968"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8259" name="Constant_20561" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="5942,sin_cached.71"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8260" name="__module.model.gpt_neox.layers.35.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8261" name="__module.model.gpt_neox.layers.35.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8262" name="__module.model.gpt_neox.layers.35.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="5945,5948,sin.141"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8263" name="Constant_52116" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8264" name="__module.model.gpt_neox.layers.35.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="5953"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8265" name="__module.model.gpt_neox.layers.35.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="5954,sin.143"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8266" name="__module.model.gpt_neox.layers.35.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5969"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8267" name="__module.model.gpt_neox.layers.35.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5970,query.283"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8268" name="Constant_436682" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8269" name="Constant_436685" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8270" name="Constant_436688" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8271" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5932,query_pass.71"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8272" name="__module.model.gpt_neox.layers.35.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5988,5998,query.285,query.287"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8273" name="Constant_436694" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8274" name="Constant_436697" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8275" name="Constant_436700" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8276" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5933,k.71"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8277" name="__module.model.gpt_neox.layers.35.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5971"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8278" name="Constant_436704" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8279" name="Constant_436703" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8280" name="__module.model.gpt_neox.layers.35.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8281" name="__module.model.gpt_neox.layers.35.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8282" name="__module.model.gpt_neox.layers.35.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5972,5977" /> </output> </layer> <layer id="8283" name="__module.model.gpt_neox.layers.35.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8284" name="__module.model.gpt_neox.layers.35.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="5974,5975,5979,5980" /> </output> </layer> <layer id="8285" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8286" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8287" name="Constant_436702" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8288" name="ScatterUpdate_436705" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8289" name="Constant_436708" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8290" name="Constant_436711" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8291" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5981,x2.143"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8292" name="Constant_467144" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8293" name="__module.model.gpt_neox.layers.35.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="5982"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8294" name="Constant_436757" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8295" name="Constant_436762" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8296" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5976,x1.143"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8297" name="__module.model.gpt_neox.layers.35.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="5984"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8298" name="__module.model.gpt_neox.layers.35.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5985"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8299" name="__module.model.gpt_neox.layers.35.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="5986,key.353"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8300" name="Constant_436808" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8301" name="Constant_436811" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8302" name="Constant_436814" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8303" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5934,key_pass.71"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8304" name="__module.model.gpt_neox.layers.35.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="5990,key.355"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8305" name="__module.model.gpt_neox.layers.35.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.35.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8306" name="Constant_535927" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8307" name="Constant_535920" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8308" name="Constant_535922" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8309" name="Constant_535924" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8310" name="Concat_535925" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8311" name="Broadcast_535928" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="8312" name="ReadValue_522421" type="ReadValue" version="opset6"> <data variable_id="past_key_values.35.valuepresent.35.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.35.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8313" name="Constant_509755" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8314" name="Gather_509756" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8315" name="Constant_436820" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8316" name="Constant_436823" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8317" name="Constant_436826" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8318" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="5928"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8319" name="Constant_20515" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="5929"> <dim>4</dim> </port> </output> </layer> <layer id="8320" name="__module.model.gpt_neox.layers.35.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="5930,value.141"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8321" name="__module.model.gpt_neox.layers.35.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.35.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8322" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="6007"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="8323" name="Constant_436830" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8324" name="Constant_436829" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8325" name="__module.model.gpt_neox.layers.35.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8326" name="__module.model.gpt_neox.layers.35.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8327" name="__module.model.gpt_neox.layers.35.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6002" /> </output> </layer> <layer id="8328" name="__module.model.gpt_neox.layers.35.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8329" name="__module.model.gpt_neox.layers.35.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8330" name="__module.model.gpt_neox.layers.35.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6000" /> </output> </layer> <layer id="8331" name="__module.model.gpt_neox.layers.35.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6004,6005" /> </output> </layer> <layer id="8332" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8333" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8334" name="Constant_436828" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8335" name="ScatterUpdate_436831" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8336" name="Constant_436832" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8337" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8338" name="ScatterUpdate_436833" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8339" name="Constant_436836" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8340" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6008"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="8341" name="Constant_436908" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8342" name="Constant_436909" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8343" name="Constant_436905" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8344" name="Constant_436904" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8345" name="ScatterUpdate_436910" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8346" name="Constant_436913" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8347" name="__module.model.gpt_neox.layers.35.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6009,6010,causal_mask.211"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8348" name="__module.model.gpt_neox.layers.35.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8349" name="__module.model.gpt_neox.layers.35.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="5997,mask_value.71" /> </output> </layer> <layer id="8350" name="__module.model.gpt_neox.layers.35.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="6011,causal_mask.213"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8351" name="Add_455899" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8352" name="ShapeOf_455900" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8353" name="Constant_402800" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8354" name="__module.model.gpt_neox.layers.35.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8355" name="__module.model.gpt_neox.layers.35.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="5995"> <dim>1</dim> </port> </output> </layer> <layer id="8356" name="__module.model.gpt_neox.layers.35.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8357" name="__module.model.gpt_neox.layers.35.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8358" name="__module.model.gpt_neox.layers.35.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8359" name="__module.model.gpt_neox.layers.35.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="6012"> <dim>4</dim> </port> </output> </layer> <layer id="8360" name="__module.model.gpt_neox.layers.35.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8361" name="__module.model.gpt_neox.layers.35.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="8362" name="__module.model.gpt_neox.layers.35.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8363" name="__module.model.gpt_neox.layers.35.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8364" name="Maximum_455901" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8365" name="__module.model.gpt_neox.layers.35.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6014,attention_mask.79"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8366" name="__module.model.gpt_neox.layers.35.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6015,6016,sdpa_result.71,tensor.141"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8367" name="Constant_20940" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6017"> <dim>4</dim> </port> </output> </layer> <layer id="8368" name="__module.model.gpt_neox.layers.35.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6018"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8369" name="__module.model.gpt_neox.layers.35.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8370" name="__module.model.gpt_neox.layers.35.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="6023,input.567"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8371" name="Constant_20958" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="16604987928" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8372" name="Convert_550169" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8373" name="Constant_20958/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="16642736664" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8374" name="Convert_550172" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8375" name="Subtract_550173" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8376" name="Constant_550174" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="16642742808" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8377" name="Constant_20958/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8378" name="__module.model.gpt_neox.layers.35.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8379" name="Constant_467145" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16491502104" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8380" name="__module.model.gpt_neox.layers.35.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6026,6028,input.569"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8381" name="__module.model.gpt_neox.layers.35/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6054"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8382" name="__module.model.gpt_neox.layers.35/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6055,6057,input.579"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8383" name="__module.model.gpt_neox.layers.36.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8384" name="__module.model.gpt_neox.layers.36.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8385" name="Constant_467146" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16642767384" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8386" name="__module.model.gpt_neox.layers.36.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8387" name="Constant_467147" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16642791960" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8388" name="__module.model.gpt_neox.layers.36.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6192,input.587"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8389" name="Constant_21558" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="16642816536" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8390" name="Convert_550214" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8391" name="Constant_21558/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="16793811480" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8392" name="Convert_550217" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8393" name="Subtract_550218" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8394" name="Constant_550219" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="16793836056" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8395" name="Constant_21558/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8396" name="__module.model.gpt_neox.layers.36.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8397" name="Constant_467148" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="16793934360" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8398" name="__module.model.gpt_neox.layers.36.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="6197,input.589"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8399" name="__module.model.gpt_neox.layers.36.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="6206,input.591"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8400" name="Constant_21578" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="16794032664" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8401" name="Convert_550223" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8402" name="Constant_21578/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="16945027608" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8403" name="Convert_550226" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8404" name="Subtract_550227" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8405" name="Constant_550228" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="16945033752" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8406" name="Constant_21578/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8407" name="__module.model.gpt_neox.layers.36.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8408" name="Constant_467149" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16945058328" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8409" name="__module.model.gpt_neox.layers.36.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6209,input.593"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8410" name="__module.model.gpt_neox.layers.36.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8411" name="__module.model.gpt_neox.layers.36.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8412" name="Constant_467150" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16945082904" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8413" name="__module.model.gpt_neox.layers.36.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8414" name="Constant_467151" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16945107480" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8415" name="__module.model.gpt_neox.layers.36.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6067,input.581"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8416" name="Constant_21037" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="16945132056" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8417" name="Convert_550196" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8418" name="Constant_21037/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="17058378264" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8419" name="Convert_550199" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8420" name="Subtract_550200" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8421" name="Constant_550201" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="17058396696" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8422" name="Constant_21037/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8423" name="__module.model.gpt_neox.layers.36.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8424" name="Constant_467152" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="17058470424" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8425" name="__module.model.gpt_neox.layers.36.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="6074,qkv.145"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8426" name="__module.model.gpt_neox.layers.36.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8427" name="__module.model.gpt_neox.layers.36.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6078,qkv.147"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="8428" name="Constant_436921" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8429" name="Constant_436924" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8430" name="Constant_436927" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8431" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6079"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8432" name="Constant_21065" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6080"> <dim>4</dim> </port> </output> </layer> <layer id="8433" name="__module.model.gpt_neox.layers.36.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6081,query.289"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8434" name="Constant_436933" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8435" name="Constant_436936" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8436" name="Constant_436939" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8437" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6088,q.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8438" name="Constant_21141" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="6100,cos_cached.73"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8439" name="__module.model.gpt_neox.layers.36.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8440" name="Constant_436945" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8441" name="Constant_436948" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8442" name="Constant_436951" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8443" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6082"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8444" name="Constant_21079" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6083"> <dim>4</dim> </port> </output> </layer> <layer id="8445" name="__module.model.gpt_neox.layers.36.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6084,key.361"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8446" name="__module.model.gpt_neox.layers.36.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8447" name="__module.model.gpt_neox.layers.36.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8448" name="__module.model.gpt_neox.layers.36.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6092" /> </output> </layer> <layer id="8449" name="Constant_535940" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8450" name="Constant_535933" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8451" name="Constant_535935" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8452" name="Constant_535937" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8453" name="Concat_535938" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8454" name="Broadcast_535941" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="8455" name="ReadValue_522423" type="ReadValue" version="opset6"> <data variable_id="past_key_values.36.keypresent.36.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.36.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8456" name="Constant_509758" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8457" name="Gather_509759" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8458" name="__module.model.gpt_neox.layers.36.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8459" name="__module.model.gpt_neox.layers.36.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8460" name="__module.model.gpt_neox.layers.36.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6094" /> </output> </layer> <layer id="8461" name="__module.model.gpt_neox.layers.36.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6093,6097,6098,seq_len.145" /> </output> </layer> <layer id="8462" name="__module.model.gpt_neox.layers.36.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8463" name="__module.model.gpt_neox.layers.36.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8464" name="__module.model.gpt_neox.layers.36.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8465" name="__module.model.gpt_neox.layers.36.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6101,6104,cos.145"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8466" name="Constant_52119" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8467" name="__module.model.gpt_neox.layers.36.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6107"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8468" name="__module.model.gpt_neox.layers.36.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6108,cos.147"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8469" name="__module.model.gpt_neox.layers.36.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6112"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8470" name="Constant_436956" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8471" name="Constant_436955" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8472" name="__module.model.gpt_neox.layers.36.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8473" name="__module.model.gpt_neox.layers.36.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8474" name="__module.model.gpt_neox.layers.36.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6113,6118" /> </output> </layer> <layer id="8475" name="__module.model.gpt_neox.layers.36.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8476" name="__module.model.gpt_neox.layers.36.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6115,6116,6120,6121" /> </output> </layer> <layer id="8477" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8478" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8479" name="Constant_436954" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8480" name="ScatterUpdate_436957" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8481" name="Constant_436960" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8482" name="Constant_436963" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8483" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6122,x2.145"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8484" name="Constant_467153" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8485" name="__module.model.gpt_neox.layers.36.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6123"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8486" name="Constant_437009" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8487" name="Constant_437014" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8488" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6117,x1.145"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8489" name="__module.model.gpt_neox.layers.36.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6125"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8490" name="Constant_21139" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="6099,sin_cached.73"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8491" name="__module.model.gpt_neox.layers.36.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8492" name="__module.model.gpt_neox.layers.36.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8493" name="__module.model.gpt_neox.layers.36.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6102,6105,sin.145"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8494" name="Constant_52122" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8495" name="__module.model.gpt_neox.layers.36.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6110"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8496" name="__module.model.gpt_neox.layers.36.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6111,sin.147"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8497" name="__module.model.gpt_neox.layers.36.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6126"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8498" name="__module.model.gpt_neox.layers.36.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6127,query.291"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8499" name="Constant_437061" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8500" name="Constant_437064" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8501" name="Constant_437067" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8502" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6089,query_pass.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8503" name="__module.model.gpt_neox.layers.36.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6145,6155,query.293,query.295"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8504" name="Constant_437073" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8505" name="Constant_437076" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8506" name="Constant_437079" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8507" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6090,k.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8508" name="__module.model.gpt_neox.layers.36.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6128"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8509" name="Constant_437083" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8510" name="Constant_437082" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8511" name="__module.model.gpt_neox.layers.36.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8512" name="__module.model.gpt_neox.layers.36.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8513" name="__module.model.gpt_neox.layers.36.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6129,6134" /> </output> </layer> <layer id="8514" name="__module.model.gpt_neox.layers.36.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8515" name="__module.model.gpt_neox.layers.36.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6131,6132,6136,6137" /> </output> </layer> <layer id="8516" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8517" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8518" name="Constant_437081" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8519" name="ScatterUpdate_437084" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8520" name="Constant_437087" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8521" name="Constant_437090" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8522" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6138,x2.147"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8523" name="Constant_467154" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8524" name="__module.model.gpt_neox.layers.36.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6139"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8525" name="Constant_437136" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8526" name="Constant_437141" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8527" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6133,x1.147"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8528" name="__module.model.gpt_neox.layers.36.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6141"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8529" name="__module.model.gpt_neox.layers.36.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6142"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8530" name="__module.model.gpt_neox.layers.36.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6143,key.363"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8531" name="Constant_437187" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8532" name="Constant_437190" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8533" name="Constant_437193" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8534" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6091,key_pass.73"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8535" name="__module.model.gpt_neox.layers.36.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6147,key.365"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8536" name="__module.model.gpt_neox.layers.36.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.36.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8537" name="Constant_535953" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8538" name="Constant_535946" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8539" name="Constant_535948" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8540" name="Constant_535950" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8541" name="Concat_535951" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8542" name="Broadcast_535954" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="8543" name="ReadValue_522425" type="ReadValue" version="opset6"> <data variable_id="past_key_values.36.valuepresent.36.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.36.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8544" name="Constant_509761" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8545" name="Gather_509762" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8546" name="Constant_437199" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8547" name="Constant_437202" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8548" name="Constant_437205" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8549" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6085"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8550" name="Constant_21093" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6086"> <dim>4</dim> </port> </output> </layer> <layer id="8551" name="__module.model.gpt_neox.layers.36.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6087,value.145"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8552" name="__module.model.gpt_neox.layers.36.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.36.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8553" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="6164"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="8554" name="Constant_437209" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8555" name="Constant_437208" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8556" name="__module.model.gpt_neox.layers.36.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8557" name="__module.model.gpt_neox.layers.36.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8558" name="__module.model.gpt_neox.layers.36.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6159" /> </output> </layer> <layer id="8559" name="__module.model.gpt_neox.layers.36.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8560" name="__module.model.gpt_neox.layers.36.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8561" name="__module.model.gpt_neox.layers.36.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6157" /> </output> </layer> <layer id="8562" name="__module.model.gpt_neox.layers.36.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6161,6162" /> </output> </layer> <layer id="8563" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8564" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8565" name="Constant_437207" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8566" name="ScatterUpdate_437210" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8567" name="Constant_437211" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8568" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8569" name="ScatterUpdate_437212" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8570" name="Constant_437215" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8571" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6165"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="8572" name="Constant_437287" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8573" name="Constant_437288" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8574" name="Constant_437284" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8575" name="Constant_437283" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8576" name="ScatterUpdate_437289" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8577" name="Constant_437292" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8578" name="__module.model.gpt_neox.layers.36.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6166,6167,causal_mask.217"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8579" name="__module.model.gpt_neox.layers.36.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8580" name="__module.model.gpt_neox.layers.36.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="6154,mask_value.73" /> </output> </layer> <layer id="8581" name="__module.model.gpt_neox.layers.36.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="6168,causal_mask.219"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8582" name="Add_455975" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8583" name="ShapeOf_455976" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8584" name="Constant_402815" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8585" name="__module.model.gpt_neox.layers.36.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8586" name="__module.model.gpt_neox.layers.36.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6152"> <dim>1</dim> </port> </output> </layer> <layer id="8587" name="__module.model.gpt_neox.layers.36.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8588" name="__module.model.gpt_neox.layers.36.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8589" name="__module.model.gpt_neox.layers.36.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8590" name="__module.model.gpt_neox.layers.36.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="6169"> <dim>4</dim> </port> </output> </layer> <layer id="8591" name="__module.model.gpt_neox.layers.36.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8592" name="__module.model.gpt_neox.layers.36.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="8593" name="__module.model.gpt_neox.layers.36.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8594" name="__module.model.gpt_neox.layers.36.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8595" name="Maximum_455977" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8596" name="__module.model.gpt_neox.layers.36.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6171,attention_mask.81"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8597" name="__module.model.gpt_neox.layers.36.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6172,6173,sdpa_result.73,tensor.145"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8598" name="Constant_21518" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6174"> <dim>4</dim> </port> </output> </layer> <layer id="8599" name="__module.model.gpt_neox.layers.36.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6175"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8600" name="__module.model.gpt_neox.layers.36.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8601" name="__module.model.gpt_neox.layers.36.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="6180,input.583"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8602" name="Constant_21536" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="17058544152" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8603" name="Convert_550205" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8604" name="Constant_21536/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="17096292888" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8605" name="Convert_550208" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8606" name="Subtract_550209" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8607" name="Constant_550210" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="17096299032" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8608" name="Constant_21536/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8609" name="__module.model.gpt_neox.layers.36.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8610" name="Constant_467155" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="16945058328" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8611" name="__module.model.gpt_neox.layers.36.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6183,6185,input.585"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8612" name="__module.model.gpt_neox.layers.36/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6211"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8613" name="__module.model.gpt_neox.layers.36/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6212,6214,input.595"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8614" name="__module.model.gpt_neox.layers.37.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8615" name="__module.model.gpt_neox.layers.37.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8616" name="Constant_467156" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17096323608" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8617" name="__module.model.gpt_neox.layers.37.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8618" name="Constant_467157" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17096348184" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8619" name="__module.model.gpt_neox.layers.37.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6349,input.603"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8620" name="Constant_22136" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="17096372760" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8621" name="Convert_550250" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8622" name="Constant_22136/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="17247367704" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8623" name="Convert_550253" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8624" name="Subtract_550254" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8625" name="Constant_550255" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="17247392280" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8626" name="Constant_22136/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8627" name="__module.model.gpt_neox.layers.37.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8628" name="Constant_467158" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="17247490584" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8629" name="__module.model.gpt_neox.layers.37.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="6354,input.605"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8630" name="__module.model.gpt_neox.layers.37.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="6363,input.607"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8631" name="Constant_22156" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="17247588888" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8632" name="Convert_550259" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8633" name="Constant_22156/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="17398583832" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8634" name="Convert_550262" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8635" name="Subtract_550263" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8636" name="Constant_550264" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="17398589976" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8637" name="Constant_22156/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8638" name="__module.model.gpt_neox.layers.37.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8639" name="Constant_467159" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17398614552" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8640" name="__module.model.gpt_neox.layers.37.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6366,input.609"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8641" name="__module.model.gpt_neox.layers.37.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8642" name="__module.model.gpt_neox.layers.37.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8643" name="Constant_467160" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17398639128" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8644" name="__module.model.gpt_neox.layers.37.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8645" name="Constant_467161" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17398663704" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8646" name="__module.model.gpt_neox.layers.37.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6224,input.597"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8647" name="Constant_21615" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="17398688280" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8648" name="Convert_550232" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8649" name="Constant_21615/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="17511934488" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8650" name="Convert_550235" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8651" name="Subtract_550236" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8652" name="Constant_550237" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="17511952920" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8653" name="Constant_21615/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8654" name="__module.model.gpt_neox.layers.37.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8655" name="Constant_467162" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="17512026648" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8656" name="__module.model.gpt_neox.layers.37.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="6231,qkv.149"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8657" name="__module.model.gpt_neox.layers.37.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8658" name="__module.model.gpt_neox.layers.37.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6235,qkv.151"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="8659" name="Constant_437300" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8660" name="Constant_437303" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8661" name="Constant_437306" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8662" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6236"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8663" name="Constant_21643" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6237"> <dim>4</dim> </port> </output> </layer> <layer id="8664" name="__module.model.gpt_neox.layers.37.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6238,query.297"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8665" name="Constant_437312" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8666" name="Constant_437315" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8667" name="Constant_437318" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8668" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6245,q.75"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8669" name="Constant_21719" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="6257,cos_cached.75"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8670" name="__module.model.gpt_neox.layers.37.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8671" name="Constant_437324" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8672" name="Constant_437327" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8673" name="Constant_437330" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8674" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6239"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8675" name="Constant_21657" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6240"> <dim>4</dim> </port> </output> </layer> <layer id="8676" name="__module.model.gpt_neox.layers.37.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6241,key.371"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8677" name="__module.model.gpt_neox.layers.37.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8678" name="__module.model.gpt_neox.layers.37.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8679" name="__module.model.gpt_neox.layers.37.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6249" /> </output> </layer> <layer id="8680" name="Constant_535966" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8681" name="Constant_535959" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8682" name="Constant_535961" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8683" name="Constant_535963" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8684" name="Concat_535964" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8685" name="Broadcast_535967" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="8686" name="ReadValue_522427" type="ReadValue" version="opset6"> <data variable_id="past_key_values.37.keypresent.37.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.37.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8687" name="Constant_509764" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8688" name="Gather_509765" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8689" name="__module.model.gpt_neox.layers.37.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8690" name="__module.model.gpt_neox.layers.37.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8691" name="__module.model.gpt_neox.layers.37.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6251" /> </output> </layer> <layer id="8692" name="__module.model.gpt_neox.layers.37.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6250,6254,6255,seq_len.149" /> </output> </layer> <layer id="8693" name="__module.model.gpt_neox.layers.37.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8694" name="__module.model.gpt_neox.layers.37.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8695" name="__module.model.gpt_neox.layers.37.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8696" name="__module.model.gpt_neox.layers.37.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6258,6261,cos.149"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8697" name="Constant_52125" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8698" name="__module.model.gpt_neox.layers.37.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6264"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8699" name="__module.model.gpt_neox.layers.37.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6265,cos.151"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8700" name="__module.model.gpt_neox.layers.37.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6269"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8701" name="Constant_437335" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8702" name="Constant_437334" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8703" name="__module.model.gpt_neox.layers.37.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8704" name="__module.model.gpt_neox.layers.37.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8705" name="__module.model.gpt_neox.layers.37.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6270,6275" /> </output> </layer> <layer id="8706" name="__module.model.gpt_neox.layers.37.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8707" name="__module.model.gpt_neox.layers.37.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6272,6273,6277,6278" /> </output> </layer> <layer id="8708" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8709" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8710" name="Constant_437333" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8711" name="ScatterUpdate_437336" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8712" name="Constant_437339" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8713" name="Constant_437342" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8714" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6279,x2.149"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8715" name="Constant_467163" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8716" name="__module.model.gpt_neox.layers.37.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6280"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8717" name="Constant_437388" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8718" name="Constant_437393" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8719" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6274,x1.149"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8720" name="__module.model.gpt_neox.layers.37.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6282"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8721" name="Constant_21717" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="6256,sin_cached.75"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8722" name="__module.model.gpt_neox.layers.37.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8723" name="__module.model.gpt_neox.layers.37.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8724" name="__module.model.gpt_neox.layers.37.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6259,6262,sin.149"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8725" name="Constant_52128" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8726" name="__module.model.gpt_neox.layers.37.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6267"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8727" name="__module.model.gpt_neox.layers.37.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6268,sin.151"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8728" name="__module.model.gpt_neox.layers.37.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6283"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8729" name="__module.model.gpt_neox.layers.37.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6284,query.299"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8730" name="Constant_437440" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8731" name="Constant_437443" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8732" name="Constant_437446" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8733" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6246,query_pass.75"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8734" name="__module.model.gpt_neox.layers.37.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6302,6312,query.301,query.303"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8735" name="Constant_437452" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8736" name="Constant_437455" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8737" name="Constant_437458" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8738" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6247,k.75"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8739" name="__module.model.gpt_neox.layers.37.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6285"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8740" name="Constant_437462" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8741" name="Constant_437461" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8742" name="__module.model.gpt_neox.layers.37.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8743" name="__module.model.gpt_neox.layers.37.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8744" name="__module.model.gpt_neox.layers.37.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6286,6291" /> </output> </layer> <layer id="8745" name="__module.model.gpt_neox.layers.37.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8746" name="__module.model.gpt_neox.layers.37.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6288,6289,6293,6294" /> </output> </layer> <layer id="8747" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8748" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8749" name="Constant_437460" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8750" name="ScatterUpdate_437463" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8751" name="Constant_437466" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8752" name="Constant_437469" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8753" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6295,x2.151"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8754" name="Constant_467164" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8755" name="__module.model.gpt_neox.layers.37.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6296"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8756" name="Constant_437515" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8757" name="Constant_437520" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8758" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6290,x1.151"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8759" name="__module.model.gpt_neox.layers.37.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6298"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8760" name="__module.model.gpt_neox.layers.37.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6299"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8761" name="__module.model.gpt_neox.layers.37.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6300,key.373"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8762" name="Constant_437566" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8763" name="Constant_437569" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8764" name="Constant_437572" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8765" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6248,key_pass.75"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8766" name="__module.model.gpt_neox.layers.37.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6304,key.375"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8767" name="__module.model.gpt_neox.layers.37.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.37.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8768" name="Constant_535979" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8769" name="Constant_535972" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8770" name="Constant_535974" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8771" name="Constant_535976" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8772" name="Concat_535977" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8773" name="Broadcast_535980" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="8774" name="ReadValue_522429" type="ReadValue" version="opset6"> <data variable_id="past_key_values.37.valuepresent.37.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.37.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8775" name="Constant_509767" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8776" name="Gather_509768" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8777" name="Constant_437578" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8778" name="Constant_437581" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8779" name="Constant_437584" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8780" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6242"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8781" name="Constant_21671" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6243"> <dim>4</dim> </port> </output> </layer> <layer id="8782" name="__module.model.gpt_neox.layers.37.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6244,value.149"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8783" name="__module.model.gpt_neox.layers.37.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.37.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8784" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="6321"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="8785" name="Constant_437588" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8786" name="Constant_437587" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8787" name="__module.model.gpt_neox.layers.37.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8788" name="__module.model.gpt_neox.layers.37.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8789" name="__module.model.gpt_neox.layers.37.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6316" /> </output> </layer> <layer id="8790" name="__module.model.gpt_neox.layers.37.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8791" name="__module.model.gpt_neox.layers.37.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8792" name="__module.model.gpt_neox.layers.37.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6314" /> </output> </layer> <layer id="8793" name="__module.model.gpt_neox.layers.37.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6318,6319" /> </output> </layer> <layer id="8794" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8795" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8796" name="Constant_437586" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8797" name="ScatterUpdate_437589" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8798" name="Constant_437590" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8799" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8800" name="ScatterUpdate_437591" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8801" name="Constant_437594" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8802" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6322"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="8803" name="Constant_437666" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8804" name="Constant_437667" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8805" name="Constant_437663" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8806" name="Constant_437662" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8807" name="ScatterUpdate_437668" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8808" name="Constant_437671" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8809" name="__module.model.gpt_neox.layers.37.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6323,6324,causal_mask.223"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8810" name="__module.model.gpt_neox.layers.37.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8811" name="__module.model.gpt_neox.layers.37.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="6311,mask_value.75" /> </output> </layer> <layer id="8812" name="__module.model.gpt_neox.layers.37.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="6325,causal_mask.225"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8813" name="Add_456051" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8814" name="ShapeOf_456052" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8815" name="Constant_402830" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8816" name="__module.model.gpt_neox.layers.37.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8817" name="__module.model.gpt_neox.layers.37.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6309"> <dim>1</dim> </port> </output> </layer> <layer id="8818" name="__module.model.gpt_neox.layers.37.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8819" name="__module.model.gpt_neox.layers.37.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8820" name="__module.model.gpt_neox.layers.37.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8821" name="__module.model.gpt_neox.layers.37.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="6326"> <dim>4</dim> </port> </output> </layer> <layer id="8822" name="__module.model.gpt_neox.layers.37.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8823" name="__module.model.gpt_neox.layers.37.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="8824" name="__module.model.gpt_neox.layers.37.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8825" name="__module.model.gpt_neox.layers.37.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8826" name="Maximum_456053" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8827" name="__module.model.gpt_neox.layers.37.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6328,attention_mask.83"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="8828" name="__module.model.gpt_neox.layers.37.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6329,6330,sdpa_result.75,tensor.149"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8829" name="Constant_22096" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6331"> <dim>4</dim> </port> </output> </layer> <layer id="8830" name="__module.model.gpt_neox.layers.37.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6332"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8831" name="__module.model.gpt_neox.layers.37.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="8832" name="__module.model.gpt_neox.layers.37.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="6337,input.599"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8833" name="Constant_22114" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="17512100376" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8834" name="Convert_550241" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8835" name="Constant_22114/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="17549849112" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8836" name="Convert_550244" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8837" name="Subtract_550245" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8838" name="Constant_550246" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="17549855256" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8839" name="Constant_22114/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8840" name="__module.model.gpt_neox.layers.37.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8841" name="Constant_467165" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17398614552" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8842" name="__module.model.gpt_neox.layers.37.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6340,6342,input.601"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8843" name="__module.model.gpt_neox.layers.37/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6368"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8844" name="__module.model.gpt_neox.layers.37/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6369,6371,input.611"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8845" name="__module.model.gpt_neox.layers.38.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8846" name="__module.model.gpt_neox.layers.38.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8847" name="Constant_467166" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17549879832" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8848" name="__module.model.gpt_neox.layers.38.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8849" name="Constant_467167" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17549904408" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8850" name="__module.model.gpt_neox.layers.38.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6506,input.619"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8851" name="Constant_22714" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="17549928984" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8852" name="Convert_550286" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8853" name="Constant_22714/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="17700923928" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8854" name="Convert_550289" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8855" name="Subtract_550290" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8856" name="Constant_550291" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="17700948504" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="8857" name="Constant_22714/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8858" name="__module.model.gpt_neox.layers.38.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8859" name="Constant_467168" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="17701046808" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8860" name="__module.model.gpt_neox.layers.38.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="6511,input.621"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8861" name="__module.model.gpt_neox.layers.38.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="6520,input.623"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8862" name="Constant_22734" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="17701145112" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8863" name="Convert_550295" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8864" name="Constant_22734/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="17852140056" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8865" name="Convert_550298" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8866" name="Subtract_550299" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8867" name="Constant_550300" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="17852146200" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="8868" name="Constant_22734/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="8869" name="__module.model.gpt_neox.layers.38.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8870" name="Constant_467169" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17852170776" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8871" name="__module.model.gpt_neox.layers.38.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6523,input.625"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8872" name="__module.model.gpt_neox.layers.38.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8873" name="__module.model.gpt_neox.layers.38.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8874" name="Constant_467170" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17852195352" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8875" name="__module.model.gpt_neox.layers.38.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8876" name="Constant_467171" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17852219928" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8877" name="__module.model.gpt_neox.layers.38.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6381,input.613"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8878" name="Constant_22193" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="17852244504" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8879" name="Convert_550268" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8880" name="Constant_22193/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="17965490712" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8881" name="Convert_550271" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8882" name="Subtract_550272" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8883" name="Constant_550273" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="17965509144" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="8884" name="Constant_22193/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="8885" name="__module.model.gpt_neox.layers.38.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8886" name="Constant_467172" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="17965582872" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8887" name="__module.model.gpt_neox.layers.38.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="6388,qkv.153"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="8888" name="__module.model.gpt_neox.layers.38.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8889" name="__module.model.gpt_neox.layers.38.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6392,qkv.155"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="8890" name="Constant_437679" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8891" name="Constant_437682" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8892" name="Constant_437685" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8893" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6393"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8894" name="Constant_22221" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6394"> <dim>4</dim> </port> </output> </layer> <layer id="8895" name="__module.model.gpt_neox.layers.38.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6395,query.305"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8896" name="Constant_437691" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8897" name="Constant_437694" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8898" name="Constant_437697" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8899" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6402,q.77"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8900" name="Constant_22297" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="6414,cos_cached.77"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8901" name="__module.model.gpt_neox.layers.38.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8902" name="Constant_437703" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8903" name="Constant_437706" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8904" name="Constant_437709" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8905" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6396"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="8906" name="Constant_22235" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6397"> <dim>4</dim> </port> </output> </layer> <layer id="8907" name="__module.model.gpt_neox.layers.38.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6398,key.381"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8908" name="__module.model.gpt_neox.layers.38.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8909" name="__module.model.gpt_neox.layers.38.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8910" name="__module.model.gpt_neox.layers.38.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6406" /> </output> </layer> <layer id="8911" name="Constant_535992" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="8912" name="Constant_535985" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8913" name="Constant_535987" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8914" name="Constant_535989" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8915" name="Concat_535990" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8916" name="Broadcast_535993" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="8917" name="ReadValue_522431" type="ReadValue" version="opset6"> <data variable_id="past_key_values.38.keypresent.38.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.38.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8918" name="Constant_509770" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8919" name="Gather_509771" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8920" name="__module.model.gpt_neox.layers.38.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8921" name="__module.model.gpt_neox.layers.38.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8922" name="__module.model.gpt_neox.layers.38.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6408" /> </output> </layer> <layer id="8923" name="__module.model.gpt_neox.layers.38.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6407,6411,6412,seq_len.153" /> </output> </layer> <layer id="8924" name="__module.model.gpt_neox.layers.38.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8925" name="__module.model.gpt_neox.layers.38.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8926" name="__module.model.gpt_neox.layers.38.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8927" name="__module.model.gpt_neox.layers.38.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6415,6418,cos.153"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8928" name="Constant_52131" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8929" name="__module.model.gpt_neox.layers.38.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6421"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8930" name="__module.model.gpt_neox.layers.38.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6422,cos.155"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8931" name="__module.model.gpt_neox.layers.38.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6426"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8932" name="Constant_437714" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8933" name="Constant_437713" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8934" name="__module.model.gpt_neox.layers.38.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8935" name="__module.model.gpt_neox.layers.38.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8936" name="__module.model.gpt_neox.layers.38.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6427,6432" /> </output> </layer> <layer id="8937" name="__module.model.gpt_neox.layers.38.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8938" name="__module.model.gpt_neox.layers.38.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6429,6430,6434,6435" /> </output> </layer> <layer id="8939" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8940" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8941" name="Constant_437712" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8942" name="ScatterUpdate_437715" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8943" name="Constant_437718" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8944" name="Constant_437721" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8945" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6436,x2.153"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8946" name="Constant_467173" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8947" name="__module.model.gpt_neox.layers.38.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6437"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8948" name="Constant_437767" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8949" name="Constant_437772" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8950" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6431,x1.153"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8951" name="__module.model.gpt_neox.layers.38.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6439"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8952" name="Constant_22295" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="6413,sin_cached.77"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="8953" name="__module.model.gpt_neox.layers.38.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8954" name="__module.model.gpt_neox.layers.38.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8955" name="__module.model.gpt_neox.layers.38.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6416,6419,sin.153"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8956" name="Constant_52134" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8957" name="__module.model.gpt_neox.layers.38.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6424"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8958" name="__module.model.gpt_neox.layers.38.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6425,sin.155"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8959" name="__module.model.gpt_neox.layers.38.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6440"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8960" name="__module.model.gpt_neox.layers.38.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6441,query.307"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8961" name="Constant_437819" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8962" name="Constant_437822" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8963" name="Constant_437825" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8964" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6403,query_pass.77"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8965" name="__module.model.gpt_neox.layers.38.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6459,6469,query.309,query.311"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8966" name="Constant_437831" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8967" name="Constant_437834" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8968" name="Constant_437837" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8969" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6404,k.77"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8970" name="__module.model.gpt_neox.layers.38.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6442"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8971" name="Constant_437841" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8972" name="Constant_437840" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="8973" name="__module.model.gpt_neox.layers.38.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8974" name="__module.model.gpt_neox.layers.38.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8975" name="__module.model.gpt_neox.layers.38.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6443,6448" /> </output> </layer> <layer id="8976" name="__module.model.gpt_neox.layers.38.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="8977" name="__module.model.gpt_neox.layers.38.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6445,6446,6450,6451" /> </output> </layer> <layer id="8978" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="8979" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8980" name="Constant_437839" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="8981" name="ScatterUpdate_437842" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8982" name="Constant_437845" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8983" name="Constant_437848" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8984" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6452,x2.155"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8985" name="Constant_467174" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="8986" name="__module.model.gpt_neox.layers.38.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6453"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8987" name="Constant_437894" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8988" name="Constant_437899" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8989" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6447,x1.155"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="8990" name="__module.model.gpt_neox.layers.38.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6455"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8991" name="__module.model.gpt_neox.layers.38.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6456"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8992" name="__module.model.gpt_neox.layers.38.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6457,key.383"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="8993" name="Constant_437945" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8994" name="Constant_437948" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="8995" name="Constant_437951" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="8996" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6405,key_pass.77"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="8997" name="__module.model.gpt_neox.layers.38.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6461,key.385"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8998" name="__module.model.gpt_neox.layers.38.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.38.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="8999" name="Constant_536005" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9000" name="Constant_535998" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9001" name="Constant_536000" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9002" name="Constant_536002" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9003" name="Concat_536003" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9004" name="Broadcast_536006" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9005" name="ReadValue_522433" type="ReadValue" version="opset6"> <data variable_id="past_key_values.38.valuepresent.38.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.38.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9006" name="Constant_509773" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9007" name="Gather_509774" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9008" name="Constant_437957" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9009" name="Constant_437960" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9010" name="Constant_437963" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9011" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6399"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9012" name="Constant_22249" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6400"> <dim>4</dim> </port> </output> </layer> <layer id="9013" name="__module.model.gpt_neox.layers.38.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6401,value.153"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9014" name="__module.model.gpt_neox.layers.38.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.38.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9015" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="6478"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9016" name="Constant_437967" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9017" name="Constant_437966" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9018" name="__module.model.gpt_neox.layers.38.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9019" name="__module.model.gpt_neox.layers.38.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9020" name="__module.model.gpt_neox.layers.38.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6473" /> </output> </layer> <layer id="9021" name="__module.model.gpt_neox.layers.38.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9022" name="__module.model.gpt_neox.layers.38.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9023" name="__module.model.gpt_neox.layers.38.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6471" /> </output> </layer> <layer id="9024" name="__module.model.gpt_neox.layers.38.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6475,6476" /> </output> </layer> <layer id="9025" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9026" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9027" name="Constant_437965" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9028" name="ScatterUpdate_437968" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9029" name="Constant_437969" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9030" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9031" name="ScatterUpdate_437970" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9032" name="Constant_437973" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9033" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6479"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9034" name="Constant_438045" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9035" name="Constant_438046" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9036" name="Constant_438042" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9037" name="Constant_438041" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9038" name="ScatterUpdate_438047" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9039" name="Constant_438050" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9040" name="__module.model.gpt_neox.layers.38.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6480,6481,causal_mask.229"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9041" name="__module.model.gpt_neox.layers.38.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9042" name="__module.model.gpt_neox.layers.38.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="6468,mask_value.77" /> </output> </layer> <layer id="9043" name="__module.model.gpt_neox.layers.38.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="6482,causal_mask.231"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9044" name="Add_456127" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9045" name="ShapeOf_456128" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9046" name="Constant_402845" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9047" name="__module.model.gpt_neox.layers.38.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9048" name="__module.model.gpt_neox.layers.38.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6466"> <dim>1</dim> </port> </output> </layer> <layer id="9049" name="__module.model.gpt_neox.layers.38.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9050" name="__module.model.gpt_neox.layers.38.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9051" name="__module.model.gpt_neox.layers.38.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9052" name="__module.model.gpt_neox.layers.38.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="6483"> <dim>4</dim> </port> </output> </layer> <layer id="9053" name="__module.model.gpt_neox.layers.38.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9054" name="__module.model.gpt_neox.layers.38.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="9055" name="__module.model.gpt_neox.layers.38.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9056" name="__module.model.gpt_neox.layers.38.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9057" name="Maximum_456129" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9058" name="__module.model.gpt_neox.layers.38.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6485,attention_mask.85"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9059" name="__module.model.gpt_neox.layers.38.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6486,6487,sdpa_result.77,tensor.153"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9060" name="Constant_22674" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6488"> <dim>4</dim> </port> </output> </layer> <layer id="9061" name="__module.model.gpt_neox.layers.38.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6489"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9062" name="__module.model.gpt_neox.layers.38.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9063" name="__module.model.gpt_neox.layers.38.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="6494,input.615"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9064" name="Constant_22692" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="17965656600" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9065" name="Convert_550277" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9066" name="Constant_22692/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="18003405336" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9067" name="Convert_550280" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9068" name="Subtract_550281" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9069" name="Constant_550282" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="18003411480" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9070" name="Constant_22692/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9071" name="__module.model.gpt_neox.layers.38.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9072" name="Constant_467175" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="17852170776" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9073" name="__module.model.gpt_neox.layers.38.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6497,6499,input.617"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9074" name="__module.model.gpt_neox.layers.38/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6525"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9075" name="__module.model.gpt_neox.layers.38/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6526,6528,input.627"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9076" name="__module.model.gpt_neox.layers.39.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9077" name="__module.model.gpt_neox.layers.39.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9078" name="Constant_467176" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18003436056" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9079" name="__module.model.gpt_neox.layers.39.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9080" name="Constant_467177" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18003460632" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9081" name="__module.model.gpt_neox.layers.39.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6663,input.635"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9082" name="Constant_23292" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="18003485208" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9083" name="Convert_550322" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9084" name="Constant_23292/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="18154480152" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9085" name="Convert_550325" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9086" name="Subtract_550326" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9087" name="Constant_550327" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="18154504728" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9088" name="Constant_23292/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9089" name="__module.model.gpt_neox.layers.39.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9090" name="Constant_467178" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="18154603032" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9091" name="__module.model.gpt_neox.layers.39.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="6668,input.637"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9092" name="__module.model.gpt_neox.layers.39.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="6677,input.639"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9093" name="Constant_23312" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="18154701336" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9094" name="Convert_550331" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9095" name="Constant_23312/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="18305696280" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9096" name="Convert_550334" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9097" name="Subtract_550335" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9098" name="Constant_550336" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="18305702424" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9099" name="Constant_23312/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9100" name="__module.model.gpt_neox.layers.39.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9101" name="Constant_467179" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18305727000" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9102" name="__module.model.gpt_neox.layers.39.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6680,input.641"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9103" name="__module.model.gpt_neox.layers.39.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9104" name="__module.model.gpt_neox.layers.39.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9105" name="Constant_467180" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18305751576" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9106" name="__module.model.gpt_neox.layers.39.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9107" name="Constant_467181" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18305776152" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9108" name="__module.model.gpt_neox.layers.39.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6538,input.629"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9109" name="Constant_22771" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="18305800728" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9110" name="Convert_550304" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9111" name="Constant_22771/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="18419046936" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9112" name="Convert_550307" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9113" name="Subtract_550308" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9114" name="Constant_550309" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="18419065368" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9115" name="Constant_22771/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9116" name="__module.model.gpt_neox.layers.39.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9117" name="Constant_467182" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="18419139096" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9118" name="__module.model.gpt_neox.layers.39.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="6545,qkv.157"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9119" name="__module.model.gpt_neox.layers.39.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9120" name="__module.model.gpt_neox.layers.39.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6549,qkv.159"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="9121" name="Constant_438058" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9122" name="Constant_438061" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9123" name="Constant_438064" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9124" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6550"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9125" name="Constant_22799" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6551"> <dim>4</dim> </port> </output> </layer> <layer id="9126" name="__module.model.gpt_neox.layers.39.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6552,query.313"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9127" name="Constant_438070" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9128" name="Constant_438073" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9129" name="Constant_438076" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9130" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6559,q.79"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9131" name="Constant_22875" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="6571,cos_cached.79"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="9132" name="__module.model.gpt_neox.layers.39.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9133" name="Constant_438082" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9134" name="Constant_438085" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9135" name="Constant_438088" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9136" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6553"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9137" name="Constant_22813" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6554"> <dim>4</dim> </port> </output> </layer> <layer id="9138" name="__module.model.gpt_neox.layers.39.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6555,key.391"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9139" name="__module.model.gpt_neox.layers.39.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9140" name="__module.model.gpt_neox.layers.39.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9141" name="__module.model.gpt_neox.layers.39.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6563" /> </output> </layer> <layer id="9142" name="Constant_536018" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9143" name="Constant_536011" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9144" name="Constant_536013" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9145" name="Constant_536015" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9146" name="Concat_536016" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9147" name="Broadcast_536019" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9148" name="ReadValue_522435" type="ReadValue" version="opset6"> <data variable_id="past_key_values.39.keypresent.39.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.39.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9149" name="Constant_509776" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9150" name="Gather_509777" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9151" name="__module.model.gpt_neox.layers.39.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9152" name="__module.model.gpt_neox.layers.39.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9153" name="__module.model.gpt_neox.layers.39.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6565" /> </output> </layer> <layer id="9154" name="__module.model.gpt_neox.layers.39.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6564,6568,6569,seq_len.157" /> </output> </layer> <layer id="9155" name="__module.model.gpt_neox.layers.39.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9156" name="__module.model.gpt_neox.layers.39.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9157" name="__module.model.gpt_neox.layers.39.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9158" name="__module.model.gpt_neox.layers.39.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6572,6575,cos.157"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9159" name="Constant_52137" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9160" name="__module.model.gpt_neox.layers.39.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6578"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9161" name="__module.model.gpt_neox.layers.39.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6579,cos.159"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9162" name="__module.model.gpt_neox.layers.39.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6583"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9163" name="Constant_438093" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9164" name="Constant_438092" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9165" name="__module.model.gpt_neox.layers.39.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9166" name="__module.model.gpt_neox.layers.39.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9167" name="__module.model.gpt_neox.layers.39.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6584,6589" /> </output> </layer> <layer id="9168" name="__module.model.gpt_neox.layers.39.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="9169" name="__module.model.gpt_neox.layers.39.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6586,6587,6591,6592" /> </output> </layer> <layer id="9170" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9171" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9172" name="Constant_438091" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9173" name="ScatterUpdate_438094" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9174" name="Constant_438097" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9175" name="Constant_438100" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9176" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6593,x2.157"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9177" name="Constant_467183" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="9178" name="__module.model.gpt_neox.layers.39.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6594"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9179" name="Constant_438146" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9180" name="Constant_438151" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9181" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6588,x1.157"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9182" name="__module.model.gpt_neox.layers.39.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6596"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9183" name="Constant_22873" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="6570,sin_cached.79"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="9184" name="__module.model.gpt_neox.layers.39.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9185" name="__module.model.gpt_neox.layers.39.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9186" name="__module.model.gpt_neox.layers.39.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6573,6576,sin.157"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9187" name="Constant_52140" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9188" name="__module.model.gpt_neox.layers.39.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6581"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9189" name="__module.model.gpt_neox.layers.39.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6582,sin.159"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9190" name="__module.model.gpt_neox.layers.39.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6597"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9191" name="__module.model.gpt_neox.layers.39.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6598,query.315"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9192" name="Constant_438198" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9193" name="Constant_438201" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9194" name="Constant_438204" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9195" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6560,query_pass.79"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="9196" name="__module.model.gpt_neox.layers.39.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6616,6626,query.317,query.319"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9197" name="Constant_438210" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9198" name="Constant_438213" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9199" name="Constant_438216" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9200" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6561,k.79"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9201" name="__module.model.gpt_neox.layers.39.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6599"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9202" name="Constant_438220" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9203" name="Constant_438219" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9204" name="__module.model.gpt_neox.layers.39.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9205" name="__module.model.gpt_neox.layers.39.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9206" name="__module.model.gpt_neox.layers.39.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6600,6605" /> </output> </layer> <layer id="9207" name="__module.model.gpt_neox.layers.39.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="9208" name="__module.model.gpt_neox.layers.39.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6602,6603,6607,6608" /> </output> </layer> <layer id="9209" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9210" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9211" name="Constant_438218" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9212" name="ScatterUpdate_438221" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9213" name="Constant_438224" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9214" name="Constant_438227" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9215" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6609,x2.159"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9216" name="Constant_467184" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="9217" name="__module.model.gpt_neox.layers.39.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6610"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9218" name="Constant_438273" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9219" name="Constant_438278" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9220" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6604,x1.159"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9221" name="__module.model.gpt_neox.layers.39.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6612"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9222" name="__module.model.gpt_neox.layers.39.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6613"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9223" name="__module.model.gpt_neox.layers.39.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6614,key.393"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9224" name="Constant_438324" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9225" name="Constant_438327" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9226" name="Constant_438330" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9227" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6562,key_pass.79"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="9228" name="__module.model.gpt_neox.layers.39.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6618,key.395"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9229" name="__module.model.gpt_neox.layers.39.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.39.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9230" name="Constant_536031" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9231" name="Constant_536024" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9232" name="Constant_536026" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9233" name="Constant_536028" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9234" name="Concat_536029" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9235" name="Broadcast_536032" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9236" name="ReadValue_522437" type="ReadValue" version="opset6"> <data variable_id="past_key_values.39.valuepresent.39.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.39.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9237" name="Constant_509779" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9238" name="Gather_509780" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9239" name="Constant_438336" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9240" name="Constant_438339" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9241" name="Constant_438342" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9242" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6556"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9243" name="Constant_22827" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6557"> <dim>4</dim> </port> </output> </layer> <layer id="9244" name="__module.model.gpt_neox.layers.39.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6558,value.157"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9245" name="__module.model.gpt_neox.layers.39.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.39.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9246" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="6635"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9247" name="Constant_438346" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9248" name="Constant_438345" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9249" name="__module.model.gpt_neox.layers.39.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9250" name="__module.model.gpt_neox.layers.39.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9251" name="__module.model.gpt_neox.layers.39.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6630" /> </output> </layer> <layer id="9252" name="__module.model.gpt_neox.layers.39.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9253" name="__module.model.gpt_neox.layers.39.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9254" name="__module.model.gpt_neox.layers.39.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6628" /> </output> </layer> <layer id="9255" name="__module.model.gpt_neox.layers.39.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6632,6633" /> </output> </layer> <layer id="9256" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9257" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9258" name="Constant_438344" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9259" name="ScatterUpdate_438347" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9260" name="Constant_438348" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9261" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9262" name="ScatterUpdate_438349" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9263" name="Constant_438352" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9264" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6636"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9265" name="Constant_438424" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9266" name="Constant_438425" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9267" name="Constant_438421" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9268" name="Constant_438420" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9269" name="ScatterUpdate_438426" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9270" name="Constant_438429" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9271" name="__module.model.gpt_neox.layers.39.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6637,6638,causal_mask.235"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9272" name="__module.model.gpt_neox.layers.39.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9273" name="__module.model.gpt_neox.layers.39.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="6625,mask_value.79" /> </output> </layer> <layer id="9274" name="__module.model.gpt_neox.layers.39.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="6639,causal_mask.237"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9275" name="Add_456203" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9276" name="ShapeOf_456204" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9277" name="Constant_402860" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9278" name="__module.model.gpt_neox.layers.39.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9279" name="__module.model.gpt_neox.layers.39.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6623"> <dim>1</dim> </port> </output> </layer> <layer id="9280" name="__module.model.gpt_neox.layers.39.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9281" name="__module.model.gpt_neox.layers.39.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9282" name="__module.model.gpt_neox.layers.39.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9283" name="__module.model.gpt_neox.layers.39.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="6640"> <dim>4</dim> </port> </output> </layer> <layer id="9284" name="__module.model.gpt_neox.layers.39.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9285" name="__module.model.gpt_neox.layers.39.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="9286" name="__module.model.gpt_neox.layers.39.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9287" name="__module.model.gpt_neox.layers.39.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9288" name="Maximum_456205" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9289" name="__module.model.gpt_neox.layers.39.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6642,attention_mask.87"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9290" name="__module.model.gpt_neox.layers.39.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6643,6644,sdpa_result.79,tensor.157"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9291" name="Constant_23252" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6645"> <dim>4</dim> </port> </output> </layer> <layer id="9292" name="__module.model.gpt_neox.layers.39.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6646"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9293" name="__module.model.gpt_neox.layers.39.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9294" name="__module.model.gpt_neox.layers.39.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="6651,input.631"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9295" name="Constant_23270" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="18419212824" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9296" name="Convert_550313" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9297" name="Constant_23270/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="18456961560" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9298" name="Convert_550316" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9299" name="Subtract_550317" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9300" name="Constant_550318" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="18456967704" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9301" name="Constant_23270/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9302" name="__module.model.gpt_neox.layers.39.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9303" name="Constant_467185" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18305727000" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9304" name="__module.model.gpt_neox.layers.39.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6654,6656,input.633"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9305" name="__module.model.gpt_neox.layers.39/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6682"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9306" name="__module.model.gpt_neox.layers.39/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6683,6685,input.643"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9307" name="__module.model.gpt_neox.layers.40.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9308" name="__module.model.gpt_neox.layers.40.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9309" name="Constant_467186" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18456992280" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9310" name="__module.model.gpt_neox.layers.40.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9311" name="Constant_467187" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18457016856" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9312" name="__module.model.gpt_neox.layers.40.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6820,input.651"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9313" name="Constant_23870" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="18457041432" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9314" name="Convert_550358" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9315" name="Constant_23870/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="18608036376" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9316" name="Convert_550361" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9317" name="Subtract_550362" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9318" name="Constant_550363" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="18608060952" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9319" name="Constant_23870/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9320" name="__module.model.gpt_neox.layers.40.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9321" name="Constant_467188" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="18608159256" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9322" name="__module.model.gpt_neox.layers.40.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="6825,input.653"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9323" name="__module.model.gpt_neox.layers.40.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="6834,input.655"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9324" name="Constant_23890" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="18608257560" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9325" name="Convert_550367" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9326" name="Constant_23890/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="18759252504" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9327" name="Convert_550370" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9328" name="Subtract_550371" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9329" name="Constant_550372" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="18759258648" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9330" name="Constant_23890/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9331" name="__module.model.gpt_neox.layers.40.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9332" name="Constant_467189" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18759283224" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9333" name="__module.model.gpt_neox.layers.40.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6837,input.657"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9334" name="__module.model.gpt_neox.layers.40.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9335" name="__module.model.gpt_neox.layers.40.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9336" name="Constant_467190" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18759307800" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9337" name="__module.model.gpt_neox.layers.40.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9338" name="Constant_467191" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18759332376" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9339" name="__module.model.gpt_neox.layers.40.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6695,input.645"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9340" name="Constant_23349" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="18759356952" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9341" name="Convert_550340" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9342" name="Constant_23349/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="18872603160" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9343" name="Convert_550343" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9344" name="Subtract_550344" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9345" name="Constant_550345" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="18872621592" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9346" name="Constant_23349/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9347" name="__module.model.gpt_neox.layers.40.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9348" name="Constant_467192" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="18872695320" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9349" name="__module.model.gpt_neox.layers.40.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="6702,qkv.161"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9350" name="__module.model.gpt_neox.layers.40.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9351" name="__module.model.gpt_neox.layers.40.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6706,qkv.163"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="9352" name="Constant_438437" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9353" name="Constant_438440" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9354" name="Constant_438443" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9355" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6707"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9356" name="Constant_23377" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6708"> <dim>4</dim> </port> </output> </layer> <layer id="9357" name="__module.model.gpt_neox.layers.40.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6709,query.321"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9358" name="Constant_438449" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9359" name="Constant_438452" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9360" name="Constant_438455" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9361" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6716,q.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9362" name="Constant_23453" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="6728,cos_cached.81"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="9363" name="__module.model.gpt_neox.layers.40.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9364" name="Constant_438461" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9365" name="Constant_438464" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9366" name="Constant_438467" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9367" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6710"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9368" name="Constant_23391" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6711"> <dim>4</dim> </port> </output> </layer> <layer id="9369" name="__module.model.gpt_neox.layers.40.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6712,key.401"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9370" name="__module.model.gpt_neox.layers.40.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9371" name="__module.model.gpt_neox.layers.40.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9372" name="__module.model.gpt_neox.layers.40.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6720" /> </output> </layer> <layer id="9373" name="Constant_536044" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9374" name="Constant_536037" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9375" name="Constant_536039" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9376" name="Constant_536041" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9377" name="Concat_536042" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9378" name="Broadcast_536045" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9379" name="ReadValue_522443" type="ReadValue" version="opset6"> <data variable_id="past_key_values.40.keypresent.40.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.40.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9380" name="Constant_509782" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9381" name="Gather_509783" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9382" name="__module.model.gpt_neox.layers.40.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9383" name="__module.model.gpt_neox.layers.40.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9384" name="__module.model.gpt_neox.layers.40.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6722" /> </output> </layer> <layer id="9385" name="__module.model.gpt_neox.layers.40.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6721,6725,6726,seq_len.161" /> </output> </layer> <layer id="9386" name="__module.model.gpt_neox.layers.40.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9387" name="__module.model.gpt_neox.layers.40.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9388" name="__module.model.gpt_neox.layers.40.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9389" name="__module.model.gpt_neox.layers.40.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6729,6732,cos.161"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9390" name="Constant_52143" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9391" name="__module.model.gpt_neox.layers.40.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6735"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9392" name="__module.model.gpt_neox.layers.40.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6736,cos.163"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9393" name="__module.model.gpt_neox.layers.40.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6740"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9394" name="Constant_438472" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9395" name="Constant_438471" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9396" name="__module.model.gpt_neox.layers.40.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9397" name="__module.model.gpt_neox.layers.40.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9398" name="__module.model.gpt_neox.layers.40.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6741,6746" /> </output> </layer> <layer id="9399" name="__module.model.gpt_neox.layers.40.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="9400" name="__module.model.gpt_neox.layers.40.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6743,6744,6748,6749" /> </output> </layer> <layer id="9401" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9402" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9403" name="Constant_438470" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9404" name="ScatterUpdate_438473" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9405" name="Constant_438476" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9406" name="Constant_438479" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9407" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6750,x2.161"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9408" name="Constant_467193" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="9409" name="__module.model.gpt_neox.layers.40.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6751"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9410" name="Constant_438525" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9411" name="Constant_438530" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9412" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6745,x1.161"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9413" name="__module.model.gpt_neox.layers.40.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6753"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9414" name="Constant_23451" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="6727,sin_cached.81"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="9415" name="__module.model.gpt_neox.layers.40.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9416" name="__module.model.gpt_neox.layers.40.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9417" name="__module.model.gpt_neox.layers.40.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6730,6733,sin.161"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9418" name="Constant_52146" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9419" name="__module.model.gpt_neox.layers.40.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6738"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9420" name="__module.model.gpt_neox.layers.40.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6739,sin.163"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9421" name="__module.model.gpt_neox.layers.40.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6754"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9422" name="__module.model.gpt_neox.layers.40.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6755,query.323"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9423" name="Constant_438577" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9424" name="Constant_438580" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9425" name="Constant_438583" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9426" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6717,query_pass.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="9427" name="__module.model.gpt_neox.layers.40.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6773,6783,query.325,query.327"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9428" name="Constant_438589" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9429" name="Constant_438592" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9430" name="Constant_438595" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9431" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6718,k.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9432" name="__module.model.gpt_neox.layers.40.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6756"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9433" name="Constant_438599" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9434" name="Constant_438598" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9435" name="__module.model.gpt_neox.layers.40.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9436" name="__module.model.gpt_neox.layers.40.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9437" name="__module.model.gpt_neox.layers.40.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6757,6762" /> </output> </layer> <layer id="9438" name="__module.model.gpt_neox.layers.40.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="9439" name="__module.model.gpt_neox.layers.40.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6759,6760,6764,6765" /> </output> </layer> <layer id="9440" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9441" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9442" name="Constant_438597" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9443" name="ScatterUpdate_438600" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9444" name="Constant_438603" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9445" name="Constant_438606" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9446" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6766,x2.163"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9447" name="Constant_467194" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="9448" name="__module.model.gpt_neox.layers.40.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6767"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9449" name="Constant_438652" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9450" name="Constant_438657" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9451" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6761,x1.163"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9452" name="__module.model.gpt_neox.layers.40.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6769"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9453" name="__module.model.gpt_neox.layers.40.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6770"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9454" name="__module.model.gpt_neox.layers.40.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6771,key.403"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9455" name="Constant_438703" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9456" name="Constant_438706" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9457" name="Constant_438709" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9458" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6719,key_pass.81"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="9459" name="__module.model.gpt_neox.layers.40.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6775,key.405"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9460" name="__module.model.gpt_neox.layers.40.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.40.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9461" name="Constant_536057" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9462" name="Constant_536050" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9463" name="Constant_536052" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9464" name="Constant_536054" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9465" name="Concat_536055" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9466" name="Broadcast_536058" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9467" name="ReadValue_522445" type="ReadValue" version="opset6"> <data variable_id="past_key_values.40.valuepresent.40.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.40.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9468" name="Constant_509785" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9469" name="Gather_509786" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9470" name="Constant_438715" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9471" name="Constant_438718" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9472" name="Constant_438721" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9473" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6713"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9474" name="Constant_23405" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6714"> <dim>4</dim> </port> </output> </layer> <layer id="9475" name="__module.model.gpt_neox.layers.40.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6715,value.161"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9476" name="__module.model.gpt_neox.layers.40.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.40.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9477" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="6792"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9478" name="Constant_438725" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9479" name="Constant_438724" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9480" name="__module.model.gpt_neox.layers.40.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9481" name="__module.model.gpt_neox.layers.40.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9482" name="__module.model.gpt_neox.layers.40.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6787" /> </output> </layer> <layer id="9483" name="__module.model.gpt_neox.layers.40.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9484" name="__module.model.gpt_neox.layers.40.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9485" name="__module.model.gpt_neox.layers.40.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6785" /> </output> </layer> <layer id="9486" name="__module.model.gpt_neox.layers.40.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6789,6790" /> </output> </layer> <layer id="9487" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9488" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9489" name="Constant_438723" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9490" name="ScatterUpdate_438726" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9491" name="Constant_438727" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9492" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9493" name="ScatterUpdate_438728" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9494" name="Constant_438731" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9495" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6793"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9496" name="Constant_438803" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9497" name="Constant_438804" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9498" name="Constant_438800" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9499" name="Constant_438799" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9500" name="ScatterUpdate_438805" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9501" name="Constant_438808" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9502" name="__module.model.gpt_neox.layers.40.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6794,6795,causal_mask.241"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9503" name="__module.model.gpt_neox.layers.40.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9504" name="__module.model.gpt_neox.layers.40.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="6782,mask_value.81" /> </output> </layer> <layer id="9505" name="__module.model.gpt_neox.layers.40.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="6796,causal_mask.243"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9506" name="Add_456279" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9507" name="ShapeOf_456280" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9508" name="Constant_402875" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9509" name="__module.model.gpt_neox.layers.40.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9510" name="__module.model.gpt_neox.layers.40.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6780"> <dim>1</dim> </port> </output> </layer> <layer id="9511" name="__module.model.gpt_neox.layers.40.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9512" name="__module.model.gpt_neox.layers.40.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9513" name="__module.model.gpt_neox.layers.40.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9514" name="__module.model.gpt_neox.layers.40.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="6797"> <dim>4</dim> </port> </output> </layer> <layer id="9515" name="__module.model.gpt_neox.layers.40.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9516" name="__module.model.gpt_neox.layers.40.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="9517" name="__module.model.gpt_neox.layers.40.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9518" name="__module.model.gpt_neox.layers.40.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9519" name="Maximum_456281" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9520" name="__module.model.gpt_neox.layers.40.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6799,attention_mask.89"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9521" name="__module.model.gpt_neox.layers.40.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6800,6801,sdpa_result.81,tensor.161"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9522" name="Constant_23830" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6802"> <dim>4</dim> </port> </output> </layer> <layer id="9523" name="__module.model.gpt_neox.layers.40.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6803"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9524" name="__module.model.gpt_neox.layers.40.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9525" name="__module.model.gpt_neox.layers.40.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="6808,input.647"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9526" name="Constant_23848" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="18872769048" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9527" name="Convert_550349" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9528" name="Constant_23848/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="18910517784" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9529" name="Convert_550352" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9530" name="Subtract_550353" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9531" name="Constant_550354" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="18910523928" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9532" name="Constant_23848/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9533" name="__module.model.gpt_neox.layers.40.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9534" name="Constant_467195" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18759283224" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9535" name="__module.model.gpt_neox.layers.40.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6811,6813,input.649"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9536" name="__module.model.gpt_neox.layers.40/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6839"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9537" name="__module.model.gpt_neox.layers.40/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6840,6842,input.659"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9538" name="__module.model.gpt_neox.layers.41.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9539" name="__module.model.gpt_neox.layers.41.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9540" name="Constant_467196" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18910548504" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9541" name="__module.model.gpt_neox.layers.41.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9542" name="Constant_467197" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="18910573080" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9543" name="__module.model.gpt_neox.layers.41.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6977,input.667"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9544" name="Constant_24448" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="18910597656" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9545" name="Convert_550394" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9546" name="Constant_24448/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="19061592600" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9547" name="Convert_550397" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9548" name="Subtract_550398" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9549" name="Constant_550399" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="19061617176" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9550" name="Constant_24448/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9551" name="__module.model.gpt_neox.layers.41.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9552" name="Constant_467198" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="19061715480" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9553" name="__module.model.gpt_neox.layers.41.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="6982,input.669"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9554" name="__module.model.gpt_neox.layers.41.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="6991,input.671"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9555" name="Constant_24468" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="19061813784" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9556" name="Convert_550403" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9557" name="Constant_24468/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="19212808728" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9558" name="Convert_550406" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9559" name="Subtract_550407" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9560" name="Constant_550408" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="19212814872" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9561" name="Constant_24468/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9562" name="__module.model.gpt_neox.layers.41.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9563" name="Constant_467199" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19212839448" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9564" name="__module.model.gpt_neox.layers.41.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6994,input.673"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9565" name="__module.model.gpt_neox.layers.41.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9566" name="__module.model.gpt_neox.layers.41.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9567" name="Constant_467200" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19212864024" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9568" name="__module.model.gpt_neox.layers.41.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9569" name="Constant_467201" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19212888600" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9570" name="__module.model.gpt_neox.layers.41.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6852,input.661"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9571" name="Constant_23927" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="19212913176" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9572" name="Convert_550376" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9573" name="Constant_23927/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="19326159384" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9574" name="Convert_550379" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9575" name="Subtract_550380" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9576" name="Constant_550381" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="19326177816" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9577" name="Constant_23927/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9578" name="__module.model.gpt_neox.layers.41.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9579" name="Constant_467202" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="19326251544" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9580" name="__module.model.gpt_neox.layers.41.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="6859,qkv.165"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9581" name="__module.model.gpt_neox.layers.41.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9582" name="__module.model.gpt_neox.layers.41.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6863,qkv.167"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="9583" name="Constant_438816" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9584" name="Constant_438819" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9585" name="Constant_438822" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9586" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6864"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9587" name="Constant_23955" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6865"> <dim>4</dim> </port> </output> </layer> <layer id="9588" name="__module.model.gpt_neox.layers.41.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6866,query.329"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9589" name="Constant_438828" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9590" name="Constant_438831" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9591" name="Constant_438834" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9592" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6873,q.83"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9593" name="Constant_24031" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="6885,cos_cached.83"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="9594" name="__module.model.gpt_neox.layers.41.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9595" name="Constant_438840" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9596" name="Constant_438843" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9597" name="Constant_438846" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9598" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6867"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9599" name="Constant_23969" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6868"> <dim>4</dim> </port> </output> </layer> <layer id="9600" name="__module.model.gpt_neox.layers.41.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6869,key.411"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9601" name="__module.model.gpt_neox.layers.41.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9602" name="__module.model.gpt_neox.layers.41.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9603" name="__module.model.gpt_neox.layers.41.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6877" /> </output> </layer> <layer id="9604" name="Constant_536070" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9605" name="Constant_536063" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9606" name="Constant_536065" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9607" name="Constant_536067" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9608" name="Concat_536068" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9609" name="Broadcast_536071" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9610" name="ReadValue_522447" type="ReadValue" version="opset6"> <data variable_id="past_key_values.41.keypresent.41.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.41.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9611" name="Constant_509788" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9612" name="Gather_509789" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9613" name="__module.model.gpt_neox.layers.41.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9614" name="__module.model.gpt_neox.layers.41.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9615" name="__module.model.gpt_neox.layers.41.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6879" /> </output> </layer> <layer id="9616" name="__module.model.gpt_neox.layers.41.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6878,6882,6883,seq_len.165" /> </output> </layer> <layer id="9617" name="__module.model.gpt_neox.layers.41.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9618" name="__module.model.gpt_neox.layers.41.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9619" name="__module.model.gpt_neox.layers.41.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9620" name="__module.model.gpt_neox.layers.41.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6886,6889,cos.165"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9621" name="Constant_52149" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9622" name="__module.model.gpt_neox.layers.41.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6892"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9623" name="__module.model.gpt_neox.layers.41.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6893,cos.167"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9624" name="__module.model.gpt_neox.layers.41.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6897"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9625" name="Constant_438851" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9626" name="Constant_438850" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9627" name="__module.model.gpt_neox.layers.41.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9628" name="__module.model.gpt_neox.layers.41.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9629" name="__module.model.gpt_neox.layers.41.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6898,6903" /> </output> </layer> <layer id="9630" name="__module.model.gpt_neox.layers.41.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="9631" name="__module.model.gpt_neox.layers.41.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6900,6901,6905,6906" /> </output> </layer> <layer id="9632" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9633" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9634" name="Constant_438849" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9635" name="ScatterUpdate_438852" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9636" name="Constant_438855" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9637" name="Constant_438858" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9638" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6907,x2.165"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9639" name="Constant_467203" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="9640" name="__module.model.gpt_neox.layers.41.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6908"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9641" name="Constant_438904" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9642" name="Constant_438909" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9643" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6902,x1.165"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9644" name="__module.model.gpt_neox.layers.41.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6910"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9645" name="Constant_24029" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="6884,sin_cached.83"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="9646" name="__module.model.gpt_neox.layers.41.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9647" name="__module.model.gpt_neox.layers.41.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9648" name="__module.model.gpt_neox.layers.41.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6887,6890,sin.165"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9649" name="Constant_52152" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9650" name="__module.model.gpt_neox.layers.41.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="6895"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9651" name="__module.model.gpt_neox.layers.41.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="6896,sin.167"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9652" name="__module.model.gpt_neox.layers.41.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6911"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9653" name="__module.model.gpt_neox.layers.41.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6912,query.331"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9654" name="Constant_438956" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9655" name="Constant_438959" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9656" name="Constant_438962" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9657" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6874,query_pass.83"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="9658" name="__module.model.gpt_neox.layers.41.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6930,6940,query.333,query.335"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9659" name="Constant_438968" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9660" name="Constant_438971" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9661" name="Constant_438974" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9662" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6875,k.83"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9663" name="__module.model.gpt_neox.layers.41.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6913"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9664" name="Constant_438978" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9665" name="Constant_438977" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9666" name="__module.model.gpt_neox.layers.41.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9667" name="__module.model.gpt_neox.layers.41.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9668" name="__module.model.gpt_neox.layers.41.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6914,6919" /> </output> </layer> <layer id="9669" name="__module.model.gpt_neox.layers.41.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="9670" name="__module.model.gpt_neox.layers.41.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="6916,6917,6921,6922" /> </output> </layer> <layer id="9671" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9672" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9673" name="Constant_438976" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9674" name="ScatterUpdate_438979" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9675" name="Constant_438982" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9676" name="Constant_438985" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9677" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6923,x2.167"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9678" name="Constant_467204" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="9679" name="__module.model.gpt_neox.layers.41.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="6924"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9680" name="Constant_439031" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9681" name="Constant_439036" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9682" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6918,x1.167"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9683" name="__module.model.gpt_neox.layers.41.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="6926"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9684" name="__module.model.gpt_neox.layers.41.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6927"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9685" name="__module.model.gpt_neox.layers.41.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="6928,key.413"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9686" name="Constant_439082" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9687" name="Constant_439085" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9688" name="Constant_439088" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9689" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6876,key_pass.83"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="9690" name="__module.model.gpt_neox.layers.41.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="6932,key.415"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9691" name="__module.model.gpt_neox.layers.41.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.41.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9692" name="Constant_536083" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9693" name="Constant_536076" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9694" name="Constant_536078" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9695" name="Constant_536080" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9696" name="Concat_536081" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9697" name="Broadcast_536084" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9698" name="ReadValue_522449" type="ReadValue" version="opset6"> <data variable_id="past_key_values.41.valuepresent.41.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.41.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9699" name="Constant_509791" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9700" name="Gather_509792" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9701" name="Constant_439094" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9702" name="Constant_439097" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9703" name="Constant_439100" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9704" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="6870"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9705" name="Constant_23983" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6871"> <dim>4</dim> </port> </output> </layer> <layer id="9706" name="__module.model.gpt_neox.layers.41.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6872,value.165"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9707" name="__module.model.gpt_neox.layers.41.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.41.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9708" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="6949"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9709" name="Constant_439104" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9710" name="Constant_439103" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9711" name="__module.model.gpt_neox.layers.41.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9712" name="__module.model.gpt_neox.layers.41.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9713" name="__module.model.gpt_neox.layers.41.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6944" /> </output> </layer> <layer id="9714" name="__module.model.gpt_neox.layers.41.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9715" name="__module.model.gpt_neox.layers.41.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9716" name="__module.model.gpt_neox.layers.41.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6942" /> </output> </layer> <layer id="9717" name="__module.model.gpt_neox.layers.41.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="6946,6947" /> </output> </layer> <layer id="9718" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9719" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9720" name="Constant_439102" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9721" name="ScatterUpdate_439105" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9722" name="Constant_439106" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9723" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9724" name="ScatterUpdate_439107" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9725" name="Constant_439110" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9726" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6950"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9727" name="Constant_439182" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9728" name="Constant_439183" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9729" name="Constant_439179" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9730" name="Constant_439178" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9731" name="ScatterUpdate_439184" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9732" name="Constant_439187" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9733" name="__module.model.gpt_neox.layers.41.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="6951,6952,causal_mask.247"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9734" name="__module.model.gpt_neox.layers.41.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9735" name="__module.model.gpt_neox.layers.41.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="6939,mask_value.83" /> </output> </layer> <layer id="9736" name="__module.model.gpt_neox.layers.41.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="6953,causal_mask.249"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9737" name="Add_456355" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9738" name="ShapeOf_456356" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9739" name="Constant_402890" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9740" name="__module.model.gpt_neox.layers.41.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9741" name="__module.model.gpt_neox.layers.41.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="6937"> <dim>1</dim> </port> </output> </layer> <layer id="9742" name="__module.model.gpt_neox.layers.41.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9743" name="__module.model.gpt_neox.layers.41.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9744" name="__module.model.gpt_neox.layers.41.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9745" name="__module.model.gpt_neox.layers.41.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="6954"> <dim>4</dim> </port> </output> </layer> <layer id="9746" name="__module.model.gpt_neox.layers.41.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9747" name="__module.model.gpt_neox.layers.41.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="9748" name="__module.model.gpt_neox.layers.41.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9749" name="__module.model.gpt_neox.layers.41.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9750" name="Maximum_456357" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9751" name="__module.model.gpt_neox.layers.41.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6956,attention_mask.91"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9752" name="__module.model.gpt_neox.layers.41.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="6957,6958,sdpa_result.83,tensor.165"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9753" name="Constant_24408" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="6959"> <dim>4</dim> </port> </output> </layer> <layer id="9754" name="__module.model.gpt_neox.layers.41.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="6960"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9755" name="__module.model.gpt_neox.layers.41.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9756" name="__module.model.gpt_neox.layers.41.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="6965,input.663"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9757" name="Constant_24426" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="19326325272" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9758" name="Convert_550385" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9759" name="Constant_24426/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="19364074008" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9760" name="Convert_550388" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9761" name="Subtract_550389" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9762" name="Constant_550390" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="19364080152" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9763" name="Constant_24426/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9764" name="__module.model.gpt_neox.layers.41.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9765" name="Constant_467205" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19212839448" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9766" name="__module.model.gpt_neox.layers.41.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6968,6970,input.665"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9767" name="__module.model.gpt_neox.layers.41/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6996"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9768" name="__module.model.gpt_neox.layers.41/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="6997,6999,input.675"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9769" name="__module.model.gpt_neox.layers.42.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9770" name="__module.model.gpt_neox.layers.42.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9771" name="Constant_467206" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19364104728" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9772" name="__module.model.gpt_neox.layers.42.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9773" name="Constant_467207" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19364129304" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9774" name="__module.model.gpt_neox.layers.42.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7134,input.683"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9775" name="Constant_25026" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="19364153880" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9776" name="Convert_550430" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9777" name="Constant_25026/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="19515148824" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9778" name="Convert_550433" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9779" name="Subtract_550434" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9780" name="Constant_550435" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="19515173400" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="9781" name="Constant_25026/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9782" name="__module.model.gpt_neox.layers.42.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9783" name="Constant_467208" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="19515271704" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9784" name="__module.model.gpt_neox.layers.42.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="7139,input.685"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9785" name="__module.model.gpt_neox.layers.42.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="7148,input.687"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9786" name="Constant_25046" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="19515370008" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9787" name="Convert_550439" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9788" name="Constant_25046/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="19666364952" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9789" name="Convert_550442" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9790" name="Subtract_550443" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9791" name="Constant_550444" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="19666371096" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9792" name="Constant_25046/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="9793" name="__module.model.gpt_neox.layers.42.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9794" name="Constant_467209" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19666395672" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9795" name="__module.model.gpt_neox.layers.42.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7151,input.689"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9796" name="__module.model.gpt_neox.layers.42.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9797" name="__module.model.gpt_neox.layers.42.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9798" name="Constant_467210" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19666420248" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9799" name="__module.model.gpt_neox.layers.42.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9800" name="Constant_467211" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19666444824" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9801" name="__module.model.gpt_neox.layers.42.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7009,input.677"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9802" name="Constant_24505" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="19666469400" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9803" name="Convert_550412" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9804" name="Constant_24505/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="19779715608" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9805" name="Convert_550415" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9806" name="Subtract_550416" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9807" name="Constant_550417" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="19779734040" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="9808" name="Constant_24505/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9809" name="__module.model.gpt_neox.layers.42.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9810" name="Constant_467212" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="19779807768" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9811" name="__module.model.gpt_neox.layers.42.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="7016,qkv.169"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="9812" name="__module.model.gpt_neox.layers.42.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9813" name="__module.model.gpt_neox.layers.42.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7020,qkv.171"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="9814" name="Constant_439195" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9815" name="Constant_439198" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9816" name="Constant_439201" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9817" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7021"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9818" name="Constant_24533" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="7022"> <dim>4</dim> </port> </output> </layer> <layer id="9819" name="__module.model.gpt_neox.layers.42.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7023,query.337"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9820" name="Constant_439207" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9821" name="Constant_439210" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9822" name="Constant_439213" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9823" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7030,q.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9824" name="Constant_24609" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="7042,cos_cached.85"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="9825" name="__module.model.gpt_neox.layers.42.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9826" name="Constant_439219" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9827" name="Constant_439222" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9828" name="Constant_439225" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9829" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7024"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9830" name="Constant_24547" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="7025"> <dim>4</dim> </port> </output> </layer> <layer id="9831" name="__module.model.gpt_neox.layers.42.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7026,key.421"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9832" name="__module.model.gpt_neox.layers.42.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9833" name="__module.model.gpt_neox.layers.42.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9834" name="__module.model.gpt_neox.layers.42.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7034" /> </output> </layer> <layer id="9835" name="Constant_536096" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9836" name="Constant_536089" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9837" name="Constant_536091" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9838" name="Constant_536093" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9839" name="Concat_536094" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9840" name="Broadcast_536097" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9841" name="ReadValue_522451" type="ReadValue" version="opset6"> <data variable_id="past_key_values.42.keypresent.42.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.42.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9842" name="Constant_509794" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9843" name="Gather_509795" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9844" name="__module.model.gpt_neox.layers.42.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9845" name="__module.model.gpt_neox.layers.42.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9846" name="__module.model.gpt_neox.layers.42.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7036" /> </output> </layer> <layer id="9847" name="__module.model.gpt_neox.layers.42.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="7035,7039,7040,seq_len.169" /> </output> </layer> <layer id="9848" name="__module.model.gpt_neox.layers.42.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9849" name="__module.model.gpt_neox.layers.42.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9850" name="__module.model.gpt_neox.layers.42.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9851" name="__module.model.gpt_neox.layers.42.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="7043,7046,cos.169"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9852" name="Constant_52155" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9853" name="__module.model.gpt_neox.layers.42.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="7049"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9854" name="__module.model.gpt_neox.layers.42.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="7050,cos.171"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9855" name="__module.model.gpt_neox.layers.42.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7054"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9856" name="Constant_439230" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9857" name="Constant_439229" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9858" name="__module.model.gpt_neox.layers.42.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9859" name="__module.model.gpt_neox.layers.42.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9860" name="__module.model.gpt_neox.layers.42.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7055,7060" /> </output> </layer> <layer id="9861" name="__module.model.gpt_neox.layers.42.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="9862" name="__module.model.gpt_neox.layers.42.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="7057,7058,7062,7063" /> </output> </layer> <layer id="9863" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9864" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9865" name="Constant_439228" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9866" name="ScatterUpdate_439231" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9867" name="Constant_439234" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9868" name="Constant_439237" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9869" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7064,x2.169"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9870" name="Constant_467213" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="9871" name="__module.model.gpt_neox.layers.42.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="7065"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9872" name="Constant_439283" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9873" name="Constant_439288" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9874" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7059,x1.169"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9875" name="__module.model.gpt_neox.layers.42.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="7067"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9876" name="Constant_24607" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="7041,sin_cached.85"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="9877" name="__module.model.gpt_neox.layers.42.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9878" name="__module.model.gpt_neox.layers.42.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9879" name="__module.model.gpt_neox.layers.42.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="7044,7047,sin.169"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9880" name="Constant_52158" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9881" name="__module.model.gpt_neox.layers.42.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="7052"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9882" name="__module.model.gpt_neox.layers.42.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="7053,sin.171"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9883" name="__module.model.gpt_neox.layers.42.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7068"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9884" name="__module.model.gpt_neox.layers.42.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7069,query.339"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9885" name="Constant_439335" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9886" name="Constant_439338" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9887" name="Constant_439341" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9888" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7031,query_pass.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="9889" name="__module.model.gpt_neox.layers.42.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="7087,7097,query.341,query.343"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9890" name="Constant_439347" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9891" name="Constant_439350" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9892" name="Constant_439353" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9893" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7032,k.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9894" name="__module.model.gpt_neox.layers.42.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7070"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9895" name="Constant_439357" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9896" name="Constant_439356" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9897" name="__module.model.gpt_neox.layers.42.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9898" name="__module.model.gpt_neox.layers.42.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9899" name="__module.model.gpt_neox.layers.42.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7071,7076" /> </output> </layer> <layer id="9900" name="__module.model.gpt_neox.layers.42.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="9901" name="__module.model.gpt_neox.layers.42.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="7073,7074,7078,7079" /> </output> </layer> <layer id="9902" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9903" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9904" name="Constant_439355" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9905" name="ScatterUpdate_439358" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9906" name="Constant_439361" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9907" name="Constant_439364" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9908" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7080,x2.171"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9909" name="Constant_467214" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="9910" name="__module.model.gpt_neox.layers.42.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="7081"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9911" name="Constant_439410" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9912" name="Constant_439415" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9913" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7075,x1.171"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="9914" name="__module.model.gpt_neox.layers.42.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="7083"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9915" name="__module.model.gpt_neox.layers.42.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7084"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9916" name="__module.model.gpt_neox.layers.42.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7085,key.423"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="9917" name="Constant_439461" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9918" name="Constant_439464" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9919" name="Constant_439467" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9920" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7033,key_pass.85"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="9921" name="__module.model.gpt_neox.layers.42.attention/aten::cat/Concat_3" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="7089,key.425"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9922" name="__module.model.gpt_neox.layers.42.attention/aten::to/Convert_1" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.42.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9923" name="Constant_536109" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9924" name="Constant_536102" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9925" name="Constant_536104" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9926" name="Constant_536106" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9927" name="Concat_536107" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9928" name="Broadcast_536110" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="9929" name="ReadValue_522453" type="ReadValue" version="opset6"> <data variable_id="past_key_values.42.valuepresent.42.value" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.42.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9930" name="Constant_509797" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="9931" name="Gather_509798" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9932" name="Constant_439473" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9933" name="Constant_439476" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="9934" name="Constant_439479" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9935" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7027"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9936" name="Constant_24561" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="7028"> <dim>4</dim> </port> </output> </layer> <layer id="9937" name="__module.model.gpt_neox.layers.42.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7029,value.169"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9938" name="__module.model.gpt_neox.layers.42.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.42.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9939" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="7106"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9940" name="Constant_439483" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9941" name="Constant_439482" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9942" name="__module.model.gpt_neox.layers.42.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9943" name="__module.model.gpt_neox.layers.42.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9944" name="__module.model.gpt_neox.layers.42.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7101" /> </output> </layer> <layer id="9945" name="__module.model.gpt_neox.layers.42.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9946" name="__module.model.gpt_neox.layers.42.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9947" name="__module.model.gpt_neox.layers.42.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7099" /> </output> </layer> <layer id="9948" name="__module.model.gpt_neox.layers.42.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="7103,7104" /> </output> </layer> <layer id="9949" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9950" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9951" name="Constant_439481" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9952" name="ScatterUpdate_439484" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9953" name="Constant_439485" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9954" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9955" name="ScatterUpdate_439486" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9956" name="Constant_439489" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9957" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="7107"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="9958" name="Constant_439561" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9959" name="Constant_439562" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9960" name="Constant_439558" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="9961" name="Constant_439557" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9962" name="ScatterUpdate_439563" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9963" name="Constant_439566" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9964" name="__module.model.gpt_neox.layers.42.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="7108,7109,causal_mask.253"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9965" name="__module.model.gpt_neox.layers.42.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="9966" name="__module.model.gpt_neox.layers.42.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="7096,mask_value.85" /> </output> </layer> <layer id="9967" name="__module.model.gpt_neox.layers.42.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="7110,causal_mask.255"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9968" name="Add_456431" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9969" name="ShapeOf_456432" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9970" name="Constant_402905" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9971" name="__module.model.gpt_neox.layers.42.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="9972" name="__module.model.gpt_neox.layers.42.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7094"> <dim>1</dim> </port> </output> </layer> <layer id="9973" name="__module.model.gpt_neox.layers.42.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9974" name="__module.model.gpt_neox.layers.42.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9975" name="__module.model.gpt_neox.layers.42.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="9976" name="__module.model.gpt_neox.layers.42.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="7111"> <dim>4</dim> </port> </output> </layer> <layer id="9977" name="__module.model.gpt_neox.layers.42.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9978" name="__module.model.gpt_neox.layers.42.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="9979" name="__module.model.gpt_neox.layers.42.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9980" name="__module.model.gpt_neox.layers.42.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9981" name="Maximum_456433" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="9982" name="__module.model.gpt_neox.layers.42.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7113,attention_mask.93"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="9983" name="__module.model.gpt_neox.layers.42.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="7114,7115,sdpa_result.85,tensor.169"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="9984" name="Constant_24986" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="7116"> <dim>4</dim> </port> </output> </layer> <layer id="9985" name="__module.model.gpt_neox.layers.42.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7117"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="9986" name="__module.model.gpt_neox.layers.42.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="9987" name="__module.model.gpt_neox.layers.42.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="7122,input.679"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9988" name="Constant_25004" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="19779881496" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9989" name="Convert_550421" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9990" name="Constant_25004/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="19817630232" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9991" name="Convert_550424" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9992" name="Subtract_550425" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9993" name="Constant_550426" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="19817636376" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="9994" name="Constant_25004/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9995" name="__module.model.gpt_neox.layers.42.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9996" name="Constant_467215" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19666395672" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9997" name="__module.model.gpt_neox.layers.42.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7125,7127,input.681"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9998" name="__module.model.gpt_neox.layers.42/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7153"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="9999" name="__module.model.gpt_neox.layers.42/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7154,7156,input.691"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10000" name="__module.model.gpt_neox.layers.43.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10001" name="__module.model.gpt_neox.layers.43.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10002" name="Constant_467216" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19817660952" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10003" name="__module.model.gpt_neox.layers.43.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10004" name="Constant_467217" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19817685528" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10005" name="__module.model.gpt_neox.layers.43.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7166,input.693"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10006" name="Constant_25083" type="Const" version="opset1"> <data element_type="u8" shape="18432, 6144" offset="19817710104" size="113246208" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10007" name="Convert_550448" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10008" name="Constant_25083/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="18432, 1" offset="19930956312" size="18432" /> <output> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="10009" name="Convert_550451" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="10010" name="Subtract_550452" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10011" name="Constant_550453" type="Const" version="opset1"> <data element_type="f32" shape="18432, 1" offset="19930974744" size="73728" /> <output> <port id="0" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </output> </layer> <layer id="10012" name="Constant_25083/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10013" name="__module.model.gpt_neox.layers.43.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>18432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="10014" name="Constant_467218" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 18432" offset="19931048472" size="73728" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="10015" name="__module.model.gpt_neox.layers.43.attention.query_key_value/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>18432</dim> </port> </input> <output> <port id="2" precision="FP32" names="7173,qkv.173"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> </output> </layer> <layer id="10016" name="__module.model.gpt_neox.layers.43.attention/prim::ListConstruct/Concat" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883168" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10017" name="__module.model.gpt_neox.layers.43.attention/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>18432</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7177,qkv"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> </output> </layer> <layer id="10018" name="Constant_439574" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10019" name="Constant_439577" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="10020" name="Constant_439580" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10021" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7184"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="10022" name="Constant_25139" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="7185"> <dim>4</dim> </port> </output> </layer> <layer id="10023" name="__module.model.gpt_neox.layers.43.attention/aten::permute/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7186,value.173"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10024" name="__module.model.gpt_neox.layers.43.attention/aten::cat/Concat" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.43.value"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10026" name="Constant_536122" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="10027" name="Constant_536115" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="12" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="10028" name="Constant_536117" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="4" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="10029" name="Constant_536119" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="20" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="10030" name="Concat_536120" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="10031" name="Broadcast_536123" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32" /> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </output> </layer> <layer id="10032" name="ReadValue_522455" type="ReadValue" version="opset6"> <data variable_id="past_key_values.43.keypresent.43.key" variable_type="f32" variable_shape="?,64,?,96" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>0</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32" names="past_key_values.43.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10033" name="Constant_509800" type="Const" version="opset1"> <data element_type="i64" shape="" offset="4" size="8" /> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="10034" name="Gather_509801" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10035" name="Constant_439586" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10036" name="Constant_439589" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="726079872" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10037" name="Constant_439592" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10038" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7181"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="10039" name="Constant_25125" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="7182"> <dim>4</dim> </port> </output> </layer> <layer id="10040" name="__module.model.gpt_neox.layers.43.attention/aten::permute/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7183,key.431"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10041" name="Constant_439598" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10042" name="Constant_439601" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10043" name="Constant_439604" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10044" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7189,k"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10045" name="Constant_25187" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="725883264" size="196608" /> <output> <port id="0" precision="FP32" names="7199,cos_cached"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="10046" name="__module.model.gpt_neox.layers.43.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10047" name="__module.model.gpt_neox.layers.43.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10048" name="__module.model.gpt_neox.layers.43.attention/aten::size/Constant_2" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10049" name="__module.model.gpt_neox.layers.43.attention/aten::size/Gather_2" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7191" /> </output> </layer> <layer id="10050" name="__module.model.gpt_neox.layers.43.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10051" name="__module.model.gpt_neox.layers.43.attention/aten::size/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10052" name="__module.model.gpt_neox.layers.43.attention/aten::size/Gather_3" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7193" /> </output> </layer> <layer id="10053" name="__module.model.gpt_neox.layers.43.attention/aten::add_/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="7192,7196,7197,seq_len.173" /> </output> </layer> <layer id="10054" name="__module.model.gpt_neox.layers.43.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10055" name="__module.model.gpt_neox.layers.43.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10056" name="__module.model.gpt_neox.layers.43.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10057" name="__module.model.gpt_neox.layers.43.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="7200,7203,cos.173"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10058" name="Constant_52161" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10059" name="__module.model.gpt_neox.layers.43.attention/aten::index/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="7206"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10060" name="__module.model.gpt_neox.layers.43.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="7207,cos"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10061" name="__module.model.gpt_neox.layers.43.attention/aten::mul/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7227"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10062" name="Constant_439609" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10063" name="Constant_439608" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="10064" name="__module.model.gpt_neox.layers.43.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10065" name="__module.model.gpt_neox.layers.43.attention/aten::size/Constant_5" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10066" name="__module.model.gpt_neox.layers.43.attention/aten::size/Gather_5" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7228,7233" /> </output> </layer> <layer id="10067" name="__module.model.gpt_neox.layers.43.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="10068" name="__module.model.gpt_neox.layers.43.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="7230,7231,7235,7236" /> </output> </layer> <layer id="10069" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10070" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10071" name="Constant_439607" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10072" name="ScatterUpdate_439610" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10073" name="Constant_439613" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="10074" name="Constant_439616" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10075" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7237,x2"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="10076" name="Constant_467219" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="10077" name="__module.model.gpt_neox.layers.43.attention/aten::neg/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="7238"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="10078" name="Constant_439662" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10079" name="Constant_439667" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10080" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7232,x1"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="10081" name="__module.model.gpt_neox.layers.43.attention/aten::cat/Concat_1" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="7240"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10082" name="Constant_25185" type="Const" version="opset1"> <data element_type="f32" shape="2048, 24" offset="726079944" size="196608" /> <output> <port id="0" precision="FP32" names="7198,sin_cached"> <dim>2048</dim> <dim>24</dim> </port> </output> </layer> <layer id="10083" name="__module.model.gpt_neox.layers.43.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10084" name="__module.model.gpt_neox.layers.43.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="726079892" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10085" name="__module.model.gpt_neox.layers.43.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1"> <data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>2048</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="FP32" names="7201,7204,sin.173"> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10086" name="Constant_52164" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10087" name="__module.model.gpt_neox.layers.43.attention/aten::index/Gather_1" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="7209"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10088" name="__module.model.gpt_neox.layers.43.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="FP32" names="7210,sin"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10089" name="__module.model.gpt_neox.layers.43.attention/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7241"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10090" name="__module.model.gpt_neox.layers.43.attention/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7242,key.433"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10091" name="Constant_439714" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10092" name="Constant_439717" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="10093" name="Constant_439720" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10094" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7190,key_pass"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="10095" name="__module.model.gpt_neox.layers.43.attention/aten::cat/Concat_2" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="7246,key.435"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10096" name="__module.model.gpt_neox.layers.43.attention/aten::to/Convert" type="Concat" version="opset1"> <data axis="-2" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="2" precision="FP32" names="present.43.key"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10174" name="__module.model.gpt_neox.layers.43.post_attention_layernorm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10175" name="__module.model.gpt_neox.layers.43.post_attention_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10176" name="Constant_467220" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19931122200" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10177" name="__module.model.gpt_neox.layers.43.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10178" name="Constant_467221" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="19931146776" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10179" name="__module.model.gpt_neox.layers.43.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7291,input.699"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10180" name="Constant_25604" type="Const" version="opset1"> <data element_type="u8" shape="24576, 6144" offset="19931171352" size="150994944" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10181" name="Convert_550466" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10182" name="Constant_25604/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="24576, 1" offset="20082166296" size="24576" /> <output> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="10183" name="Convert_550469" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="10184" name="Subtract_550470" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10185" name="Constant_550471" type="Const" version="opset1"> <data element_type="f32" shape="24576, 1" offset="20082190872" size="98304" /> <output> <port id="0" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </output> </layer> <layer id="10186" name="Constant_25604/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10187" name="__module.model.gpt_neox.layers.43.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>24576</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="10188" name="Constant_467222" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 24576" offset="20082289176" size="98304" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="10189" name="__module.model.gpt_neox.layers.43.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32" names="7296,input.701"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="10190" name="__module.model.gpt_neox.layers.43.mlp.act/aten::mul/Multiply_5" type="Gelu" version="opset7"> <data approximation_mode="TANH" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32" names="7305,input.703"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> </output> </layer> <layer id="10191" name="Constant_25624" type="Const" version="opset1"> <data element_type="u8" shape="6144, 24576" offset="20082387480" size="150994944" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="10192" name="Convert_550475" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="10193" name="Constant_25624/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="20233382424" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="10194" name="Convert_550478" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="10195" name="Subtract_550479" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="10196" name="Constant_550480" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="20233388568" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="10197" name="Constant_25624/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </output> </layer> <layer id="10198" name="__module.model.gpt_neox.layers.43.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>24576</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>24576</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10199" name="Constant_467223" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="20233413144" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10200" name="__module.model.gpt_neox.layers.43.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7308,input.705"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10201" name="Constant_439726" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10202" name="Constant_439729" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883200" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10203" name="Constant_439732" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10204" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>288</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7178"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="10205" name="Constant_25111" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="7179"> <dim>4</dim> </port> </output> </layer> <layer id="10206" name="__module.model.gpt_neox.layers.43.attention/aten::permute/Transpose_2" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7180,query.345"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10207" name="Constant_439738" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10208" name="Constant_439741" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10209" name="Constant_439744" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10210" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7187,q"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10211" name="__module.model.gpt_neox.layers.43.attention/aten::mul/Multiply_2" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7211"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10212" name="Constant_439748" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10213" name="Constant_439747" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="10214" name="__module.model.gpt_neox.layers.43.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10215" name="__module.model.gpt_neox.layers.43.attention/aten::size/Constant_7" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10216" name="__module.model.gpt_neox.layers.43.attention/aten::size/Gather_7" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7212,7217" /> </output> </layer> <layer id="10217" name="__module.model.gpt_neox.layers.43.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1"> <data auto_broadcast="numpy" m_pythondiv="true" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" /> </output> </layer> <layer id="10218" name="__module.model.gpt_neox.layers.43.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1"> <input> <port id="0" precision="I32" /> </input> <output> <port id="1" precision="I32" names="7214,7215,7219,7220" /> </output> </layer> <layer id="10219" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10220" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Unsqueeze_37" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10221" name="Constant_439746" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10222" name="ScatterUpdate_439749" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10223" name="Constant_439752" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="10224" name="Constant_439755" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10225" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7221,x2.173"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="10226" name="Constant_467224" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 1, 1" offset="726079940" size="4" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="10227" name="__module.model.gpt_neox.layers.43.attention/aten::neg/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32" names="7222"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="10228" name="Constant_439801" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10229" name="Constant_439806" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10230" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7216,x1.173"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </output> </layer> <layer id="10231" name="__module.model.gpt_neox.layers.43.attention/aten::cat/Concat_4" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>12</dim> </port> </input> <output> <port id="2" precision="FP32" names="7224"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10232" name="__module.model.gpt_neox.layers.43.attention/aten::mul/Multiply_3" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7225"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10233" name="__module.model.gpt_neox.layers.43.attention/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </input> <output> <port id="2" precision="FP32" names="7226,query.347"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> </output> </layer> <layer id="10234" name="Constant_439852" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883248" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10235" name="Constant_439855" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="726079908" size="32" /> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="10236" name="Constant_439858" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10237" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I64"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="FP32" names="7188,query_pass"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </output> </layer> <layer id="10238" name="__module.model.gpt_neox.layers.43.attention/aten::cat/Concat_5" type="Concat" version="opset1"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>24</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>72</dim> </port> </input> <output> <port id="2" precision="FP32" names="7244,7254,query,query.349"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10239" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_11" type="Const" version="opset1"> <data element_type="boolean" shape="1, 1, 2048, 2048" offset="726276552" size="4194304" /> <output> <port id="0" precision="BOOL" names="7263"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> </output> </layer> <layer id="10240" name="Constant_439862" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="10241" name="Constant_439861" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="730470868" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="10242" name="__module.model.gpt_neox.layers.43.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10243" name="__module.model.gpt_neox.layers.43.attention/aten::size/Constant_8" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10244" name="__module.model.gpt_neox.layers.43.attention/aten::size/Gather_8" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7258" /> </output> </layer> <layer id="10245" name="__module.model.gpt_neox.layers.43.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10246" name="__module.model.gpt_neox.layers.43.attention/aten::size/Constant_9" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10247" name="__module.model.gpt_neox.layers.43.attention/aten::size/Gather_9" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7256" /> </output> </layer> <layer id="10248" name="__module.model.gpt_neox.layers.43.attention/aten::sub/Subtract" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32" names="7260,7261" /> </output> </layer> <layer id="10249" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10250" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10251" name="Constant_439860" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10252" name="ScatterUpdate_439863" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="10253" name="Constant_439864" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470856" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="10254" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="I32" /> <port id="1" precision="I32" /> </input> <output> <port id="2" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10255" name="ScatterUpdate_439865" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="10256" name="Constant_439868" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470880" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="10257" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>2048</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> <port id="2" precision="I32"> <dim>3</dim> </port> <port id="3" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="4" precision="BOOL" names="7264"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> </output> </layer> <layer id="10258" name="Constant_439940" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10259" name="Constant_439941" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883184" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10260" name="Constant_439937" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="726079896" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="10261" name="Constant_439936" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10262" name="ScatterUpdate_439942" type="ScatterUpdate" version="opset3"> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10263" name="Constant_439945" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10264" name="__module.model.gpt_neox.layers.43.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1"> <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>2048</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> <port id="3" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="4" precision="BOOL" names="7265,7266,causal_mask.259"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="10265" name="__module.model.gpt_neox.layers.43.attention/aten::where/Convert_1" type="Const" version="opset1"> <data element_type="f32" shape="" offset="0" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="10266" name="__module.model.gpt_neox.layers.43.attention/aten::full/Convert" type="Const" version="opset1"> <data element_type="f32" shape="" offset="730470892" size="4" /> <output> <port id="0" precision="FP32" names="7253,mask_value" /> </output> </layer> <layer id="10267" name="__module.model.gpt_neox.layers.43.attention/aten::where/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32" /> </input> <output> <port id="3" precision="FP32" names="7267,causal_mask.261"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="10268" name="Add_456507" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="10269" name="ShapeOf_456508" type="ShapeOf" version="opset3"> <data output_type="i32" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10270" name="Constant_402920" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="0" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10271" name="__module.model.gpt_neox.layers.43.attention/aten::size/Constant_10" type="Const" version="opset1"> <data element_type="i32" shape="" offset="0" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10272" name="__module.model.gpt_neox.layers.43.attention/aten::size/Gather_10" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I32" names="7251"> <dim>1</dim> </port> </output> </layer> <layer id="10273" name="__module.model.gpt_neox.layers.43.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10274" name="__module.model.gpt_neox.layers.43.attention/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10275" name="__module.model.gpt_neox.layers.43.attention/prim::ListConstruct/Reshape_2_1" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10276" name="__module.model.gpt_neox.layers.43.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I32"> <dim>1</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> <port id="2" precision="I32"> <dim>1</dim> </port> <port id="3" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I32" names="7268"> <dim>4</dim> </port> </output> </layer> <layer id="10277" name="__module.model.gpt_neox.layers.43.attention/aten::expand/Broadcast" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="730470908" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10278" name="__module.model.gpt_neox.layers.43.attention/aten::expand/Equal" type="Equal" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="BOOL"> <dim>4</dim> </port> </output> </layer> <layer id="10279" name="__module.model.gpt_neox.layers.43.attention/aten::expand/Broadcast_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883216" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10280" name="__module.model.gpt_neox.layers.43.attention/aten::expand/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> <port id="2" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="3" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10281" name="Maximum_456509" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I32"> <dim>4</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="10282" name="__module.model.gpt_neox.layers.43.attention/aten::add/Add_2" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7270,attention_mask"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="10283" name="__module.model.gpt_neox.layers.43.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> <data causal="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="2" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="3" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="4" precision="FP32" names="7271,7272,sdpa_result,tensor.173"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10284" name="Constant_25564" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="725883232" size="16" /> <output> <port id="0" precision="I32" names="7273"> <dim>4</dim> </port> </output> </layer> <layer id="10285" name="__module.model.gpt_neox.layers.43.attention/aten::permute/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="7274"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> </output> </layer> <layer id="10286" name="__module.model.gpt_neox.layers.43.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1"> <data element_type="i32" shape="3" offset="730470924" size="12" /> <output> <port id="0" precision="I32"> <dim>3</dim> </port> </output> </layer> <layer id="10287" name="__module.model.gpt_neox.layers.43.attention/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> <dim>96</dim> </port> <port id="1" precision="I32"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="7279,input.695"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10288" name="Constant_25582" type="Const" version="opset1"> <data element_type="u8" shape="6144, 6144" offset="20233437720" size="37748736" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10289" name="Convert_550457" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10290" name="Constant_25582/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="6144, 1" offset="20271186456" size="6144" /> <output> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="10291" name="Convert_550460" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="10292" name="Subtract_550461" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10293" name="Constant_550462" type="Const" version="opset1"> <data element_type="f32" shape="6144, 1" offset="20271192600" size="24576" /> <output> <port id="0" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </output> </layer> <layer id="10294" name="Constant_25582/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10295" name="__module.model.gpt_neox.layers.43.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>6144</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10296" name="Constant_467225" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="20233413144" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10297" name="__module.model.gpt_neox.layers.43.attention.dense/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7282,7284,input.697"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10298" name="__module.model.gpt_neox.layers.43/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7310"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10299" name="__module.model.gpt_neox.layers.43/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7311,7313,input.707"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10300" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Constant" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="310106396" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="10301" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10302" name="Constant_467226" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="20271217176" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10303" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Multiply" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10304" name="Constant_467227" type="Const" version="opset1"> <data element_type="f32" shape="1, 1, 6144" offset="20271241752" size="24576" /> <output> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10305" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="7319,7321,input"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10306" name="Constant_25652" type="Const" version="opset1"> <data element_type="u8" shape="50432, 6144" offset="20271266328" size="309854208" /> <output> <port id="0" precision="U8"> <dim>50432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10307" name="Convert_550484" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>50432</dim> <dim>6144</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10308" name="Constant_25652/zero_point" type="Const" version="opset1"> <data element_type="u8" shape="50432, 1" offset="20581120536" size="50432" /> <output> <port id="0" precision="U8"> <dim>50432</dim> <dim>1</dim> </port> </output> </layer> <layer id="10309" name="Convert_550487" type="Convert" version="opset1"> <data destination_type="f32" /> <input> <port id="0" precision="U8"> <dim>50432</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>50432</dim> <dim>1</dim> </port> </output> </layer> <layer id="10310" name="Subtract_550488" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>50432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10311" name="Constant_550489" type="Const" version="opset1"> <data element_type="f32" shape="50432, 1" offset="20581170968" size="201728" /> <output> <port id="0" precision="FP32"> <dim>50432</dim> <dim>1</dim> </port> </output> </layer> <layer id="10312" name="Constant_25652/fq_weights_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>50432</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> </output> </layer> <layer id="10313" name="__module.model.embed_out/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>6144</dim> </port> <port id="1" precision="FP32"> <dim>50432</dim> <dim>6144</dim> </port> </input> <output> <port id="2" precision="FP32" names="logits"> <dim>-1</dim> <dim>-1</dim> <dim>50432</dim> </port> </output> </layer> <layer id="10173" name="Assign_522304" type="Assign" version="opset6"> <data variable_id="past_key_values.0.keypresent.0.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10172" name="Assign_522306" type="Assign" version="opset6"> <data variable_id="past_key_values.0.valuepresent.0.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10171" name="Assign_522308" type="Assign" version="opset6"> <data variable_id="past_key_values.1.keypresent.1.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10170" name="Assign_522310" type="Assign" version="opset6"> <data variable_id="past_key_values.1.valuepresent.1.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10169" name="Assign_522312" type="Assign" version="opset6"> <data variable_id="past_key_values.10.keypresent.10.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10168" name="Assign_522314" type="Assign" version="opset6"> <data variable_id="past_key_values.10.valuepresent.10.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10167" name="Assign_522316" type="Assign" version="opset6"> <data variable_id="past_key_values.11.keypresent.11.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10166" name="Assign_522318" type="Assign" version="opset6"> <data variable_id="past_key_values.11.valuepresent.11.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10165" name="Assign_522320" type="Assign" version="opset6"> <data variable_id="past_key_values.12.keypresent.12.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10164" name="Assign_522322" type="Assign" version="opset6"> <data variable_id="past_key_values.12.valuepresent.12.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10163" name="Assign_522324" type="Assign" version="opset6"> <data variable_id="past_key_values.13.keypresent.13.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10162" name="Assign_522326" type="Assign" version="opset6"> <data variable_id="past_key_values.13.valuepresent.13.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10161" name="Assign_522328" type="Assign" version="opset6"> <data variable_id="past_key_values.14.keypresent.14.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10160" name="Assign_522330" type="Assign" version="opset6"> <data variable_id="past_key_values.14.valuepresent.14.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10159" name="Assign_522332" type="Assign" version="opset6"> <data variable_id="past_key_values.15.keypresent.15.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10158" name="Assign_522334" type="Assign" version="opset6"> <data variable_id="past_key_values.15.valuepresent.15.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10157" name="Assign_522336" type="Assign" version="opset6"> <data variable_id="past_key_values.16.keypresent.16.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10156" name="Assign_522338" type="Assign" version="opset6"> <data variable_id="past_key_values.16.valuepresent.16.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10155" name="Assign_522340" type="Assign" version="opset6"> <data variable_id="past_key_values.17.keypresent.17.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10154" name="Assign_522342" type="Assign" version="opset6"> <data variable_id="past_key_values.17.valuepresent.17.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10153" name="Assign_522344" type="Assign" version="opset6"> <data variable_id="past_key_values.18.keypresent.18.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10152" name="Assign_522346" type="Assign" version="opset6"> <data variable_id="past_key_values.18.valuepresent.18.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10151" name="Assign_522348" type="Assign" version="opset6"> <data variable_id="past_key_values.19.keypresent.19.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10150" name="Assign_522350" type="Assign" version="opset6"> <data variable_id="past_key_values.19.valuepresent.19.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10149" name="Assign_522352" type="Assign" version="opset6"> <data variable_id="past_key_values.2.keypresent.2.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10148" name="Assign_522354" type="Assign" version="opset6"> <data variable_id="past_key_values.2.valuepresent.2.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10147" name="Assign_522356" type="Assign" version="opset6"> <data variable_id="past_key_values.20.keypresent.20.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10146" name="Assign_522358" type="Assign" version="opset6"> <data variable_id="past_key_values.20.valuepresent.20.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10145" name="Assign_522360" type="Assign" version="opset6"> <data variable_id="past_key_values.21.keypresent.21.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10144" name="Assign_522362" type="Assign" version="opset6"> <data variable_id="past_key_values.21.valuepresent.21.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10143" name="Assign_522364" type="Assign" version="opset6"> <data variable_id="past_key_values.22.keypresent.22.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10142" name="Assign_522366" type="Assign" version="opset6"> <data variable_id="past_key_values.22.valuepresent.22.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10141" name="Assign_522368" type="Assign" version="opset6"> <data variable_id="past_key_values.23.keypresent.23.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10140" name="Assign_522370" type="Assign" version="opset6"> <data variable_id="past_key_values.23.valuepresent.23.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10139" name="Assign_522372" type="Assign" version="opset6"> <data variable_id="past_key_values.24.keypresent.24.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10138" name="Assign_522374" type="Assign" version="opset6"> <data variable_id="past_key_values.24.valuepresent.24.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10137" name="Assign_522376" type="Assign" version="opset6"> <data variable_id="past_key_values.25.keypresent.25.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10136" name="Assign_522378" type="Assign" version="opset6"> <data variable_id="past_key_values.25.valuepresent.25.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10135" name="Assign_522380" type="Assign" version="opset6"> <data variable_id="past_key_values.26.keypresent.26.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10134" name="Assign_522382" type="Assign" version="opset6"> <data variable_id="past_key_values.26.valuepresent.26.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10133" name="Assign_522384" type="Assign" version="opset6"> <data variable_id="past_key_values.27.keypresent.27.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10132" name="Assign_522386" type="Assign" version="opset6"> <data variable_id="past_key_values.27.valuepresent.27.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10131" name="Assign_522388" type="Assign" version="opset6"> <data variable_id="past_key_values.28.keypresent.28.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10130" name="Assign_522390" type="Assign" version="opset6"> <data variable_id="past_key_values.28.valuepresent.28.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10129" name="Assign_522392" type="Assign" version="opset6"> <data variable_id="past_key_values.29.keypresent.29.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10128" name="Assign_522394" type="Assign" version="opset6"> <data variable_id="past_key_values.29.valuepresent.29.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10127" name="Assign_522396" type="Assign" version="opset6"> <data variable_id="past_key_values.3.keypresent.3.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10126" name="Assign_522398" type="Assign" version="opset6"> <data variable_id="past_key_values.3.valuepresent.3.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10125" name="Assign_522400" type="Assign" version="opset6"> <data variable_id="past_key_values.30.keypresent.30.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10124" name="Assign_522402" type="Assign" version="opset6"> <data variable_id="past_key_values.30.valuepresent.30.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10123" name="Assign_522404" type="Assign" version="opset6"> <data variable_id="past_key_values.31.keypresent.31.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10122" name="Assign_522406" type="Assign" version="opset6"> <data variable_id="past_key_values.31.valuepresent.31.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10121" name="Assign_522408" type="Assign" version="opset6"> <data variable_id="past_key_values.32.keypresent.32.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10120" name="Assign_522410" type="Assign" version="opset6"> <data variable_id="past_key_values.32.valuepresent.32.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10119" name="Assign_522412" type="Assign" version="opset6"> <data variable_id="past_key_values.33.keypresent.33.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10118" name="Assign_522414" type="Assign" version="opset6"> <data variable_id="past_key_values.33.valuepresent.33.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10117" name="Assign_522416" type="Assign" version="opset6"> <data variable_id="past_key_values.34.keypresent.34.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10116" name="Assign_522418" type="Assign" version="opset6"> <data variable_id="past_key_values.34.valuepresent.34.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10115" name="Assign_522420" type="Assign" version="opset6"> <data variable_id="past_key_values.35.keypresent.35.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10114" name="Assign_522422" type="Assign" version="opset6"> <data variable_id="past_key_values.35.valuepresent.35.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10113" name="Assign_522424" type="Assign" version="opset6"> <data variable_id="past_key_values.36.keypresent.36.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10112" name="Assign_522426" type="Assign" version="opset6"> <data variable_id="past_key_values.36.valuepresent.36.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10111" name="Assign_522428" type="Assign" version="opset6"> <data variable_id="past_key_values.37.keypresent.37.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10110" name="Assign_522430" type="Assign" version="opset6"> <data variable_id="past_key_values.37.valuepresent.37.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10109" name="Assign_522432" type="Assign" version="opset6"> <data variable_id="past_key_values.38.keypresent.38.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10108" name="Assign_522434" type="Assign" version="opset6"> <data variable_id="past_key_values.38.valuepresent.38.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10107" name="Assign_522436" type="Assign" version="opset6"> <data variable_id="past_key_values.39.keypresent.39.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10106" name="Assign_522438" type="Assign" version="opset6"> <data variable_id="past_key_values.39.valuepresent.39.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10105" name="Assign_522440" type="Assign" version="opset6"> <data variable_id="past_key_values.4.keypresent.4.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10104" name="Assign_522442" type="Assign" version="opset6"> <data variable_id="past_key_values.4.valuepresent.4.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10103" name="Assign_522444" type="Assign" version="opset6"> <data variable_id="past_key_values.40.keypresent.40.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10102" name="Assign_522446" type="Assign" version="opset6"> <data variable_id="past_key_values.40.valuepresent.40.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10101" name="Assign_522448" type="Assign" version="opset6"> <data variable_id="past_key_values.41.keypresent.41.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10100" name="Assign_522450" type="Assign" version="opset6"> <data variable_id="past_key_values.41.valuepresent.41.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10099" name="Assign_522452" type="Assign" version="opset6"> <data variable_id="past_key_values.42.keypresent.42.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10098" name="Assign_522454" type="Assign" version="opset6"> <data variable_id="past_key_values.42.valuepresent.42.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10097" name="Assign_522456" type="Assign" version="opset6"> <data variable_id="past_key_values.43.keypresent.43.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10025" name="Assign_522458" type="Assign" version="opset6"> <data variable_id="past_key_values.43.valuepresent.43.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2241" name="Assign_522460" type="Assign" version="opset6"> <data variable_id="past_key_values.5.keypresent.5.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2240" name="Assign_522462" type="Assign" version="opset6"> <data variable_id="past_key_values.5.valuepresent.5.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2239" name="Assign_522464" type="Assign" version="opset6"> <data variable_id="past_key_values.6.keypresent.6.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2238" name="Assign_522466" type="Assign" version="opset6"> <data variable_id="past_key_values.6.valuepresent.6.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2237" name="Assign_522468" type="Assign" version="opset6"> <data variable_id="past_key_values.7.keypresent.7.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2236" name="Assign_522470" type="Assign" version="opset6"> <data variable_id="past_key_values.7.valuepresent.7.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2235" name="Assign_522472" type="Assign" version="opset6"> <data variable_id="past_key_values.8.keypresent.8.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2234" name="Assign_522474" type="Assign" version="opset6"> <data variable_id="past_key_values.8.valuepresent.8.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2233" name="Assign_522476" type="Assign" version="opset6"> <data variable_id="past_key_values.9.keypresent.9.key" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="2161" name="Assign_522478" type="Assign" version="opset6"> <data variable_id="past_key_values.9.valuepresent.9.value" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>-1</dim> <dim>64</dim> <dim>-1</dim> <dim>96</dim> </port> </output> </layer> <layer id="10314" name="Result_192830" type="Result" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>50432</dim> </port> </input> </layer> </layers> <edges> <edge from-layer="0" from-port="0" to-layer="3232" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="8545" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4156" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4068" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="5311" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="3925" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="8457" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="3837" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="3694" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="3606" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6004" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="3463" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="8314" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="3375" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4299" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6147" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="450" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="3144" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="8226" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="3001" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="8083" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="2770" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="2682" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7995" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="2539" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="2451" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7852" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="2250" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="8919" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9381" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9469" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9238" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9612" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="5685" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9150" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9700" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="5542" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9007" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9843" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="5223" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="9931" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="5773" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="2913" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="10034" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="362" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="5080" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4992" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="5916" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4849" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4761" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="8776" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4618" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4530" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="102" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="8688" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="4387" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1055" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1517" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7071" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1374" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1286" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6466" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6378" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1143" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7533" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="2170" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7159" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="912" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7302" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="593" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="5454" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="824" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7390" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="681" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7621" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6609" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1605" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6235" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6697" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1748" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="16" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1836" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="2067" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6928" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="194" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="7764" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="1979" to-port="1" /> <edge from-layer="0" from-port="0" to-layer="6840" to-port="1" /> <edge from-layer="1" from-port="0" to-layer="111" to-port="0" /> <edge from-layer="2" from-port="0" to-layer="239" to-port="0" /> <edge from-layer="3" from-port="0" to-layer="24" to-port="0" /> <edge from-layer="3" from-port="0" to-layer="233" to-port="0" /> <edge from-layer="3" from-port="0" to-layer="5" to-port="0" /> <edge from-layer="4" from-port="0" to-layer="13" to-port="0" /> <edge from-layer="5" from-port="1" to-layer="8" to-port="0" /> <edge from-layer="6" from-port="0" to-layer="8" to-port="1" /> <edge from-layer="7" from-port="0" to-layer="8" to-port="2" /> <edge from-layer="8" from-port="3" to-layer="8079" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7067" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="5769" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6231" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9003" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="8222" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6924" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6693" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9146" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7760" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6143" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6836" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="190" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9234" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7991" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9377" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7848" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6000" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7386" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="5307" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="8541" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="8772" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7298" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7529" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="5450" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="8915" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="8684" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6462" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7155" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="5912" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="8453" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="7617" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6374" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="8310" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="6605" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1832" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="446" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="3228" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="3140" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2997" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2909" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2766" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2678" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2535" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2447" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2246" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2166" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="2063" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1975" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9608" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1744" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1601" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1513" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1370" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1282" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1139" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="1051" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="908" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="820" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="677" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="589" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="12" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="3459" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9465" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9696" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="5681" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="5538" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9839" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="9927" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="5219" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="358" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="10030" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="5076" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4988" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4845" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4757" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4526" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4614" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4383" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4295" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4152" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="98" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="4064" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="3921" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="3833" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="3690" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="3602" to-port="0" /> <edge from-layer="8" from-port="3" to-layer="3371" to-port="0" /> <edge from-layer="9" from-port="0" to-layer="12" to-port="1" /> <edge from-layer="10" from-port="0" to-layer="12" to-port="2" /> <edge from-layer="11" from-port="0" to-layer="12" to-port="3" /> <edge from-layer="12" from-port="4" to-layer="13" to-port="1" /> <edge from-layer="13" from-port="2" to-layer="14" to-port="0" /> <edge from-layer="14" from-port="1" to-layer="16" to-port="0" /> <edge from-layer="15" from-port="0" to-layer="16" to-port="2" /> <edge from-layer="16" from-port="3" to-layer="2160" to-port="0" /> <edge from-layer="17" from-port="0" to-layer="18" to-port="0" /> <edge from-layer="18" from-port="1" to-layer="21" to-port="0" /> <edge from-layer="19" from-port="0" to-layer="20" to-port="0" /> <edge from-layer="20" from-port="1" to-layer="21" to-port="1" /> <edge from-layer="21" from-port="2" to-layer="23" to-port="0" /> <edge from-layer="22" from-port="0" to-layer="23" to-port="1" /> <edge from-layer="23" from-port="2" to-layer="26" to-port="0" /> <edge from-layer="24" from-port="1" to-layer="26" to-port="1" /> <edge from-layer="25" from-port="0" to-layer="26" to-port="2" /> <edge from-layer="26" from-port="3" to-layer="287" to-port="1" /> <edge from-layer="26" from-port="3" to-layer="28" to-port="0" /> <edge from-layer="26" from-port="3" to-layer="55" to-port="0" /> <edge from-layer="27" from-port="0" to-layer="28" to-port="1" /> <edge from-layer="28" from-port="2" to-layer="30" to-port="0" /> <edge from-layer="29" from-port="0" to-layer="30" to-port="1" /> <edge from-layer="30" from-port="2" to-layer="32" to-port="0" /> <edge from-layer="31" from-port="0" to-layer="32" to-port="1" /> <edge from-layer="32" from-port="2" to-layer="40" to-port="0" /> <edge from-layer="33" from-port="0" to-layer="34" to-port="0" /> <edge from-layer="34" from-port="1" to-layer="37" to-port="0" /> <edge from-layer="35" from-port="0" to-layer="36" to-port="0" /> <edge from-layer="36" from-port="1" to-layer="37" to-port="1" /> <edge from-layer="37" from-port="2" to-layer="39" to-port="0" /> <edge from-layer="38" from-port="0" to-layer="39" to-port="1" /> <edge from-layer="39" from-port="2" to-layer="40" to-port="1" /> <edge from-layer="40" from-port="2" to-layer="42" to-port="0" /> <edge from-layer="41" from-port="0" to-layer="42" to-port="1" /> <edge from-layer="42" from-port="2" to-layer="43" to-port="0" /> <edge from-layer="43" from-port="1" to-layer="51" to-port="0" /> <edge from-layer="44" from-port="0" to-layer="45" to-port="0" /> <edge from-layer="45" from-port="1" to-layer="48" to-port="0" /> <edge from-layer="46" from-port="0" to-layer="47" to-port="0" /> <edge from-layer="47" from-port="1" to-layer="48" to-port="1" /> <edge from-layer="48" from-port="2" to-layer="50" to-port="0" /> <edge from-layer="49" from-port="0" to-layer="50" to-port="1" /> <edge from-layer="50" from-port="2" to-layer="51" to-port="1" /> <edge from-layer="51" from-port="2" to-layer="53" to-port="0" /> <edge from-layer="52" from-port="0" to-layer="53" to-port="1" /> <edge from-layer="53" from-port="2" to-layer="286" to-port="0" /> <edge from-layer="54" from-port="0" to-layer="55" to-port="1" /> <edge from-layer="55" from-port="2" to-layer="57" to-port="0" /> <edge from-layer="56" from-port="0" to-layer="57" to-port="1" /> <edge from-layer="57" from-port="2" to-layer="59" to-port="0" /> <edge from-layer="58" from-port="0" to-layer="59" to-port="1" /> <edge from-layer="59" from-port="2" to-layer="67" to-port="0" /> <edge from-layer="60" from-port="0" to-layer="61" to-port="0" /> <edge from-layer="61" from-port="1" to-layer="64" to-port="0" /> <edge from-layer="62" from-port="0" to-layer="63" to-port="0" /> <edge from-layer="63" from-port="1" to-layer="64" to-port="1" /> <edge from-layer="64" from-port="2" to-layer="66" to-port="0" /> <edge from-layer="65" from-port="0" to-layer="66" to-port="1" /> <edge from-layer="66" from-port="2" to-layer="67" to-port="1" /> <edge from-layer="67" from-port="2" to-layer="69" to-port="0" /> <edge from-layer="68" from-port="0" to-layer="69" to-port="1" /> <edge from-layer="69" from-port="2" to-layer="71" to-port="0" /> <edge from-layer="70" from-port="0" to-layer="71" to-port="1" /> <edge from-layer="71" from-port="2" to-layer="75" to-port="0" /> <edge from-layer="71" from-port="2" to-layer="198" to-port="0" /> <edge from-layer="71" from-port="2" to-layer="87" to-port="0" /> <edge from-layer="72" from-port="0" to-layer="75" to-port="1" /> <edge from-layer="73" from-port="0" to-layer="75" to-port="2" /> <edge from-layer="74" from-port="0" to-layer="75" to-port="3" /> <edge from-layer="75" from-port="4" to-layer="77" to-port="0" /> <edge from-layer="76" from-port="0" to-layer="77" to-port="1" /> <edge from-layer="77" from-port="2" to-layer="151" to-port="0" /> <edge from-layer="77" from-port="2" to-layer="81" to-port="0" /> <edge from-layer="78" from-port="0" to-layer="81" to-port="1" /> <edge from-layer="79" from-port="0" to-layer="81" to-port="2" /> <edge from-layer="80" from-port="0" to-layer="81" to-port="3" /> <edge from-layer="81" from-port="4" to-layer="116" to-port="0" /> <edge from-layer="81" from-port="4" to-layer="119" to-port="0" /> <edge from-layer="81" from-port="4" to-layer="137" to-port="0" /> <edge from-layer="81" from-port="4" to-layer="132" to-port="0" /> <edge from-layer="82" from-port="0" to-layer="110" to-port="0" /> <edge from-layer="83" from-port="0" to-layer="110" to-port="1" /> <edge from-layer="84" from-port="0" to-layer="87" to-port="1" /> <edge from-layer="85" from-port="0" to-layer="87" to-port="2" /> <edge from-layer="86" from-port="0" to-layer="87" to-port="3" /> <edge from-layer="87" from-port="4" to-layer="89" to-port="0" /> <edge from-layer="88" from-port="0" to-layer="89" to-port="1" /> <edge from-layer="89" from-port="2" to-layer="183" to-port="0" /> <edge from-layer="89" from-port="2" to-layer="90" to-port="0" /> <edge from-layer="89" from-port="2" to-layer="156" to-port="0" /> <edge from-layer="90" from-port="1" to-layer="93" to-port="0" /> <edge from-layer="91" from-port="0" to-layer="10052" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="10049" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="8910" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="2904" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="5919" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="5214" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="3609" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="353" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="5226" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="9846" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="5688" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="9834" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="3597" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="5676" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4071" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4752" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="9153" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4290" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4302" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4521" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="9372" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="9141" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="93" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4533" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4059" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="9384" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="9615" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="3840" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="105" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4764" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4983" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="4995" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="3828" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="8922" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="9603" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="5907" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="365" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1277" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7755" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="6369" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="5457" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1751" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1739" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7536" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="6381" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7524" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1520" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1508" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1289" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7305" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7293" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="3378" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="5445" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1058" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1046" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7074" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="6600" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="6612" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7062" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="827" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="815" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="596" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="6843" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="6831" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="584" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="2916" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1970" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="245" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="2454" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="2673" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="8217" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="6150" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="2685" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="8229" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7998" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="6138" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="8448" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="8460" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="3135" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="3147" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="3366" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="8679" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="8691" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="2442" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7986" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="2188" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="2185" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="1982" to-port="1" /> <edge from-layer="91" from-port="0" to-layer="7767" to-port="1" /> <edge from-layer="92" from-port="0" to-layer="93" to-port="2" /> <edge from-layer="93" from-port="3" to-layer="106" to-port="0" /> <edge from-layer="94" from-port="0" to-layer="99" to-port="0" /> <edge from-layer="95" from-port="0" to-layer="98" to-port="1" /> <edge from-layer="96" from-port="0" to-layer="98" to-port="2" /> <edge from-layer="97" from-port="0" to-layer="98" to-port="3" /> <edge from-layer="98" from-port="4" to-layer="99" to-port="1" /> <edge from-layer="99" from-port="2" to-layer="100" to-port="0" /> <edge from-layer="100" from-port="1" to-layer="102" to-port="0" /> <edge from-layer="101" from-port="0" to-layer="102" to-port="2" /> <edge from-layer="102" from-port="3" to-layer="185" to-port="0" /> <edge from-layer="102" from-port="3" to-layer="103" to-port="0" /> <edge from-layer="103" from-port="1" to-layer="105" to-port="0" /> <edge from-layer="104" from-port="0" to-layer="105" to-port="2" /> <edge from-layer="105" from-port="3" to-layer="106" to-port="1" /> <edge from-layer="106" from-port="2" to-layer="108" to-port="0" /> <edge from-layer="107" from-port="0" to-layer="108" to-port="1" /> <edge from-layer="108" from-port="2" to-layer="142" to-port="2" /> <edge from-layer="108" from-port="2" to-layer="110" to-port="2" /> <edge from-layer="109" from-port="0" to-layer="110" to-port="3" /> <edge from-layer="110" from-port="4" to-layer="113" to-port="0" /> <edge from-layer="111" from-port="1" to-layer="3644" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="3616" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1324" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8726" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="7340" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8698" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="3413" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5926" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8929" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1296" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="3847" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="144" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="7312" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="3875" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8957" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5492" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="4078" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1093" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="4106" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="9160" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1065" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="9188" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="4309" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="400" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="7109" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="4337" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="7081" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1989" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1786" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2720" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2692" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="6157" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8033" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="7774" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5464" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2489" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8236" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2461" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8005" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2017" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2223" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2195" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="6185" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="7802" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1527" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8264" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1758" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2923" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="7571" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="2951" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="7543" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8467" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="6388" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="8495" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="3154" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="3182" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="1555" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5954" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="6416" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="3385" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="9419" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5030" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="10059" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="9391" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="6878" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="6647" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="9881" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="9622" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="862" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="6850" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5233" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="4771" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="9650" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5002" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="372" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="113" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="4799" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="603" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="834" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5723" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5261" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="631" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="4568" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="6619" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="10087" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="9853" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="5695" to-port="1" /> <edge from-layer="111" from-port="1" to-layer="4540" to-port="1" /> <edge from-layer="112" from-port="0" to-layer="113" to-port="2" /> <edge from-layer="113" from-port="3" to-layer="115" to-port="0" /> <edge from-layer="114" from-port="0" to-layer="145" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="10088" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="9623" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8468" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2924" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8265" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="7544" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2952" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="632" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5955" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="3155" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1759" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8237" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="604" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="10060" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="244" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="7803" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="7775" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5234" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="6648" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2196" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2018" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2224" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5465" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="9854" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1990" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="6158" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2721" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="9651" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8006" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2462" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2490" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="9882" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8034" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="6186" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="115" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="7572" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="2693" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1787" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="4569" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1297" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="6389" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="835" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="4772" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="3848" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="3876" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8958" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="6417" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="7110" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="9392" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1094" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="4079" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="863" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8930" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="4107" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1066" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5724" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="9161" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="6620" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="4541" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="9189" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="401" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="4310" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="4338" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="7082" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5262" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="9420" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="3183" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1556" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8496" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5493" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="6851" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5031" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1528" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5927" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="3386" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5003" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="3414" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="7341" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8699" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="5696" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="8727" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="7313" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="373" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="3617" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="1325" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="6879" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="3645" to-port="1" /> <edge from-layer="114" from-port="0" to-layer="4800" to-port="1" /> <edge from-layer="115" from-port="2" to-layer="116" to-port="1" /> <edge from-layer="115" from-port="2" to-layer="157" to-port="1" /> <edge from-layer="116" from-port="2" to-layer="147" to-port="0" /> <edge from-layer="117" from-port="0" to-layer="129" to-port="0" /> <edge from-layer="118" from-port="0" to-layer="129" to-port="1" /> <edge from-layer="119" from-port="1" to-layer="122" to-port="0" /> <edge from-layer="120" from-port="0" to-layer="4085" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="880" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="4547" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5702" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="9206" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="4355" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="7088" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="4316" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="9167" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="4124" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1072" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="4586" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5741" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1111" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5279" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="7127" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="6434" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8975" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="3893" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="3854" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8936" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5048" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="6665" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="9668" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="9860" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="9899" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5240" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="9629" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="610" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="10066" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="10216" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="649" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5510" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="6857" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5009" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="6626" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="9437" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="6896" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="4817" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="4778" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="841" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="9398" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8051" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2930" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8282" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8243" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1765" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2738" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2699" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1804" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5471" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="7589" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="6203" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="7550" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8012" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2507" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2468" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2293" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1996" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="6164" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2202" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2035" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="7820" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="3392" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1303" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="3662" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="3623" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1342" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8744" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="7319" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8705" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="3431" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="7358" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="6395" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="7781" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5933" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1534" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8513" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="3200" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="1573" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="3161" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="8474" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="5972" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="2969" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="379" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="122" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="162" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="418" to-port="1" /> <edge from-layer="121" from-port="0" to-layer="122" to-port="2" /> <edge from-layer="122" from-port="3" to-layer="124" to-port="0" /> <edge from-layer="123" from-port="0" to-layer="163" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8937" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="6165" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5049" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="3393" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5511" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="6204" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="4125" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="124" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2931" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="4086" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8745" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="7551" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8706" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="7590" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="7128" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2739" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="611" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2700" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="4548" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="4779" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="380" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="9168" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="3162" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5973" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5934" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="4818" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="9207" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="4587" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="650" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5472" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2970" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5010" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="7320" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5742" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="3432" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8976" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5280" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5703" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="4356" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="7359" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="4317" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8283" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1997" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="7821" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="6627" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="6666" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1073" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="9669" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="842" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="3663" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1805" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1766" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="9630" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="9438" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="6858" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="3624" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1574" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1535" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8244" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8013" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="5241" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="6897" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8052" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1343" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1304" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2294" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="7089" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="10217" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="6396" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="10067" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="6435" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2508" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2469" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="9399" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8514" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="3894" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="881" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="3855" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="419" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2203" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="1112" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="9861" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="3201" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="7782" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="8475" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="9900" to-port="1" /> <edge from-layer="123" from-port="0" to-layer="2036" to-port="1" /> <edge from-layer="124" from-port="2" to-layer="125" to-port="0" /> <edge from-layer="125" from-port="1" to-layer="127" to-port="0" /> <edge from-layer="126" from-port="0" to-layer="127" to-port="1" /> <edge from-layer="127" from-port="2" to-layer="129" to-port="2" /> <edge from-layer="128" from-port="0" to-layer="129" to-port="3" /> <edge from-layer="129" from-port="4" to-layer="137" to-port="2" /> <edge from-layer="129" from-port="4" to-layer="132" to-port="1" /> <edge from-layer="130" from-port="0" to-layer="132" to-port="2" /> <edge from-layer="131" from-port="0" to-layer="132" to-port="3" /> <edge from-layer="132" from-port="4" to-layer="134" to-port="0" /> <edge from-layer="133" from-port="0" to-layer="134" to-port="1" /> <edge from-layer="134" from-port="2" to-layer="138" to-port="0" /> <edge from-layer="135" from-port="0" to-layer="137" to-port="1" /> <edge from-layer="136" from-port="0" to-layer="137" to-port="3" /> <edge from-layer="137" from-port="4" to-layer="138" to-port="1" /> <edge from-layer="138" from-port="2" to-layer="146" to-port="0" /> <edge from-layer="139" from-port="0" to-layer="142" to-port="0" /> <edge from-layer="140" from-port="0" to-layer="142" to-port="1" /> <edge from-layer="141" from-port="0" to-layer="142" to-port="3" /> <edge from-layer="142" from-port="4" to-layer="144" to-port="0" /> <edge from-layer="143" from-port="0" to-layer="144" to-port="2" /> <edge from-layer="144" from-port="3" to-layer="145" to-port="0" /> <edge from-layer="145" from-port="2" to-layer="146" to-port="1" /> <edge from-layer="145" from-port="2" to-layer="178" to-port="1" /> <edge from-layer="146" from-port="2" to-layer="147" to-port="1" /> <edge from-layer="147" from-port="2" to-layer="152" to-port="0" /> <edge from-layer="148" from-port="0" to-layer="151" to-port="1" /> <edge from-layer="149" from-port="0" to-layer="151" to-port="2" /> <edge from-layer="150" from-port="0" to-layer="151" to-port="3" /> <edge from-layer="151" from-port="4" to-layer="152" to-port="1" /> <edge from-layer="152" from-port="2" to-layer="271" to-port="0" /> <edge from-layer="152" from-port="2" to-layer="209" to-port="0" /> <edge from-layer="153" from-port="0" to-layer="156" to-port="1" /> <edge from-layer="154" from-port="0" to-layer="156" to-port="2" /> <edge from-layer="155" from-port="0" to-layer="156" to-port="3" /> <edge from-layer="156" from-port="4" to-layer="176" to-port="0" /> <edge from-layer="156" from-port="4" to-layer="171" to-port="0" /> <edge from-layer="156" from-port="4" to-layer="157" to-port="0" /> <edge from-layer="156" from-port="4" to-layer="160" to-port="0" /> <edge from-layer="157" from-port="2" to-layer="179" to-port="0" /> <edge from-layer="158" from-port="0" to-layer="168" to-port="0" /> <edge from-layer="159" from-port="0" to-layer="168" to-port="1" /> <edge from-layer="160" from-port="1" to-layer="162" to-port="0" /> <edge from-layer="161" from-port="0" to-layer="162" to-port="2" /> <edge from-layer="162" from-port="3" to-layer="163" to-port="0" /> <edge from-layer="163" from-port="2" to-layer="164" to-port="0" /> <edge from-layer="164" from-port="1" to-layer="166" to-port="0" /> <edge from-layer="165" from-port="0" to-layer="166" to-port="1" /> <edge from-layer="166" from-port="2" to-layer="168" to-port="2" /> <edge from-layer="167" from-port="0" to-layer="168" to-port="3" /> <edge from-layer="168" from-port="4" to-layer="176" to-port="2" /> <edge from-layer="168" from-port="4" to-layer="171" to-port="1" /> <edge from-layer="169" from-port="0" to-layer="171" to-port="2" /> <edge from-layer="170" from-port="0" to-layer="171" to-port="3" /> <edge from-layer="171" from-port="4" to-layer="173" to-port="0" /> <edge from-layer="172" from-port="0" to-layer="173" to-port="1" /> <edge from-layer="173" from-port="2" to-layer="177" to-port="0" /> <edge from-layer="174" from-port="0" to-layer="176" to-port="1" /> <edge from-layer="175" from-port="0" to-layer="176" to-port="3" /> <edge from-layer="176" from-port="4" to-layer="177" to-port="1" /> <edge from-layer="177" from-port="2" to-layer="178" to-port="0" /> <edge from-layer="178" from-port="2" to-layer="179" to-port="1" /> <edge from-layer="179" from-port="2" to-layer="184" to-port="0" /> <edge from-layer="180" from-port="0" to-layer="183" to-port="1" /> <edge from-layer="181" from-port="0" to-layer="183" to-port="2" /> <edge from-layer="182" from-port="0" to-layer="183" to-port="3" /> <edge from-layer="183" from-port="4" to-layer="184" to-port="1" /> <edge from-layer="184" from-port="2" to-layer="185" to-port="1" /> <edge from-layer="185" from-port="2" to-layer="205" to-port="0" /> <edge from-layer="185" from-port="2" to-layer="271" to-port="1" /> <edge from-layer="185" from-port="2" to-layer="10173" to-port="0" /> <edge from-layer="186" from-port="0" to-layer="191" to-port="0" /> <edge from-layer="187" from-port="0" to-layer="190" to-port="1" /> <edge from-layer="188" from-port="0" to-layer="190" to-port="2" /> <edge from-layer="189" from-port="0" to-layer="190" to-port="3" /> <edge from-layer="190" from-port="4" to-layer="191" to-port="1" /> <edge from-layer="191" from-port="2" to-layer="192" to-port="0" /> <edge from-layer="192" from-port="1" to-layer="194" to-port="0" /> <edge from-layer="193" from-port="0" to-layer="194" to-port="2" /> <edge from-layer="194" from-port="3" to-layer="201" to-port="0" /> <edge from-layer="195" from-port="0" to-layer="198" to-port="1" /> <edge from-layer="196" from-port="0" to-layer="198" to-port="2" /> <edge from-layer="197" from-port="0" to-layer="198" to-port="3" /> <edge from-layer="198" from-port="4" to-layer="200" to-port="0" /> <edge from-layer="199" from-port="0" to-layer="200" to-port="1" /> <edge from-layer="200" from-port="2" to-layer="201" to-port="1" /> <edge from-layer="201" from-port="2" to-layer="271" to-port="2" /> <edge from-layer="201" from-port="2" to-layer="10172" to-port="0" /> <edge from-layer="202" from-port="0" to-layer="221" to-port="0" /> <edge from-layer="203" from-port="0" to-layer="216" to-port="0" /> <edge from-layer="204" from-port="0" to-layer="216" to-port="1" /> <edge from-layer="204" from-port="0" to-layer="219" to-port="1" /> <edge from-layer="205" from-port="1" to-layer="208" to-port="0" /> <edge from-layer="206" from-port="0" to-layer="8792" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6479" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="8327" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9482" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="7406" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="7403" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="5324" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9020" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9023" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="5327" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="8099" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="7172" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="7175" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="8096" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="7865" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="8558" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3248" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="208" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="8561" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="7637" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="928" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9254" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3245" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6941" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="7634" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9251" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="211" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="7868" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="925" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="8789" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="8330" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6944" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="5555" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="2783" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="2786" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9485" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="466" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="463" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3014" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3017" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="4172" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="1390" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="5558" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="4400" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="4403" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="5786" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="5789" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="4631" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="4634" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="2080" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="1621" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="1849" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="1852" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3707" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3710" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3479" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="1618" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3476" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="4169" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="2083" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="2321" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="2324" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="2552" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3938" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="3941" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="2555" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="1156" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6248" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="5096" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6251" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="1387" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="5093" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="10247" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="694" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6710" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="4865" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="4862" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6713" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="10244" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6020" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6482" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9947" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9713" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="1159" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9944" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="6017" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="697" to-port="1" /> <edge from-layer="206" from-port="0" to-layer="9716" to-port="1" /> <edge from-layer="207" from-port="0" to-layer="208" to-port="2" /> <edge from-layer="208" from-port="3" to-layer="218" to-port="0" /> <edge from-layer="208" from-port="3" to-layer="212" to-port="0" /> <edge from-layer="209" from-port="1" to-layer="260" to-port="0" /> <edge from-layer="209" from-port="1" to-layer="211" to-port="0" /> <edge from-layer="210" from-port="0" to-layer="211" to-port="2" /> <edge from-layer="211" from-port="3" to-layer="212" to-port="1" /> <edge from-layer="212" from-port="2" to-layer="214" to-port="0" /> <edge from-layer="213" from-port="0" to-layer="218" to-port="1" /> <edge from-layer="213" from-port="0" to-layer="214" to-port="1" /> <edge from-layer="214" from-port="2" to-layer="216" to-port="2" /> <edge from-layer="215" from-port="0" to-layer="216" to-port="3" /> <edge from-layer="215" from-port="0" to-layer="219" to-port="3" /> <edge from-layer="216" from-port="4" to-layer="221" to-port="1" /> <edge from-layer="217" from-port="0" to-layer="219" to-port="0" /> <edge from-layer="218" from-port="2" to-layer="219" to-port="2" /> <edge from-layer="218" from-port="2" to-layer="226" to-port="2" /> <edge from-layer="219" from-port="4" to-layer="221" to-port="2" /> <edge from-layer="220" from-port="0" to-layer="221" to-port="3" /> <edge from-layer="221" from-port="4" to-layer="228" to-port="0" /> <edge from-layer="222" from-port="0" to-layer="228" to-port="1" /> <edge from-layer="223" from-port="0" to-layer="226" to-port="0" /> <edge from-layer="224" from-port="0" to-layer="226" to-port="1" /> <edge from-layer="225" from-port="0" to-layer="226" to-port="3" /> <edge from-layer="226" from-port="4" to-layer="228" to-port="2" /> <edge from-layer="227" from-port="0" to-layer="228" to-port="3" /> <edge from-layer="228" from-port="4" to-layer="231" to-port="0" /> <edge from-layer="229" from-port="0" to-layer="231" to-port="1" /> <edge from-layer="230" from-port="0" to-layer="231" to-port="2" /> <edge from-layer="231" from-port="3" to-layer="256" to-port="0" /> <edge from-layer="232" from-port="0" to-layer="253" to-port="0" /> <edge from-layer="233" from-port="1" to-layer="236" to-port="0" /> <edge from-layer="234" from-port="0" to-layer="236" to-port="1" /> <edge from-layer="235" from-port="0" to-layer="236" to-port="2" /> <edge from-layer="236" from-port="3" to-layer="238" to-port="0" /> <edge from-layer="237" from-port="0" to-layer="238" to-port="1" /> <edge from-layer="238" from-port="2" to-layer="239" to-port="1" /> <edge from-layer="239" from-port="2" to-layer="243" to-port="0" /> <edge from-layer="240" from-port="0" to-layer="243" to-port="1" /> <edge from-layer="241" from-port="0" to-layer="243" to-port="2" /> <edge from-layer="242" from-port="0" to-layer="243" to-port="3" /> <edge from-layer="243" from-port="4" to-layer="244" to-port="0" /> <edge from-layer="244" from-port="2" to-layer="245" to-port="0" /> <edge from-layer="245" from-port="2" to-layer="249" to-port="0" /> <edge from-layer="246" from-port="0" to-layer="249" to-port="1" /> <edge from-layer="247" from-port="0" to-layer="249" to-port="2" /> <edge from-layer="248" from-port="0" to-layer="249" to-port="3" /> <edge from-layer="249" from-port="4" to-layer="250" to-port="0" /> <edge from-layer="250" from-port="1" to-layer="252" to-port="0" /> <edge from-layer="251" from-port="0" to-layer="252" to-port="1" /> <edge from-layer="252" from-port="2" to-layer="253" to-port="1" /> <edge from-layer="253" from-port="2" to-layer="255" to-port="0" /> <edge from-layer="254" from-port="0" to-layer="255" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="1411" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="2807" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="5579" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="949" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="6734" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="6965" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="6503" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="1180" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="6272" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="6041" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="7196" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="5810" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="3038" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="1642" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="7427" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="2576" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="7658" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="2345" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="487" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="2104" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="1873" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="5348" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="7889" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="10268" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="8582" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="8813" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="3269" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="9044" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="9275" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="9506" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="9737" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="718" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="9968" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="8120" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="5117" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="4886" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="4655" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="4424" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="4193" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="3962" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="3731" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="3500" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="8351" to-port="1" /> <edge from-layer="255" from-port="2" to-layer="256" to-port="1" /> <edge from-layer="256" from-port="2" to-layer="270" to-port="0" /> <edge from-layer="256" from-port="2" to-layer="257" to-port="0" /> <edge from-layer="257" from-port="1" to-layer="269" to-port="0" /> <edge from-layer="258" from-port="0" to-layer="260" to-port="1" /> <edge from-layer="259" from-port="0" to-layer="260" to-port="2" /> <edge from-layer="260" from-port="3" to-layer="264" to-port="0" /> <edge from-layer="261" from-port="0" to-layer="264" to-port="1" /> <edge from-layer="262" from-port="0" to-layer="264" to-port="2" /> <edge from-layer="263" from-port="0" to-layer="264" to-port="3" /> <edge from-layer="264" from-port="4" to-layer="268" to-port="2" /> <edge from-layer="264" from-port="4" to-layer="266" to-port="0" /> <edge from-layer="265" from-port="0" to-layer="266" to-port="1" /> <edge from-layer="266" from-port="2" to-layer="268" to-port="0" /> <edge from-layer="267" from-port="0" to-layer="268" to-port="1" /> <edge from-layer="268" from-port="3" to-layer="269" to-port="1" /> <edge from-layer="269" from-port="2" to-layer="270" to-port="1" /> <edge from-layer="270" from-port="2" to-layer="271" to-port="3" /> <edge from-layer="271" from-port="4" to-layer="273" to-port="0" /> <edge from-layer="272" from-port="0" to-layer="273" to-port="1" /> <edge from-layer="273" from-port="2" to-layer="275" to-port="0" /> <edge from-layer="274" from-port="0" to-layer="275" to-port="1" /> <edge from-layer="275" from-port="2" to-layer="283" to-port="0" /> <edge from-layer="276" from-port="0" to-layer="277" to-port="0" /> <edge from-layer="277" from-port="1" to-layer="280" to-port="0" /> <edge from-layer="278" from-port="0" to-layer="279" to-port="0" /> <edge from-layer="279" from-port="1" to-layer="280" to-port="1" /> <edge from-layer="280" from-port="2" to-layer="282" to-port="0" /> <edge from-layer="281" from-port="0" to-layer="282" to-port="1" /> <edge from-layer="282" from-port="2" to-layer="283" to-port="1" /> <edge from-layer="283" from-port="2" to-layer="285" to-port="0" /> <edge from-layer="284" from-port="0" to-layer="285" to-port="1" /> <edge from-layer="285" from-port="2" to-layer="286" to-port="1" /> <edge from-layer="286" from-port="2" to-layer="287" to-port="0" /> <edge from-layer="287" from-port="2" to-layer="289" to-port="0" /> <edge from-layer="287" from-port="2" to-layer="316" to-port="0" /> <edge from-layer="287" from-port="2" to-layer="518" to-port="1" /> <edge from-layer="288" from-port="0" to-layer="289" to-port="1" /> <edge from-layer="289" from-port="2" to-layer="291" to-port="0" /> <edge from-layer="290" from-port="0" to-layer="291" to-port="1" /> <edge from-layer="291" from-port="2" to-layer="293" to-port="0" /> <edge from-layer="292" from-port="0" to-layer="293" to-port="1" /> <edge from-layer="293" from-port="2" to-layer="301" to-port="0" /> <edge from-layer="294" from-port="0" to-layer="295" to-port="0" /> <edge from-layer="295" from-port="1" to-layer="298" to-port="0" /> <edge from-layer="296" from-port="0" to-layer="297" to-port="0" /> <edge from-layer="297" from-port="1" to-layer="298" to-port="1" /> <edge from-layer="298" from-port="2" to-layer="300" to-port="0" /> <edge from-layer="299" from-port="0" to-layer="300" to-port="1" /> <edge from-layer="300" from-port="2" to-layer="301" to-port="1" /> <edge from-layer="301" from-port="2" to-layer="303" to-port="0" /> <edge from-layer="302" from-port="0" to-layer="303" to-port="1" /> <edge from-layer="303" from-port="2" to-layer="304" to-port="0" /> <edge from-layer="304" from-port="1" to-layer="312" to-port="0" /> <edge from-layer="305" from-port="0" to-layer="306" to-port="0" /> <edge from-layer="306" from-port="1" to-layer="309" to-port="0" /> <edge from-layer="307" from-port="0" to-layer="308" to-port="0" /> <edge from-layer="308" from-port="1" to-layer="309" to-port="1" /> <edge from-layer="309" from-port="2" to-layer="311" to-port="0" /> <edge from-layer="310" from-port="0" to-layer="311" to-port="1" /> <edge from-layer="311" from-port="2" to-layer="312" to-port="1" /> <edge from-layer="312" from-port="2" to-layer="314" to-port="0" /> <edge from-layer="313" from-port="0" to-layer="314" to-port="1" /> <edge from-layer="314" from-port="2" to-layer="517" to-port="0" /> <edge from-layer="315" from-port="0" to-layer="316" to-port="1" /> <edge from-layer="316" from-port="2" to-layer="318" to-port="0" /> <edge from-layer="317" from-port="0" to-layer="318" to-port="1" /> <edge from-layer="318" from-port="2" to-layer="320" to-port="0" /> <edge from-layer="319" from-port="0" to-layer="320" to-port="1" /> <edge from-layer="320" from-port="2" to-layer="328" to-port="0" /> <edge from-layer="321" from-port="0" to-layer="322" to-port="0" /> <edge from-layer="322" from-port="1" to-layer="325" to-port="0" /> <edge from-layer="323" from-port="0" to-layer="324" to-port="0" /> <edge from-layer="324" from-port="1" to-layer="325" to-port="1" /> <edge from-layer="325" from-port="2" to-layer="327" to-port="0" /> <edge from-layer="326" from-port="0" to-layer="327" to-port="1" /> <edge from-layer="327" from-port="2" to-layer="328" to-port="1" /> <edge from-layer="328" from-port="2" to-layer="330" to-port="0" /> <edge from-layer="329" from-port="0" to-layer="330" to-port="1" /> <edge from-layer="330" from-port="2" to-layer="332" to-port="0" /> <edge from-layer="331" from-port="0" to-layer="332" to-port="1" /> <edge from-layer="332" from-port="2" to-layer="454" to-port="0" /> <edge from-layer="332" from-port="2" to-layer="336" to-port="0" /> <edge from-layer="332" from-port="2" to-layer="348" to-port="0" /> <edge from-layer="333" from-port="0" to-layer="336" to-port="1" /> <edge from-layer="334" from-port="0" to-layer="336" to-port="2" /> <edge from-layer="335" from-port="0" to-layer="336" to-port="3" /> <edge from-layer="336" from-port="4" to-layer="338" to-port="0" /> <edge from-layer="337" from-port="0" to-layer="338" to-port="1" /> <edge from-layer="338" from-port="2" to-layer="342" to-port="0" /> <edge from-layer="338" from-port="2" to-layer="407" to-port="0" /> <edge from-layer="339" from-port="0" to-layer="342" to-port="1" /> <edge from-layer="340" from-port="0" to-layer="342" to-port="2" /> <edge from-layer="341" from-port="0" to-layer="342" to-port="3" /> <edge from-layer="342" from-port="4" to-layer="388" to-port="0" /> <edge from-layer="342" from-port="4" to-layer="374" to-port="0" /> <edge from-layer="342" from-port="4" to-layer="377" to-port="0" /> <edge from-layer="342" from-port="4" to-layer="393" to-port="0" /> <edge from-layer="343" from-port="0" to-layer="370" to-port="0" /> <edge from-layer="344" from-port="0" to-layer="370" to-port="1" /> <edge from-layer="345" from-port="0" to-layer="348" to-port="1" /> <edge from-layer="346" from-port="0" to-layer="348" to-port="2" /> <edge from-layer="347" from-port="0" to-layer="348" to-port="3" /> <edge from-layer="348" from-port="4" to-layer="350" to-port="0" /> <edge from-layer="349" from-port="0" to-layer="350" to-port="1" /> <edge from-layer="350" from-port="2" to-layer="412" to-port="0" /> <edge from-layer="350" from-port="2" to-layer="439" to-port="0" /> <edge from-layer="350" from-port="2" to-layer="351" to-port="0" /> <edge from-layer="351" from-port="1" to-layer="353" to-port="0" /> <edge from-layer="352" from-port="0" to-layer="353" to-port="2" /> <edge from-layer="353" from-port="3" to-layer="366" to-port="0" /> <edge from-layer="354" from-port="0" to-layer="359" to-port="0" /> <edge from-layer="355" from-port="0" to-layer="358" to-port="1" /> <edge from-layer="356" from-port="0" to-layer="358" to-port="2" /> <edge from-layer="357" from-port="0" to-layer="358" to-port="3" /> <edge from-layer="358" from-port="4" to-layer="359" to-port="1" /> <edge from-layer="359" from-port="2" to-layer="360" to-port="0" /> <edge from-layer="360" from-port="1" to-layer="362" to-port="0" /> <edge from-layer="361" from-port="0" to-layer="362" to-port="2" /> <edge from-layer="362" from-port="3" to-layer="441" to-port="0" /> <edge from-layer="362" from-port="3" to-layer="363" to-port="0" /> <edge from-layer="363" from-port="1" to-layer="365" to-port="0" /> <edge from-layer="364" from-port="0" to-layer="365" to-port="2" /> <edge from-layer="365" from-port="3" to-layer="366" to-port="1" /> <edge from-layer="366" from-port="2" to-layer="368" to-port="0" /> <edge from-layer="367" from-port="0" to-layer="368" to-port="1" /> <edge from-layer="368" from-port="2" to-layer="398" to-port="2" /> <edge from-layer="368" from-port="2" to-layer="370" to-port="2" /> <edge from-layer="369" from-port="0" to-layer="370" to-port="3" /> <edge from-layer="370" from-port="4" to-layer="372" to-port="0" /> <edge from-layer="371" from-port="0" to-layer="372" to-port="2" /> <edge from-layer="372" from-port="3" to-layer="373" to-port="0" /> <edge from-layer="373" from-port="2" to-layer="413" to-port="1" /> <edge from-layer="373" from-port="2" to-layer="374" to-port="1" /> <edge from-layer="374" from-port="2" to-layer="403" to-port="0" /> <edge from-layer="375" from-port="0" to-layer="385" to-port="0" /> <edge from-layer="376" from-port="0" to-layer="385" to-port="1" /> <edge from-layer="377" from-port="1" to-layer="379" to-port="0" /> <edge from-layer="378" from-port="0" to-layer="379" to-port="2" /> <edge from-layer="379" from-port="3" to-layer="380" to-port="0" /> <edge from-layer="380" from-port="2" to-layer="381" to-port="0" /> <edge from-layer="381" from-port="1" to-layer="383" to-port="0" /> <edge from-layer="382" from-port="0" to-layer="383" to-port="1" /> <edge from-layer="383" from-port="2" to-layer="385" to-port="2" /> <edge from-layer="384" from-port="0" to-layer="385" to-port="3" /> <edge from-layer="385" from-port="4" to-layer="393" to-port="2" /> <edge from-layer="385" from-port="4" to-layer="388" to-port="1" /> <edge from-layer="386" from-port="0" to-layer="388" to-port="2" /> <edge from-layer="387" from-port="0" to-layer="388" to-port="3" /> <edge from-layer="388" from-port="4" to-layer="390" to-port="0" /> <edge from-layer="389" from-port="0" to-layer="390" to-port="1" /> <edge from-layer="390" from-port="2" to-layer="394" to-port="0" /> <edge from-layer="391" from-port="0" to-layer="393" to-port="1" /> <edge from-layer="392" from-port="0" to-layer="393" to-port="3" /> <edge from-layer="393" from-port="4" to-layer="394" to-port="1" /> <edge from-layer="394" from-port="2" to-layer="402" to-port="0" /> <edge from-layer="395" from-port="0" to-layer="398" to-port="0" /> <edge from-layer="396" from-port="0" to-layer="398" to-port="1" /> <edge from-layer="397" from-port="0" to-layer="398" to-port="3" /> <edge from-layer="398" from-port="4" to-layer="400" to-port="0" /> <edge from-layer="399" from-port="0" to-layer="400" to-port="2" /> <edge from-layer="400" from-port="3" to-layer="401" to-port="0" /> <edge from-layer="401" from-port="2" to-layer="434" to-port="1" /> <edge from-layer="401" from-port="2" to-layer="402" to-port="1" /> <edge from-layer="402" from-port="2" to-layer="403" to-port="1" /> <edge from-layer="403" from-port="2" to-layer="408" to-port="0" /> <edge from-layer="404" from-port="0" to-layer="407" to-port="1" /> <edge from-layer="405" from-port="0" to-layer="407" to-port="2" /> <edge from-layer="406" from-port="0" to-layer="407" to-port="3" /> <edge from-layer="407" from-port="4" to-layer="408" to-port="1" /> <edge from-layer="408" from-port="2" to-layer="464" to-port="0" /> <edge from-layer="408" from-port="2" to-layer="502" to-port="0" /> <edge from-layer="409" from-port="0" to-layer="412" to-port="1" /> <edge from-layer="410" from-port="0" to-layer="412" to-port="2" /> <edge from-layer="411" from-port="0" to-layer="412" to-port="3" /> <edge from-layer="412" from-port="4" to-layer="416" to-port="0" /> <edge from-layer="412" from-port="4" to-layer="427" to-port="0" /> <edge from-layer="412" from-port="4" to-layer="413" to-port="0" /> <edge from-layer="412" from-port="4" to-layer="432" to-port="0" /> <edge from-layer="413" from-port="2" to-layer="435" to-port="0" /> <edge from-layer="414" from-port="0" to-layer="424" to-port="0" /> <edge from-layer="415" from-port="0" to-layer="424" to-port="1" /> <edge from-layer="416" from-port="1" to-layer="418" to-port="0" /> <edge from-layer="417" from-port="0" to-layer="418" to-port="2" /> <edge from-layer="418" from-port="3" to-layer="419" to-port="0" /> <edge from-layer="419" from-port="2" to-layer="420" to-port="0" /> <edge from-layer="420" from-port="1" to-layer="422" to-port="0" /> <edge from-layer="421" from-port="0" to-layer="422" to-port="1" /> <edge from-layer="422" from-port="2" to-layer="424" to-port="2" /> <edge from-layer="423" from-port="0" to-layer="424" to-port="3" /> <edge from-layer="424" from-port="4" to-layer="432" to-port="2" /> <edge from-layer="424" from-port="4" to-layer="427" to-port="1" /> <edge from-layer="425" from-port="0" to-layer="427" to-port="2" /> <edge from-layer="426" from-port="0" to-layer="427" to-port="3" /> <edge from-layer="427" from-port="4" to-layer="429" to-port="0" /> <edge from-layer="428" from-port="0" to-layer="429" to-port="1" /> <edge from-layer="429" from-port="2" to-layer="433" to-port="0" /> <edge from-layer="430" from-port="0" to-layer="432" to-port="1" /> <edge from-layer="431" from-port="0" to-layer="432" to-port="3" /> <edge from-layer="432" from-port="4" to-layer="433" to-port="1" /> <edge from-layer="433" from-port="2" to-layer="434" to-port="0" /> <edge from-layer="434" from-port="2" to-layer="435" to-port="1" /> <edge from-layer="435" from-port="2" to-layer="440" to-port="0" /> <edge from-layer="436" from-port="0" to-layer="439" to-port="1" /> <edge from-layer="437" from-port="0" to-layer="439" to-port="2" /> <edge from-layer="438" from-port="0" to-layer="439" to-port="3" /> <edge from-layer="439" from-port="4" to-layer="440" to-port="1" /> <edge from-layer="440" from-port="2" to-layer="441" to-port="1" /> <edge from-layer="441" from-port="2" to-layer="461" to-port="0" /> <edge from-layer="441" from-port="2" to-layer="502" to-port="1" /> <edge from-layer="441" from-port="2" to-layer="10171" to-port="0" /> <edge from-layer="442" from-port="0" to-layer="447" to-port="0" /> <edge from-layer="443" from-port="0" to-layer="446" to-port="1" /> <edge from-layer="444" from-port="0" to-layer="446" to-port="2" /> <edge from-layer="445" from-port="0" to-layer="446" to-port="3" /> <edge from-layer="446" from-port="4" to-layer="447" to-port="1" /> <edge from-layer="447" from-port="2" to-layer="448" to-port="0" /> <edge from-layer="448" from-port="1" to-layer="450" to-port="0" /> <edge from-layer="449" from-port="0" to-layer="450" to-port="2" /> <edge from-layer="450" from-port="3" to-layer="457" to-port="0" /> <edge from-layer="451" from-port="0" to-layer="454" to-port="1" /> <edge from-layer="452" from-port="0" to-layer="454" to-port="2" /> <edge from-layer="453" from-port="0" to-layer="454" to-port="3" /> <edge from-layer="454" from-port="4" to-layer="456" to-port="0" /> <edge from-layer="455" from-port="0" to-layer="456" to-port="1" /> <edge from-layer="456" from-port="2" to-layer="457" to-port="1" /> <edge from-layer="457" from-port="2" to-layer="502" to-port="2" /> <edge from-layer="457" from-port="2" to-layer="10170" to-port="0" /> <edge from-layer="458" from-port="0" to-layer="476" to-port="0" /> <edge from-layer="459" from-port="0" to-layer="471" to-port="0" /> <edge from-layer="460" from-port="0" to-layer="474" to-port="1" /> <edge from-layer="460" from-port="0" to-layer="471" to-port="1" /> <edge from-layer="461" from-port="1" to-layer="463" to-port="0" /> <edge from-layer="462" from-port="0" to-layer="463" to-port="2" /> <edge from-layer="463" from-port="3" to-layer="467" to-port="0" /> <edge from-layer="463" from-port="3" to-layer="473" to-port="0" /> <edge from-layer="464" from-port="1" to-layer="491" to-port="0" /> <edge from-layer="464" from-port="1" to-layer="466" to-port="0" /> <edge from-layer="465" from-port="0" to-layer="466" to-port="2" /> <edge from-layer="466" from-port="3" to-layer="467" to-port="1" /> <edge from-layer="467" from-port="2" to-layer="469" to-port="0" /> <edge from-layer="468" from-port="0" to-layer="473" to-port="1" /> <edge from-layer="468" from-port="0" to-layer="469" to-port="1" /> <edge from-layer="469" from-port="2" to-layer="471" to-port="2" /> <edge from-layer="470" from-port="0" to-layer="471" to-port="3" /> <edge from-layer="470" from-port="0" to-layer="474" to-port="3" /> <edge from-layer="471" from-port="4" to-layer="476" to-port="1" /> <edge from-layer="472" from-port="0" to-layer="474" to-port="0" /> <edge from-layer="473" from-port="2" to-layer="474" to-port="2" /> <edge from-layer="473" from-port="2" to-layer="481" to-port="2" /> <edge from-layer="474" from-port="4" to-layer="476" to-port="2" /> <edge from-layer="475" from-port="0" to-layer="476" to-port="3" /> <edge from-layer="476" from-port="4" to-layer="483" to-port="0" /> <edge from-layer="477" from-port="0" to-layer="483" to-port="1" /> <edge from-layer="478" from-port="0" to-layer="481" to-port="0" /> <edge from-layer="479" from-port="0" to-layer="481" to-port="1" /> <edge from-layer="480" from-port="0" to-layer="481" to-port="3" /> <edge from-layer="481" from-port="4" to-layer="483" to-port="2" /> <edge from-layer="482" from-port="0" to-layer="483" to-port="3" /> <edge from-layer="483" from-port="4" to-layer="486" to-port="0" /> <edge from-layer="484" from-port="0" to-layer="486" to-port="1" /> <edge from-layer="485" from-port="0" to-layer="486" to-port="2" /> <edge from-layer="486" from-port="3" to-layer="487" to-port="0" /> <edge from-layer="487" from-port="2" to-layer="501" to-port="0" /> <edge from-layer="487" from-port="2" to-layer="488" to-port="0" /> <edge from-layer="488" from-port="1" to-layer="500" to-port="0" /> <edge from-layer="489" from-port="0" to-layer="491" to-port="1" /> <edge from-layer="490" from-port="0" to-layer="491" to-port="2" /> <edge from-layer="491" from-port="3" to-layer="495" to-port="0" /> <edge from-layer="492" from-port="0" to-layer="495" to-port="1" /> <edge from-layer="493" from-port="0" to-layer="495" to-port="2" /> <edge from-layer="494" from-port="0" to-layer="495" to-port="3" /> <edge from-layer="495" from-port="4" to-layer="497" to-port="0" /> <edge from-layer="495" from-port="4" to-layer="499" to-port="2" /> <edge from-layer="496" from-port="0" to-layer="497" to-port="1" /> <edge from-layer="497" from-port="2" to-layer="499" to-port="0" /> <edge from-layer="498" from-port="0" to-layer="499" to-port="1" /> <edge from-layer="499" from-port="3" to-layer="500" to-port="1" /> <edge from-layer="500" from-port="2" to-layer="501" to-port="1" /> <edge from-layer="501" from-port="2" to-layer="502" to-port="3" /> <edge from-layer="502" from-port="4" to-layer="504" to-port="0" /> <edge from-layer="503" from-port="0" to-layer="504" to-port="1" /> <edge from-layer="504" from-port="2" to-layer="506" to-port="0" /> <edge from-layer="505" from-port="0" to-layer="506" to-port="1" /> <edge from-layer="506" from-port="2" to-layer="514" to-port="0" /> <edge from-layer="507" from-port="0" to-layer="508" to-port="0" /> <edge from-layer="508" from-port="1" to-layer="511" to-port="0" /> <edge from-layer="509" from-port="0" to-layer="510" to-port="0" /> <edge from-layer="510" from-port="1" to-layer="511" to-port="1" /> <edge from-layer="511" from-port="2" to-layer="513" to-port="0" /> <edge from-layer="512" from-port="0" to-layer="513" to-port="1" /> <edge from-layer="513" from-port="2" to-layer="514" to-port="1" /> <edge from-layer="514" from-port="2" to-layer="516" to-port="0" /> <edge from-layer="515" from-port="0" to-layer="516" to-port="1" /> <edge from-layer="516" from-port="2" to-layer="517" to-port="1" /> <edge from-layer="517" from-port="2" to-layer="518" to-port="0" /> <edge from-layer="518" from-port="2" to-layer="547" to-port="0" /> <edge from-layer="518" from-port="2" to-layer="749" to-port="1" /> <edge from-layer="518" from-port="2" to-layer="520" to-port="0" /> <edge from-layer="519" from-port="0" to-layer="520" to-port="1" /> <edge from-layer="520" from-port="2" to-layer="522" to-port="0" /> <edge from-layer="521" from-port="0" to-layer="522" to-port="1" /> <edge from-layer="522" from-port="2" to-layer="524" to-port="0" /> <edge from-layer="523" from-port="0" to-layer="524" to-port="1" /> <edge from-layer="524" from-port="2" to-layer="532" to-port="0" /> <edge from-layer="525" from-port="0" to-layer="526" to-port="0" /> <edge from-layer="526" from-port="1" to-layer="529" to-port="0" /> <edge from-layer="527" from-port="0" to-layer="528" to-port="0" /> <edge from-layer="528" from-port="1" to-layer="529" to-port="1" /> <edge from-layer="529" from-port="2" to-layer="531" to-port="0" /> <edge from-layer="530" from-port="0" to-layer="531" to-port="1" /> <edge from-layer="531" from-port="2" to-layer="532" to-port="1" /> <edge from-layer="532" from-port="2" to-layer="534" to-port="0" /> <edge from-layer="533" from-port="0" to-layer="534" to-port="1" /> <edge from-layer="534" from-port="2" to-layer="535" to-port="0" /> <edge from-layer="535" from-port="1" to-layer="543" to-port="0" /> <edge from-layer="536" from-port="0" to-layer="537" to-port="0" /> <edge from-layer="537" from-port="1" to-layer="540" to-port="0" /> <edge from-layer="538" from-port="0" to-layer="539" to-port="0" /> <edge from-layer="539" from-port="1" to-layer="540" to-port="1" /> <edge from-layer="540" from-port="2" to-layer="542" to-port="0" /> <edge from-layer="541" from-port="0" to-layer="542" to-port="1" /> <edge from-layer="542" from-port="2" to-layer="543" to-port="1" /> <edge from-layer="543" from-port="2" to-layer="545" to-port="0" /> <edge from-layer="544" from-port="0" to-layer="545" to-port="1" /> <edge from-layer="545" from-port="2" to-layer="748" to-port="0" /> <edge from-layer="546" from-port="0" to-layer="547" to-port="1" /> <edge from-layer="547" from-port="2" to-layer="549" to-port="0" /> <edge from-layer="548" from-port="0" to-layer="549" to-port="1" /> <edge from-layer="549" from-port="2" to-layer="551" to-port="0" /> <edge from-layer="550" from-port="0" to-layer="551" to-port="1" /> <edge from-layer="551" from-port="2" to-layer="559" to-port="0" /> <edge from-layer="552" from-port="0" to-layer="553" to-port="0" /> <edge from-layer="553" from-port="1" to-layer="556" to-port="0" /> <edge from-layer="554" from-port="0" to-layer="555" to-port="0" /> <edge from-layer="555" from-port="1" to-layer="556" to-port="1" /> <edge from-layer="556" from-port="2" to-layer="558" to-port="0" /> <edge from-layer="557" from-port="0" to-layer="558" to-port="1" /> <edge from-layer="558" from-port="2" to-layer="559" to-port="1" /> <edge from-layer="559" from-port="2" to-layer="561" to-port="0" /> <edge from-layer="560" from-port="0" to-layer="561" to-port="1" /> <edge from-layer="561" from-port="2" to-layer="563" to-port="0" /> <edge from-layer="562" from-port="0" to-layer="563" to-port="1" /> <edge from-layer="563" from-port="2" to-layer="567" to-port="0" /> <edge from-layer="563" from-port="2" to-layer="685" to-port="0" /> <edge from-layer="563" from-port="2" to-layer="579" to-port="0" /> <edge from-layer="564" from-port="0" to-layer="567" to-port="1" /> <edge from-layer="565" from-port="0" to-layer="567" to-port="2" /> <edge from-layer="566" from-port="0" to-layer="567" to-port="3" /> <edge from-layer="567" from-port="4" to-layer="569" to-port="0" /> <edge from-layer="568" from-port="0" to-layer="569" to-port="1" /> <edge from-layer="569" from-port="2" to-layer="573" to-port="0" /> <edge from-layer="569" from-port="2" to-layer="638" to-port="0" /> <edge from-layer="570" from-port="0" to-layer="573" to-port="1" /> <edge from-layer="571" from-port="0" to-layer="573" to-port="2" /> <edge from-layer="572" from-port="0" to-layer="573" to-port="3" /> <edge from-layer="573" from-port="4" to-layer="624" to-port="0" /> <edge from-layer="573" from-port="4" to-layer="605" to-port="0" /> <edge from-layer="573" from-port="4" to-layer="608" to-port="0" /> <edge from-layer="573" from-port="4" to-layer="619" to-port="0" /> <edge from-layer="574" from-port="0" to-layer="601" to-port="0" /> <edge from-layer="575" from-port="0" to-layer="601" to-port="1" /> <edge from-layer="576" from-port="0" to-layer="579" to-port="1" /> <edge from-layer="577" from-port="0" to-layer="579" to-port="2" /> <edge from-layer="578" from-port="0" to-layer="579" to-port="3" /> <edge from-layer="579" from-port="4" to-layer="581" to-port="0" /> <edge from-layer="580" from-port="0" to-layer="581" to-port="1" /> <edge from-layer="581" from-port="2" to-layer="643" to-port="0" /> <edge from-layer="581" from-port="2" to-layer="582" to-port="0" /> <edge from-layer="581" from-port="2" to-layer="670" to-port="0" /> <edge from-layer="582" from-port="1" to-layer="584" to-port="0" /> <edge from-layer="583" from-port="0" to-layer="584" to-port="2" /> <edge from-layer="584" from-port="3" to-layer="597" to-port="0" /> <edge from-layer="585" from-port="0" to-layer="590" to-port="0" /> <edge from-layer="586" from-port="0" to-layer="589" to-port="1" /> <edge from-layer="587" from-port="0" to-layer="589" to-port="2" /> <edge from-layer="588" from-port="0" to-layer="589" to-port="3" /> <edge from-layer="589" from-port="4" to-layer="590" to-port="1" /> <edge from-layer="590" from-port="2" to-layer="591" to-port="0" /> <edge from-layer="591" from-port="1" to-layer="593" to-port="0" /> <edge from-layer="592" from-port="0" to-layer="593" to-port="2" /> <edge from-layer="593" from-port="3" to-layer="672" to-port="0" /> <edge from-layer="593" from-port="3" to-layer="594" to-port="0" /> <edge from-layer="594" from-port="1" to-layer="596" to-port="0" /> <edge from-layer="595" from-port="0" to-layer="596" to-port="2" /> <edge from-layer="596" from-port="3" to-layer="597" to-port="1" /> <edge from-layer="597" from-port="2" to-layer="599" to-port="0" /> <edge from-layer="598" from-port="0" to-layer="599" to-port="1" /> <edge from-layer="599" from-port="2" to-layer="601" to-port="2" /> <edge from-layer="599" from-port="2" to-layer="629" to-port="2" /> <edge from-layer="600" from-port="0" to-layer="601" to-port="3" /> <edge from-layer="601" from-port="4" to-layer="603" to-port="0" /> <edge from-layer="602" from-port="0" to-layer="603" to-port="2" /> <edge from-layer="603" from-port="3" to-layer="604" to-port="0" /> <edge from-layer="604" from-port="2" to-layer="644" to-port="1" /> <edge from-layer="604" from-port="2" to-layer="605" to-port="1" /> <edge from-layer="605" from-port="2" to-layer="634" to-port="0" /> <edge from-layer="606" from-port="0" to-layer="616" to-port="0" /> <edge from-layer="607" from-port="0" to-layer="616" to-port="1" /> <edge from-layer="608" from-port="1" to-layer="610" to-port="0" /> <edge from-layer="609" from-port="0" to-layer="610" to-port="2" /> <edge from-layer="610" from-port="3" to-layer="611" to-port="0" /> <edge from-layer="611" from-port="2" to-layer="612" to-port="0" /> <edge from-layer="612" from-port="1" to-layer="614" to-port="0" /> <edge from-layer="613" from-port="0" to-layer="614" to-port="1" /> <edge from-layer="614" from-port="2" to-layer="616" to-port="2" /> <edge from-layer="615" from-port="0" to-layer="616" to-port="3" /> <edge from-layer="616" from-port="4" to-layer="619" to-port="1" /> <edge from-layer="616" from-port="4" to-layer="624" to-port="2" /> <edge from-layer="617" from-port="0" to-layer="619" to-port="2" /> <edge from-layer="618" from-port="0" to-layer="619" to-port="3" /> <edge from-layer="619" from-port="4" to-layer="621" to-port="0" /> <edge from-layer="620" from-port="0" to-layer="621" to-port="1" /> <edge from-layer="621" from-port="2" to-layer="625" to-port="0" /> <edge from-layer="622" from-port="0" to-layer="624" to-port="1" /> <edge from-layer="623" from-port="0" to-layer="624" to-port="3" /> <edge from-layer="624" from-port="4" to-layer="625" to-port="1" /> <edge from-layer="625" from-port="2" to-layer="633" to-port="0" /> <edge from-layer="626" from-port="0" to-layer="629" to-port="0" /> <edge from-layer="627" from-port="0" to-layer="629" to-port="1" /> <edge from-layer="628" from-port="0" to-layer="629" to-port="3" /> <edge from-layer="629" from-port="4" to-layer="631" to-port="0" /> <edge from-layer="630" from-port="0" to-layer="631" to-port="2" /> <edge from-layer="631" from-port="3" to-layer="632" to-port="0" /> <edge from-layer="632" from-port="2" to-layer="665" to-port="1" /> <edge from-layer="632" from-port="2" to-layer="633" to-port="1" /> <edge from-layer="633" from-port="2" to-layer="634" to-port="1" /> <edge from-layer="634" from-port="2" to-layer="639" to-port="0" /> <edge from-layer="635" from-port="0" to-layer="638" to-port="1" /> <edge from-layer="636" from-port="0" to-layer="638" to-port="2" /> <edge from-layer="637" from-port="0" to-layer="638" to-port="3" /> <edge from-layer="638" from-port="4" to-layer="639" to-port="1" /> <edge from-layer="639" from-port="2" to-layer="695" to-port="0" /> <edge from-layer="639" from-port="2" to-layer="733" to-port="0" /> <edge from-layer="640" from-port="0" to-layer="643" to-port="1" /> <edge from-layer="641" from-port="0" to-layer="643" to-port="2" /> <edge from-layer="642" from-port="0" to-layer="643" to-port="3" /> <edge from-layer="643" from-port="4" to-layer="644" to-port="0" /> <edge from-layer="643" from-port="4" to-layer="647" to-port="0" /> <edge from-layer="643" from-port="4" to-layer="658" to-port="0" /> <edge from-layer="643" from-port="4" to-layer="663" to-port="0" /> <edge from-layer="644" from-port="2" to-layer="666" to-port="0" /> <edge from-layer="645" from-port="0" to-layer="655" to-port="0" /> <edge from-layer="646" from-port="0" to-layer="655" to-port="1" /> <edge from-layer="647" from-port="1" to-layer="649" to-port="0" /> <edge from-layer="648" from-port="0" to-layer="649" to-port="2" /> <edge from-layer="649" from-port="3" to-layer="650" to-port="0" /> <edge from-layer="650" from-port="2" to-layer="651" to-port="0" /> <edge from-layer="651" from-port="1" to-layer="653" to-port="0" /> <edge from-layer="652" from-port="0" to-layer="653" to-port="1" /> <edge from-layer="653" from-port="2" to-layer="655" to-port="2" /> <edge from-layer="654" from-port="0" to-layer="655" to-port="3" /> <edge from-layer="655" from-port="4" to-layer="658" to-port="1" /> <edge from-layer="655" from-port="4" to-layer="663" to-port="2" /> <edge from-layer="656" from-port="0" to-layer="658" to-port="2" /> <edge from-layer="657" from-port="0" to-layer="658" to-port="3" /> <edge from-layer="658" from-port="4" to-layer="660" to-port="0" /> <edge from-layer="659" from-port="0" to-layer="660" to-port="1" /> <edge from-layer="660" from-port="2" to-layer="664" to-port="0" /> <edge from-layer="661" from-port="0" to-layer="663" to-port="1" /> <edge from-layer="662" from-port="0" to-layer="663" to-port="3" /> <edge from-layer="663" from-port="4" to-layer="664" to-port="1" /> <edge from-layer="664" from-port="2" to-layer="665" to-port="0" /> <edge from-layer="665" from-port="2" to-layer="666" to-port="1" /> <edge from-layer="666" from-port="2" to-layer="671" to-port="0" /> <edge from-layer="667" from-port="0" to-layer="670" to-port="1" /> <edge from-layer="668" from-port="0" to-layer="670" to-port="2" /> <edge from-layer="669" from-port="0" to-layer="670" to-port="3" /> <edge from-layer="670" from-port="4" to-layer="671" to-port="1" /> <edge from-layer="671" from-port="2" to-layer="672" to-port="1" /> <edge from-layer="672" from-port="2" to-layer="733" to-port="1" /> <edge from-layer="672" from-port="2" to-layer="692" to-port="0" /> <edge from-layer="672" from-port="2" to-layer="10149" to-port="0" /> <edge from-layer="673" from-port="0" to-layer="678" to-port="0" /> <edge from-layer="674" from-port="0" to-layer="677" to-port="1" /> <edge from-layer="675" from-port="0" to-layer="677" to-port="2" /> <edge from-layer="676" from-port="0" to-layer="677" to-port="3" /> <edge from-layer="677" from-port="4" to-layer="678" to-port="1" /> <edge from-layer="678" from-port="2" to-layer="679" to-port="0" /> <edge from-layer="679" from-port="1" to-layer="681" to-port="0" /> <edge from-layer="680" from-port="0" to-layer="681" to-port="2" /> <edge from-layer="681" from-port="3" to-layer="688" to-port="0" /> <edge from-layer="682" from-port="0" to-layer="685" to-port="1" /> <edge from-layer="683" from-port="0" to-layer="685" to-port="2" /> <edge from-layer="684" from-port="0" to-layer="685" to-port="3" /> <edge from-layer="685" from-port="4" to-layer="687" to-port="0" /> <edge from-layer="686" from-port="0" to-layer="687" to-port="1" /> <edge from-layer="687" from-port="2" to-layer="688" to-port="1" /> <edge from-layer="688" from-port="2" to-layer="10148" to-port="0" /> <edge from-layer="688" from-port="2" to-layer="733" to-port="2" /> <edge from-layer="689" from-port="0" to-layer="707" to-port="0" /> <edge from-layer="690" from-port="0" to-layer="702" to-port="0" /> <edge from-layer="691" from-port="0" to-layer="705" to-port="1" /> <edge from-layer="691" from-port="0" to-layer="702" to-port="1" /> <edge from-layer="692" from-port="1" to-layer="694" to-port="0" /> <edge from-layer="693" from-port="0" to-layer="694" to-port="2" /> <edge from-layer="694" from-port="3" to-layer="698" to-port="0" /> <edge from-layer="694" from-port="3" to-layer="704" to-port="0" /> <edge from-layer="695" from-port="1" to-layer="722" to-port="0" /> <edge from-layer="695" from-port="1" to-layer="697" to-port="0" /> <edge from-layer="696" from-port="0" to-layer="697" to-port="2" /> <edge from-layer="697" from-port="3" to-layer="698" to-port="1" /> <edge from-layer="698" from-port="2" to-layer="700" to-port="0" /> <edge from-layer="699" from-port="0" to-layer="704" to-port="1" /> <edge from-layer="699" from-port="0" to-layer="700" to-port="1" /> <edge from-layer="700" from-port="2" to-layer="702" to-port="2" /> <edge from-layer="701" from-port="0" to-layer="705" to-port="3" /> <edge from-layer="701" from-port="0" to-layer="702" to-port="3" /> <edge from-layer="702" from-port="4" to-layer="707" to-port="1" /> <edge from-layer="703" from-port="0" to-layer="705" to-port="0" /> <edge from-layer="704" from-port="2" to-layer="705" to-port="2" /> <edge from-layer="704" from-port="2" to-layer="712" to-port="2" /> <edge from-layer="705" from-port="4" to-layer="707" to-port="2" /> <edge from-layer="706" from-port="0" to-layer="707" to-port="3" /> <edge from-layer="707" from-port="4" to-layer="714" to-port="0" /> <edge from-layer="708" from-port="0" to-layer="714" to-port="1" /> <edge from-layer="709" from-port="0" to-layer="712" to-port="0" /> <edge from-layer="710" from-port="0" to-layer="712" to-port="1" /> <edge from-layer="711" from-port="0" to-layer="712" to-port="3" /> <edge from-layer="712" from-port="4" to-layer="714" to-port="2" /> <edge from-layer="713" from-port="0" to-layer="714" to-port="3" /> <edge from-layer="714" from-port="4" to-layer="717" to-port="0" /> <edge from-layer="715" from-port="0" to-layer="717" to-port="1" /> <edge from-layer="716" from-port="0" to-layer="717" to-port="2" /> <edge from-layer="717" from-port="3" to-layer="718" to-port="0" /> <edge from-layer="718" from-port="2" to-layer="732" to-port="0" /> <edge from-layer="718" from-port="2" to-layer="719" to-port="0" /> <edge from-layer="719" from-port="1" to-layer="731" to-port="0" /> <edge from-layer="720" from-port="0" to-layer="722" to-port="1" /> <edge from-layer="721" from-port="0" to-layer="722" to-port="2" /> <edge from-layer="722" from-port="3" to-layer="726" to-port="0" /> <edge from-layer="723" from-port="0" to-layer="726" to-port="1" /> <edge from-layer="724" from-port="0" to-layer="726" to-port="2" /> <edge from-layer="725" from-port="0" to-layer="726" to-port="3" /> <edge from-layer="726" from-port="4" to-layer="728" to-port="0" /> <edge from-layer="726" from-port="4" to-layer="730" to-port="2" /> <edge from-layer="727" from-port="0" to-layer="728" to-port="1" /> <edge from-layer="728" from-port="2" to-layer="730" to-port="0" /> <edge from-layer="729" from-port="0" to-layer="730" to-port="1" /> <edge from-layer="730" from-port="3" to-layer="731" to-port="1" /> <edge from-layer="731" from-port="2" to-layer="732" to-port="1" /> <edge from-layer="732" from-port="2" to-layer="733" to-port="3" /> <edge from-layer="733" from-port="4" to-layer="735" to-port="0" /> <edge from-layer="734" from-port="0" to-layer="735" to-port="1" /> <edge from-layer="735" from-port="2" to-layer="737" to-port="0" /> <edge from-layer="736" from-port="0" to-layer="737" to-port="1" /> <edge from-layer="737" from-port="2" to-layer="745" to-port="0" /> <edge from-layer="738" from-port="0" to-layer="739" to-port="0" /> <edge from-layer="739" from-port="1" to-layer="742" to-port="0" /> <edge from-layer="740" from-port="0" to-layer="741" to-port="0" /> <edge from-layer="741" from-port="1" to-layer="742" to-port="1" /> <edge from-layer="742" from-port="2" to-layer="744" to-port="0" /> <edge from-layer="743" from-port="0" to-layer="744" to-port="1" /> <edge from-layer="744" from-port="2" to-layer="745" to-port="1" /> <edge from-layer="745" from-port="2" to-layer="747" to-port="0" /> <edge from-layer="746" from-port="0" to-layer="747" to-port="1" /> <edge from-layer="747" from-port="2" to-layer="748" to-port="1" /> <edge from-layer="748" from-port="2" to-layer="749" to-port="0" /> <edge from-layer="749" from-port="2" to-layer="778" to-port="0" /> <edge from-layer="749" from-port="2" to-layer="980" to-port="1" /> <edge from-layer="749" from-port="2" to-layer="751" to-port="0" /> <edge from-layer="750" from-port="0" to-layer="751" to-port="1" /> <edge from-layer="751" from-port="2" to-layer="753" to-port="0" /> <edge from-layer="752" from-port="0" to-layer="753" to-port="1" /> <edge from-layer="753" from-port="2" to-layer="755" to-port="0" /> <edge from-layer="754" from-port="0" to-layer="755" to-port="1" /> <edge from-layer="755" from-port="2" to-layer="763" to-port="0" /> <edge from-layer="756" from-port="0" to-layer="757" to-port="0" /> <edge from-layer="757" from-port="1" to-layer="760" to-port="0" /> <edge from-layer="758" from-port="0" to-layer="759" to-port="0" /> <edge from-layer="759" from-port="1" to-layer="760" to-port="1" /> <edge from-layer="760" from-port="2" to-layer="762" to-port="0" /> <edge from-layer="761" from-port="0" to-layer="762" to-port="1" /> <edge from-layer="762" from-port="2" to-layer="763" to-port="1" /> <edge from-layer="763" from-port="2" to-layer="765" to-port="0" /> <edge from-layer="764" from-port="0" to-layer="765" to-port="1" /> <edge from-layer="765" from-port="2" to-layer="766" to-port="0" /> <edge from-layer="766" from-port="1" to-layer="774" to-port="0" /> <edge from-layer="767" from-port="0" to-layer="768" to-port="0" /> <edge from-layer="768" from-port="1" to-layer="771" to-port="0" /> <edge from-layer="769" from-port="0" to-layer="770" to-port="0" /> <edge from-layer="770" from-port="1" to-layer="771" to-port="1" /> <edge from-layer="771" from-port="2" to-layer="773" to-port="0" /> <edge from-layer="772" from-port="0" to-layer="773" to-port="1" /> <edge from-layer="773" from-port="2" to-layer="774" to-port="1" /> <edge from-layer="774" from-port="2" to-layer="776" to-port="0" /> <edge from-layer="775" from-port="0" to-layer="776" to-port="1" /> <edge from-layer="776" from-port="2" to-layer="979" to-port="0" /> <edge from-layer="777" from-port="0" to-layer="778" to-port="1" /> <edge from-layer="778" from-port="2" to-layer="780" to-port="0" /> <edge from-layer="779" from-port="0" to-layer="780" to-port="1" /> <edge from-layer="780" from-port="2" to-layer="782" to-port="0" /> <edge from-layer="781" from-port="0" to-layer="782" to-port="1" /> <edge from-layer="782" from-port="2" to-layer="790" to-port="0" /> <edge from-layer="783" from-port="0" to-layer="784" to-port="0" /> <edge from-layer="784" from-port="1" to-layer="787" to-port="0" /> <edge from-layer="785" from-port="0" to-layer="786" to-port="0" /> <edge from-layer="786" from-port="1" to-layer="787" to-port="1" /> <edge from-layer="787" from-port="2" to-layer="789" to-port="0" /> <edge from-layer="788" from-port="0" to-layer="789" to-port="1" /> <edge from-layer="789" from-port="2" to-layer="790" to-port="1" /> <edge from-layer="790" from-port="2" to-layer="792" to-port="0" /> <edge from-layer="791" from-port="0" to-layer="792" to-port="1" /> <edge from-layer="792" from-port="2" to-layer="794" to-port="0" /> <edge from-layer="793" from-port="0" to-layer="794" to-port="1" /> <edge from-layer="794" from-port="2" to-layer="810" to-port="0" /> <edge from-layer="794" from-port="2" to-layer="798" to-port="0" /> <edge from-layer="794" from-port="2" to-layer="916" to-port="0" /> <edge from-layer="795" from-port="0" to-layer="798" to-port="1" /> <edge from-layer="796" from-port="0" to-layer="798" to-port="2" /> <edge from-layer="797" from-port="0" to-layer="798" to-port="3" /> <edge from-layer="798" from-port="4" to-layer="800" to-port="0" /> <edge from-layer="799" from-port="0" to-layer="800" to-port="1" /> <edge from-layer="800" from-port="2" to-layer="869" to-port="0" /> <edge from-layer="800" from-port="2" to-layer="804" to-port="0" /> <edge from-layer="801" from-port="0" to-layer="804" to-port="1" /> <edge from-layer="802" from-port="0" to-layer="804" to-port="2" /> <edge from-layer="803" from-port="0" to-layer="804" to-port="3" /> <edge from-layer="804" from-port="4" to-layer="839" to-port="0" /> <edge from-layer="804" from-port="4" to-layer="836" to-port="0" /> <edge from-layer="804" from-port="4" to-layer="850" to-port="0" /> <edge from-layer="804" from-port="4" to-layer="855" to-port="0" /> <edge from-layer="805" from-port="0" to-layer="832" to-port="0" /> <edge from-layer="806" from-port="0" to-layer="832" to-port="1" /> <edge from-layer="807" from-port="0" to-layer="810" to-port="1" /> <edge from-layer="808" from-port="0" to-layer="810" to-port="2" /> <edge from-layer="809" from-port="0" to-layer="810" to-port="3" /> <edge from-layer="810" from-port="4" to-layer="812" to-port="0" /> <edge from-layer="811" from-port="0" to-layer="812" to-port="1" /> <edge from-layer="812" from-port="2" to-layer="813" to-port="0" /> <edge from-layer="812" from-port="2" to-layer="874" to-port="0" /> <edge from-layer="812" from-port="2" to-layer="901" to-port="0" /> <edge from-layer="813" from-port="1" to-layer="815" to-port="0" /> <edge from-layer="814" from-port="0" to-layer="815" to-port="2" /> <edge from-layer="815" from-port="3" to-layer="828" to-port="0" /> <edge from-layer="816" from-port="0" to-layer="821" to-port="0" /> <edge from-layer="817" from-port="0" to-layer="820" to-port="1" /> <edge from-layer="818" from-port="0" to-layer="820" to-port="2" /> <edge from-layer="819" from-port="0" to-layer="820" to-port="3" /> <edge from-layer="820" from-port="4" to-layer="821" to-port="1" /> <edge from-layer="821" from-port="2" to-layer="822" to-port="0" /> <edge from-layer="822" from-port="1" to-layer="824" to-port="0" /> <edge from-layer="823" from-port="0" to-layer="824" to-port="2" /> <edge from-layer="824" from-port="3" to-layer="825" to-port="0" /> <edge from-layer="824" from-port="3" to-layer="903" to-port="0" /> <edge from-layer="825" from-port="1" to-layer="827" to-port="0" /> <edge from-layer="826" from-port="0" to-layer="827" to-port="2" /> <edge from-layer="827" from-port="3" to-layer="828" to-port="1" /> <edge from-layer="828" from-port="2" to-layer="830" to-port="0" /> <edge from-layer="829" from-port="0" to-layer="830" to-port="1" /> <edge from-layer="830" from-port="2" to-layer="832" to-port="2" /> <edge from-layer="830" from-port="2" to-layer="860" to-port="2" /> <edge from-layer="831" from-port="0" to-layer="832" to-port="3" /> <edge from-layer="832" from-port="4" to-layer="834" to-port="0" /> <edge from-layer="833" from-port="0" to-layer="834" to-port="2" /> <edge from-layer="834" from-port="3" to-layer="835" to-port="0" /> <edge from-layer="835" from-port="2" to-layer="875" to-port="1" /> <edge from-layer="835" from-port="2" to-layer="836" to-port="1" /> <edge from-layer="836" from-port="2" to-layer="865" to-port="0" /> <edge from-layer="837" from-port="0" to-layer="847" to-port="0" /> <edge from-layer="838" from-port="0" to-layer="847" to-port="1" /> <edge from-layer="839" from-port="1" to-layer="841" to-port="0" /> <edge from-layer="840" from-port="0" to-layer="841" to-port="2" /> <edge from-layer="841" from-port="3" to-layer="842" to-port="0" /> <edge from-layer="842" from-port="2" to-layer="843" to-port="0" /> <edge from-layer="843" from-port="1" to-layer="845" to-port="0" /> <edge from-layer="844" from-port="0" to-layer="845" to-port="1" /> <edge from-layer="845" from-port="2" to-layer="847" to-port="2" /> <edge from-layer="846" from-port="0" to-layer="847" to-port="3" /> <edge from-layer="847" from-port="4" to-layer="850" to-port="1" /> <edge from-layer="847" from-port="4" to-layer="855" to-port="2" /> <edge from-layer="848" from-port="0" to-layer="850" to-port="2" /> <edge from-layer="849" from-port="0" to-layer="850" to-port="3" /> <edge from-layer="850" from-port="4" to-layer="852" to-port="0" /> <edge from-layer="851" from-port="0" to-layer="852" to-port="1" /> <edge from-layer="852" from-port="2" to-layer="856" to-port="0" /> <edge from-layer="853" from-port="0" to-layer="855" to-port="1" /> <edge from-layer="854" from-port="0" to-layer="855" to-port="3" /> <edge from-layer="855" from-port="4" to-layer="856" to-port="1" /> <edge from-layer="856" from-port="2" to-layer="864" to-port="0" /> <edge from-layer="857" from-port="0" to-layer="860" to-port="0" /> <edge from-layer="858" from-port="0" to-layer="860" to-port="1" /> <edge from-layer="859" from-port="0" to-layer="860" to-port="3" /> <edge from-layer="860" from-port="4" to-layer="862" to-port="0" /> <edge from-layer="861" from-port="0" to-layer="862" to-port="2" /> <edge from-layer="862" from-port="3" to-layer="863" to-port="0" /> <edge from-layer="863" from-port="2" to-layer="896" to-port="1" /> <edge from-layer="863" from-port="2" to-layer="864" to-port="1" /> <edge from-layer="864" from-port="2" to-layer="865" to-port="1" /> <edge from-layer="865" from-port="2" to-layer="870" to-port="0" /> <edge from-layer="866" from-port="0" to-layer="869" to-port="1" /> <edge from-layer="867" from-port="0" to-layer="869" to-port="2" /> <edge from-layer="868" from-port="0" to-layer="869" to-port="3" /> <edge from-layer="869" from-port="4" to-layer="870" to-port="1" /> <edge from-layer="870" from-port="2" to-layer="926" to-port="0" /> <edge from-layer="870" from-port="2" to-layer="964" to-port="0" /> <edge from-layer="871" from-port="0" to-layer="874" to-port="1" /> <edge from-layer="872" from-port="0" to-layer="874" to-port="2" /> <edge from-layer="873" from-port="0" to-layer="874" to-port="3" /> <edge from-layer="874" from-port="4" to-layer="875" to-port="0" /> <edge from-layer="874" from-port="4" to-layer="878" to-port="0" /> <edge from-layer="874" from-port="4" to-layer="894" to-port="0" /> <edge from-layer="874" from-port="4" to-layer="889" to-port="0" /> <edge from-layer="875" from-port="2" to-layer="897" to-port="0" /> <edge from-layer="876" from-port="0" to-layer="886" to-port="0" /> <edge from-layer="877" from-port="0" to-layer="886" to-port="1" /> <edge from-layer="878" from-port="1" to-layer="880" to-port="0" /> <edge from-layer="879" from-port="0" to-layer="880" to-port="2" /> <edge from-layer="880" from-port="3" to-layer="881" to-port="0" /> <edge from-layer="881" from-port="2" to-layer="882" to-port="0" /> <edge from-layer="882" from-port="1" to-layer="884" to-port="0" /> <edge from-layer="883" from-port="0" to-layer="884" to-port="1" /> <edge from-layer="884" from-port="2" to-layer="886" to-port="2" /> <edge from-layer="885" from-port="0" to-layer="886" to-port="3" /> <edge from-layer="886" from-port="4" to-layer="889" to-port="1" /> <edge from-layer="886" from-port="4" to-layer="894" to-port="2" /> <edge from-layer="887" from-port="0" to-layer="889" to-port="2" /> <edge from-layer="888" from-port="0" to-layer="889" to-port="3" /> <edge from-layer="889" from-port="4" to-layer="891" to-port="0" /> <edge from-layer="890" from-port="0" to-layer="891" to-port="1" /> <edge from-layer="891" from-port="2" to-layer="895" to-port="0" /> <edge from-layer="892" from-port="0" to-layer="894" to-port="1" /> <edge from-layer="893" from-port="0" to-layer="894" to-port="3" /> <edge from-layer="894" from-port="4" to-layer="895" to-port="1" /> <edge from-layer="895" from-port="2" to-layer="896" to-port="0" /> <edge from-layer="896" from-port="2" to-layer="897" to-port="1" /> <edge from-layer="897" from-port="2" to-layer="902" to-port="0" /> <edge from-layer="898" from-port="0" to-layer="901" to-port="1" /> <edge from-layer="899" from-port="0" to-layer="901" to-port="2" /> <edge from-layer="900" from-port="0" to-layer="901" to-port="3" /> <edge from-layer="901" from-port="4" to-layer="902" to-port="1" /> <edge from-layer="902" from-port="2" to-layer="903" to-port="1" /> <edge from-layer="903" from-port="2" to-layer="10127" to-port="0" /> <edge from-layer="903" from-port="2" to-layer="964" to-port="1" /> <edge from-layer="903" from-port="2" to-layer="923" to-port="0" /> <edge from-layer="904" from-port="0" to-layer="909" to-port="0" /> <edge from-layer="905" from-port="0" to-layer="908" to-port="1" /> <edge from-layer="906" from-port="0" to-layer="908" to-port="2" /> <edge from-layer="907" from-port="0" to-layer="908" to-port="3" /> <edge from-layer="908" from-port="4" to-layer="909" to-port="1" /> <edge from-layer="909" from-port="2" to-layer="910" to-port="0" /> <edge from-layer="910" from-port="1" to-layer="912" to-port="0" /> <edge from-layer="911" from-port="0" to-layer="912" to-port="2" /> <edge from-layer="912" from-port="3" to-layer="919" to-port="0" /> <edge from-layer="913" from-port="0" to-layer="916" to-port="1" /> <edge from-layer="914" from-port="0" to-layer="916" to-port="2" /> <edge from-layer="915" from-port="0" to-layer="916" to-port="3" /> <edge from-layer="916" from-port="4" to-layer="918" to-port="0" /> <edge from-layer="917" from-port="0" to-layer="918" to-port="1" /> <edge from-layer="918" from-port="2" to-layer="919" to-port="1" /> <edge from-layer="919" from-port="2" to-layer="964" to-port="2" /> <edge from-layer="919" from-port="2" to-layer="10126" to-port="0" /> <edge from-layer="920" from-port="0" to-layer="938" to-port="0" /> <edge from-layer="921" from-port="0" to-layer="933" to-port="0" /> <edge from-layer="922" from-port="0" to-layer="933" to-port="1" /> <edge from-layer="922" from-port="0" to-layer="936" to-port="1" /> <edge from-layer="923" from-port="1" to-layer="925" to-port="0" /> <edge from-layer="924" from-port="0" to-layer="925" to-port="2" /> <edge from-layer="925" from-port="3" to-layer="929" to-port="0" /> <edge from-layer="925" from-port="3" to-layer="935" to-port="0" /> <edge from-layer="926" from-port="1" to-layer="953" to-port="0" /> <edge from-layer="926" from-port="1" to-layer="928" to-port="0" /> <edge from-layer="927" from-port="0" to-layer="928" to-port="2" /> <edge from-layer="928" from-port="3" to-layer="929" to-port="1" /> <edge from-layer="929" from-port="2" to-layer="931" to-port="0" /> <edge from-layer="930" from-port="0" to-layer="931" to-port="1" /> <edge from-layer="930" from-port="0" to-layer="935" to-port="1" /> <edge from-layer="931" from-port="2" to-layer="933" to-port="2" /> <edge from-layer="932" from-port="0" to-layer="933" to-port="3" /> <edge from-layer="932" from-port="0" to-layer="936" to-port="3" /> <edge from-layer="933" from-port="4" to-layer="938" to-port="1" /> <edge from-layer="934" from-port="0" to-layer="936" to-port="0" /> <edge from-layer="935" from-port="2" to-layer="943" to-port="2" /> <edge from-layer="935" from-port="2" to-layer="936" to-port="2" /> <edge from-layer="936" from-port="4" to-layer="938" to-port="2" /> <edge from-layer="937" from-port="0" to-layer="938" to-port="3" /> <edge from-layer="938" from-port="4" to-layer="945" to-port="0" /> <edge from-layer="939" from-port="0" to-layer="945" to-port="1" /> <edge from-layer="940" from-port="0" to-layer="943" to-port="0" /> <edge from-layer="941" from-port="0" to-layer="943" to-port="1" /> <edge from-layer="942" from-port="0" to-layer="943" to-port="3" /> <edge from-layer="943" from-port="4" to-layer="945" to-port="2" /> <edge from-layer="944" from-port="0" to-layer="945" to-port="3" /> <edge from-layer="945" from-port="4" to-layer="948" to-port="0" /> <edge from-layer="946" from-port="0" to-layer="948" to-port="1" /> <edge from-layer="947" from-port="0" to-layer="948" to-port="2" /> <edge from-layer="948" from-port="3" to-layer="949" to-port="0" /> <edge from-layer="949" from-port="2" to-layer="950" to-port="0" /> <edge from-layer="949" from-port="2" to-layer="963" to-port="0" /> <edge from-layer="950" from-port="1" to-layer="962" to-port="0" /> <edge from-layer="951" from-port="0" to-layer="953" to-port="1" /> <edge from-layer="952" from-port="0" to-layer="953" to-port="2" /> <edge from-layer="953" from-port="3" to-layer="957" to-port="0" /> <edge from-layer="954" from-port="0" to-layer="957" to-port="1" /> <edge from-layer="955" from-port="0" to-layer="957" to-port="2" /> <edge from-layer="956" from-port="0" to-layer="957" to-port="3" /> <edge from-layer="957" from-port="4" to-layer="961" to-port="2" /> <edge from-layer="957" from-port="4" to-layer="959" to-port="0" /> <edge from-layer="958" from-port="0" to-layer="959" to-port="1" /> <edge from-layer="959" from-port="2" to-layer="961" to-port="0" /> <edge from-layer="960" from-port="0" to-layer="961" to-port="1" /> <edge from-layer="961" from-port="3" to-layer="962" to-port="1" /> <edge from-layer="962" from-port="2" to-layer="963" to-port="1" /> <edge from-layer="963" from-port="2" to-layer="964" to-port="3" /> <edge from-layer="964" from-port="4" to-layer="966" to-port="0" /> <edge from-layer="965" from-port="0" to-layer="966" to-port="1" /> <edge from-layer="966" from-port="2" to-layer="968" to-port="0" /> <edge from-layer="967" from-port="0" to-layer="968" to-port="1" /> <edge from-layer="968" from-port="2" to-layer="976" to-port="0" /> <edge from-layer="969" from-port="0" to-layer="970" to-port="0" /> <edge from-layer="970" from-port="1" to-layer="973" to-port="0" /> <edge from-layer="971" from-port="0" to-layer="972" to-port="0" /> <edge from-layer="972" from-port="1" to-layer="973" to-port="1" /> <edge from-layer="973" from-port="2" to-layer="975" to-port="0" /> <edge from-layer="974" from-port="0" to-layer="975" to-port="1" /> <edge from-layer="975" from-port="2" to-layer="976" to-port="1" /> <edge from-layer="976" from-port="2" to-layer="978" to-port="0" /> <edge from-layer="977" from-port="0" to-layer="978" to-port="1" /> <edge from-layer="978" from-port="2" to-layer="979" to-port="1" /> <edge from-layer="979" from-port="2" to-layer="980" to-port="0" /> <edge from-layer="980" from-port="2" to-layer="982" to-port="0" /> <edge from-layer="980" from-port="2" to-layer="1009" to-port="0" /> <edge from-layer="980" from-port="2" to-layer="1211" to-port="1" /> <edge from-layer="981" from-port="0" to-layer="982" to-port="1" /> <edge from-layer="982" from-port="2" to-layer="984" to-port="0" /> <edge from-layer="983" from-port="0" to-layer="984" to-port="1" /> <edge from-layer="984" from-port="2" to-layer="986" to-port="0" /> <edge from-layer="985" from-port="0" to-layer="986" to-port="1" /> <edge from-layer="986" from-port="2" to-layer="994" to-port="0" /> <edge from-layer="987" from-port="0" to-layer="988" to-port="0" /> <edge from-layer="988" from-port="1" to-layer="991" to-port="0" /> <edge from-layer="989" from-port="0" to-layer="990" to-port="0" /> <edge from-layer="990" from-port="1" to-layer="991" to-port="1" /> <edge from-layer="991" from-port="2" to-layer="993" to-port="0" /> <edge from-layer="992" from-port="0" to-layer="993" to-port="1" /> <edge from-layer="993" from-port="2" to-layer="994" to-port="1" /> <edge from-layer="994" from-port="2" to-layer="996" to-port="0" /> <edge from-layer="995" from-port="0" to-layer="996" to-port="1" /> <edge from-layer="996" from-port="2" to-layer="997" to-port="0" /> <edge from-layer="997" from-port="1" to-layer="1005" to-port="0" /> <edge from-layer="998" from-port="0" to-layer="999" to-port="0" /> <edge from-layer="999" from-port="1" to-layer="1002" to-port="0" /> <edge from-layer="1000" from-port="0" to-layer="1001" to-port="0" /> <edge from-layer="1001" from-port="1" to-layer="1002" to-port="1" /> <edge from-layer="1002" from-port="2" to-layer="1004" to-port="0" /> <edge from-layer="1003" from-port="0" to-layer="1004" to-port="1" /> <edge from-layer="1004" from-port="2" to-layer="1005" to-port="1" /> <edge from-layer="1005" from-port="2" to-layer="1007" to-port="0" /> <edge from-layer="1006" from-port="0" to-layer="1007" to-port="1" /> <edge from-layer="1007" from-port="2" to-layer="1210" to-port="0" /> <edge from-layer="1008" from-port="0" to-layer="1009" to-port="1" /> <edge from-layer="1009" from-port="2" to-layer="1011" to-port="0" /> <edge from-layer="1010" from-port="0" to-layer="1011" to-port="1" /> <edge from-layer="1011" from-port="2" to-layer="1013" to-port="0" /> <edge from-layer="1012" from-port="0" to-layer="1013" to-port="1" /> <edge from-layer="1013" from-port="2" to-layer="1021" to-port="0" /> <edge from-layer="1014" from-port="0" to-layer="1015" to-port="0" /> <edge from-layer="1015" from-port="1" to-layer="1018" to-port="0" /> <edge from-layer="1016" from-port="0" to-layer="1017" to-port="0" /> <edge from-layer="1017" from-port="1" to-layer="1018" to-port="1" /> <edge from-layer="1018" from-port="2" to-layer="1020" to-port="0" /> <edge from-layer="1019" from-port="0" to-layer="1020" to-port="1" /> <edge from-layer="1020" from-port="2" to-layer="1021" to-port="1" /> <edge from-layer="1021" from-port="2" to-layer="1023" to-port="0" /> <edge from-layer="1022" from-port="0" to-layer="1023" to-port="1" /> <edge from-layer="1023" from-port="2" to-layer="1025" to-port="0" /> <edge from-layer="1024" from-port="0" to-layer="1025" to-port="1" /> <edge from-layer="1025" from-port="2" to-layer="1029" to-port="0" /> <edge from-layer="1025" from-port="2" to-layer="1147" to-port="0" /> <edge from-layer="1025" from-port="2" to-layer="1041" to-port="0" /> <edge from-layer="1026" from-port="0" to-layer="1029" to-port="1" /> <edge from-layer="1027" from-port="0" to-layer="1029" to-port="2" /> <edge from-layer="1028" from-port="0" to-layer="1029" to-port="3" /> <edge from-layer="1029" from-port="4" to-layer="1031" to-port="0" /> <edge from-layer="1030" from-port="0" to-layer="1031" to-port="1" /> <edge from-layer="1031" from-port="2" to-layer="1035" to-port="0" /> <edge from-layer="1031" from-port="2" to-layer="1100" to-port="0" /> <edge from-layer="1032" from-port="0" to-layer="1035" to-port="1" /> <edge from-layer="1033" from-port="0" to-layer="1035" to-port="2" /> <edge from-layer="1034" from-port="0" to-layer="1035" to-port="3" /> <edge from-layer="1035" from-port="4" to-layer="1081" to-port="0" /> <edge from-layer="1035" from-port="4" to-layer="1070" to-port="0" /> <edge from-layer="1035" from-port="4" to-layer="1086" to-port="0" /> <edge from-layer="1035" from-port="4" to-layer="1067" to-port="0" /> <edge from-layer="1036" from-port="0" to-layer="1063" to-port="0" /> <edge from-layer="1037" from-port="0" to-layer="1063" to-port="1" /> <edge from-layer="1038" from-port="0" to-layer="1041" to-port="1" /> <edge from-layer="1039" from-port="0" to-layer="1041" to-port="2" /> <edge from-layer="1040" from-port="0" to-layer="1041" to-port="3" /> <edge from-layer="1041" from-port="4" to-layer="1043" to-port="0" /> <edge from-layer="1042" from-port="0" to-layer="1043" to-port="1" /> <edge from-layer="1043" from-port="2" to-layer="1044" to-port="0" /> <edge from-layer="1043" from-port="2" to-layer="1105" to-port="0" /> <edge from-layer="1043" from-port="2" to-layer="1132" to-port="0" /> <edge from-layer="1044" from-port="1" to-layer="1046" to-port="0" /> <edge from-layer="1045" from-port="0" to-layer="1046" to-port="2" /> <edge from-layer="1046" from-port="3" to-layer="1059" to-port="0" /> <edge from-layer="1047" from-port="0" to-layer="1052" to-port="0" /> <edge from-layer="1048" from-port="0" to-layer="1051" to-port="1" /> <edge from-layer="1049" from-port="0" to-layer="1051" to-port="2" /> <edge from-layer="1050" from-port="0" to-layer="1051" to-port="3" /> <edge from-layer="1051" from-port="4" to-layer="1052" to-port="1" /> <edge from-layer="1052" from-port="2" to-layer="1053" to-port="0" /> <edge from-layer="1053" from-port="1" to-layer="1055" to-port="0" /> <edge from-layer="1054" from-port="0" to-layer="1055" to-port="2" /> <edge from-layer="1055" from-port="3" to-layer="1134" to-port="0" /> <edge from-layer="1055" from-port="3" to-layer="1056" to-port="0" /> <edge from-layer="1056" from-port="1" to-layer="1058" to-port="0" /> <edge from-layer="1057" from-port="0" to-layer="1058" to-port="2" /> <edge from-layer="1058" from-port="3" to-layer="1059" to-port="1" /> <edge from-layer="1059" from-port="2" to-layer="1061" to-port="0" /> <edge from-layer="1060" from-port="0" to-layer="1061" to-port="1" /> <edge from-layer="1061" from-port="2" to-layer="1063" to-port="2" /> <edge from-layer="1061" from-port="2" to-layer="1091" to-port="2" /> <edge from-layer="1062" from-port="0" to-layer="1063" to-port="3" /> <edge from-layer="1063" from-port="4" to-layer="1065" to-port="0" /> <edge from-layer="1064" from-port="0" to-layer="1065" to-port="2" /> <edge from-layer="1065" from-port="3" to-layer="1066" to-port="0" /> <edge from-layer="1066" from-port="2" to-layer="1106" to-port="1" /> <edge from-layer="1066" from-port="2" to-layer="1067" to-port="1" /> <edge from-layer="1067" from-port="2" to-layer="1096" to-port="0" /> <edge from-layer="1068" from-port="0" to-layer="1078" to-port="0" /> <edge from-layer="1069" from-port="0" to-layer="1078" to-port="1" /> <edge from-layer="1070" from-port="1" to-layer="1072" to-port="0" /> <edge from-layer="1071" from-port="0" to-layer="1072" to-port="2" /> <edge from-layer="1072" from-port="3" to-layer="1073" to-port="0" /> <edge from-layer="1073" from-port="2" to-layer="1074" to-port="0" /> <edge from-layer="1074" from-port="1" to-layer="1076" to-port="0" /> <edge from-layer="1075" from-port="0" to-layer="1076" to-port="1" /> <edge from-layer="1076" from-port="2" to-layer="1078" to-port="2" /> <edge from-layer="1077" from-port="0" to-layer="1078" to-port="3" /> <edge from-layer="1078" from-port="4" to-layer="1086" to-port="2" /> <edge from-layer="1078" from-port="4" to-layer="1081" to-port="1" /> <edge from-layer="1079" from-port="0" to-layer="1081" to-port="2" /> <edge from-layer="1080" from-port="0" to-layer="1081" to-port="3" /> <edge from-layer="1081" from-port="4" to-layer="1083" to-port="0" /> <edge from-layer="1082" from-port="0" to-layer="1083" to-port="1" /> <edge from-layer="1083" from-port="2" to-layer="1087" to-port="0" /> <edge from-layer="1084" from-port="0" to-layer="1086" to-port="1" /> <edge from-layer="1085" from-port="0" to-layer="1086" to-port="3" /> <edge from-layer="1086" from-port="4" to-layer="1087" to-port="1" /> <edge from-layer="1087" from-port="2" to-layer="1095" to-port="0" /> <edge from-layer="1088" from-port="0" to-layer="1091" to-port="0" /> <edge from-layer="1089" from-port="0" to-layer="1091" to-port="1" /> <edge from-layer="1090" from-port="0" to-layer="1091" to-port="3" /> <edge from-layer="1091" from-port="4" to-layer="1093" to-port="0" /> <edge from-layer="1092" from-port="0" to-layer="1093" to-port="2" /> <edge from-layer="1093" from-port="3" to-layer="1094" to-port="0" /> <edge from-layer="1094" from-port="2" to-layer="1127" to-port="1" /> <edge from-layer="1094" from-port="2" to-layer="1095" to-port="1" /> <edge from-layer="1095" from-port="2" to-layer="1096" to-port="1" /> <edge from-layer="1096" from-port="2" to-layer="1101" to-port="0" /> <edge from-layer="1097" from-port="0" to-layer="1100" to-port="1" /> <edge from-layer="1098" from-port="0" to-layer="1100" to-port="2" /> <edge from-layer="1099" from-port="0" to-layer="1100" to-port="3" /> <edge from-layer="1100" from-port="4" to-layer="1101" to-port="1" /> <edge from-layer="1101" from-port="2" to-layer="1195" to-port="0" /> <edge from-layer="1101" from-port="2" to-layer="1157" to-port="0" /> <edge from-layer="1102" from-port="0" to-layer="1105" to-port="1" /> <edge from-layer="1103" from-port="0" to-layer="1105" to-port="2" /> <edge from-layer="1104" from-port="0" to-layer="1105" to-port="3" /> <edge from-layer="1105" from-port="4" to-layer="1106" to-port="0" /> <edge from-layer="1105" from-port="4" to-layer="1125" to-port="0" /> <edge from-layer="1105" from-port="4" to-layer="1120" to-port="0" /> <edge from-layer="1105" from-port="4" to-layer="1109" to-port="0" /> <edge from-layer="1106" from-port="2" to-layer="1128" to-port="0" /> <edge from-layer="1107" from-port="0" to-layer="1117" to-port="0" /> <edge from-layer="1108" from-port="0" to-layer="1117" to-port="1" /> <edge from-layer="1109" from-port="1" to-layer="1111" to-port="0" /> <edge from-layer="1110" from-port="0" to-layer="1111" to-port="2" /> <edge from-layer="1111" from-port="3" to-layer="1112" to-port="0" /> <edge from-layer="1112" from-port="2" to-layer="1113" to-port="0" /> <edge from-layer="1113" from-port="1" to-layer="1115" to-port="0" /> <edge from-layer="1114" from-port="0" to-layer="1115" to-port="1" /> <edge from-layer="1115" from-port="2" to-layer="1117" to-port="2" /> <edge from-layer="1116" from-port="0" to-layer="1117" to-port="3" /> <edge from-layer="1117" from-port="4" to-layer="1120" to-port="1" /> <edge from-layer="1117" from-port="4" to-layer="1125" to-port="2" /> <edge from-layer="1118" from-port="0" to-layer="1120" to-port="2" /> <edge from-layer="1119" from-port="0" to-layer="1120" to-port="3" /> <edge from-layer="1120" from-port="4" to-layer="1122" to-port="0" /> <edge from-layer="1121" from-port="0" to-layer="1122" to-port="1" /> <edge from-layer="1122" from-port="2" to-layer="1126" to-port="0" /> <edge from-layer="1123" from-port="0" to-layer="1125" to-port="1" /> <edge from-layer="1124" from-port="0" to-layer="1125" to-port="3" /> <edge from-layer="1125" from-port="4" to-layer="1126" to-port="1" /> <edge from-layer="1126" from-port="2" to-layer="1127" to-port="0" /> <edge from-layer="1127" from-port="2" to-layer="1128" to-port="1" /> <edge from-layer="1128" from-port="2" to-layer="1133" to-port="0" /> <edge from-layer="1129" from-port="0" to-layer="1132" to-port="1" /> <edge from-layer="1130" from-port="0" to-layer="1132" to-port="2" /> <edge from-layer="1131" from-port="0" to-layer="1132" to-port="3" /> <edge from-layer="1132" from-port="4" to-layer="1133" to-port="1" /> <edge from-layer="1133" from-port="2" to-layer="1134" to-port="1" /> <edge from-layer="1134" from-port="2" to-layer="1195" to-port="1" /> <edge from-layer="1134" from-port="2" to-layer="10105" to-port="0" /> <edge from-layer="1134" from-port="2" to-layer="1154" to-port="0" /> <edge from-layer="1135" from-port="0" to-layer="1140" to-port="0" /> <edge from-layer="1136" from-port="0" to-layer="1139" to-port="1" /> <edge from-layer="1137" from-port="0" to-layer="1139" to-port="2" /> <edge from-layer="1138" from-port="0" to-layer="1139" to-port="3" /> <edge from-layer="1139" from-port="4" to-layer="1140" to-port="1" /> <edge from-layer="1140" from-port="2" to-layer="1141" to-port="0" /> <edge from-layer="1141" from-port="1" to-layer="1143" to-port="0" /> <edge from-layer="1142" from-port="0" to-layer="1143" to-port="2" /> <edge from-layer="1143" from-port="3" to-layer="1150" to-port="0" /> <edge from-layer="1144" from-port="0" to-layer="1147" to-port="1" /> <edge from-layer="1145" from-port="0" to-layer="1147" to-port="2" /> <edge from-layer="1146" from-port="0" to-layer="1147" to-port="3" /> <edge from-layer="1147" from-port="4" to-layer="1149" to-port="0" /> <edge from-layer="1148" from-port="0" to-layer="1149" to-port="1" /> <edge from-layer="1149" from-port="2" to-layer="1150" to-port="1" /> <edge from-layer="1150" from-port="2" to-layer="10104" to-port="0" /> <edge from-layer="1150" from-port="2" to-layer="1195" to-port="2" /> <edge from-layer="1151" from-port="0" to-layer="1169" to-port="0" /> <edge from-layer="1152" from-port="0" to-layer="1164" to-port="0" /> <edge from-layer="1153" from-port="0" to-layer="1164" to-port="1" /> <edge from-layer="1153" from-port="0" to-layer="1167" to-port="1" /> <edge from-layer="1154" from-port="1" to-layer="1156" to-port="0" /> <edge from-layer="1155" from-port="0" to-layer="1156" to-port="2" /> <edge from-layer="1156" from-port="3" to-layer="1166" to-port="0" /> <edge from-layer="1156" from-port="3" to-layer="1160" to-port="0" /> <edge from-layer="1157" from-port="1" to-layer="1184" to-port="0" /> <edge from-layer="1157" from-port="1" to-layer="1159" to-port="0" /> <edge from-layer="1158" from-port="0" to-layer="1159" to-port="2" /> <edge from-layer="1159" from-port="3" to-layer="1160" to-port="1" /> <edge from-layer="1160" from-port="2" to-layer="1162" to-port="0" /> <edge from-layer="1161" from-port="0" to-layer="1162" to-port="1" /> <edge from-layer="1161" from-port="0" to-layer="1166" to-port="1" /> <edge from-layer="1162" from-port="2" to-layer="1164" to-port="2" /> <edge from-layer="1163" from-port="0" to-layer="1167" to-port="3" /> <edge from-layer="1163" from-port="0" to-layer="1164" to-port="3" /> <edge from-layer="1164" from-port="4" to-layer="1169" to-port="1" /> <edge from-layer="1165" from-port="0" to-layer="1167" to-port="0" /> <edge from-layer="1166" from-port="2" to-layer="1167" to-port="2" /> <edge from-layer="1166" from-port="2" to-layer="1174" to-port="2" /> <edge from-layer="1167" from-port="4" to-layer="1169" to-port="2" /> <edge from-layer="1168" from-port="0" to-layer="1169" to-port="3" /> <edge from-layer="1169" from-port="4" to-layer="1176" to-port="0" /> <edge from-layer="1170" from-port="0" to-layer="1176" to-port="1" /> <edge from-layer="1171" from-port="0" to-layer="1174" to-port="0" /> <edge from-layer="1172" from-port="0" to-layer="1174" to-port="1" /> <edge from-layer="1173" from-port="0" to-layer="1174" to-port="3" /> <edge from-layer="1174" from-port="4" to-layer="1176" to-port="2" /> <edge from-layer="1175" from-port="0" to-layer="1176" to-port="3" /> <edge from-layer="1176" from-port="4" to-layer="1179" to-port="0" /> <edge from-layer="1177" from-port="0" to-layer="1179" to-port="1" /> <edge from-layer="1178" from-port="0" to-layer="1179" to-port="2" /> <edge from-layer="1179" from-port="3" to-layer="1180" to-port="0" /> <edge from-layer="1180" from-port="2" to-layer="1181" to-port="0" /> <edge from-layer="1180" from-port="2" to-layer="1194" to-port="0" /> <edge from-layer="1181" from-port="1" to-layer="1193" to-port="0" /> <edge from-layer="1182" from-port="0" to-layer="1184" to-port="1" /> <edge from-layer="1183" from-port="0" to-layer="1184" to-port="2" /> <edge from-layer="1184" from-port="3" to-layer="1188" to-port="0" /> <edge from-layer="1185" from-port="0" to-layer="1188" to-port="1" /> <edge from-layer="1186" from-port="0" to-layer="1188" to-port="2" /> <edge from-layer="1187" from-port="0" to-layer="1188" to-port="3" /> <edge from-layer="1188" from-port="4" to-layer="1190" to-port="0" /> <edge from-layer="1188" from-port="4" to-layer="1192" to-port="2" /> <edge from-layer="1189" from-port="0" to-layer="1190" to-port="1" /> <edge from-layer="1190" from-port="2" to-layer="1192" to-port="0" /> <edge from-layer="1191" from-port="0" to-layer="1192" to-port="1" /> <edge from-layer="1192" from-port="3" to-layer="1193" to-port="1" /> <edge from-layer="1193" from-port="2" to-layer="1194" to-port="1" /> <edge from-layer="1194" from-port="2" to-layer="1195" to-port="3" /> <edge from-layer="1195" from-port="4" to-layer="1197" to-port="0" /> <edge from-layer="1196" from-port="0" to-layer="1197" to-port="1" /> <edge from-layer="1197" from-port="2" to-layer="1199" to-port="0" /> <edge from-layer="1198" from-port="0" to-layer="1199" to-port="1" /> <edge from-layer="1199" from-port="2" to-layer="1207" to-port="0" /> <edge from-layer="1200" from-port="0" to-layer="1201" to-port="0" /> <edge from-layer="1201" from-port="1" to-layer="1204" to-port="0" /> <edge from-layer="1202" from-port="0" to-layer="1203" to-port="0" /> <edge from-layer="1203" from-port="1" to-layer="1204" to-port="1" /> <edge from-layer="1204" from-port="2" to-layer="1206" to-port="0" /> <edge from-layer="1205" from-port="0" to-layer="1206" to-port="1" /> <edge from-layer="1206" from-port="2" to-layer="1207" to-port="1" /> <edge from-layer="1207" from-port="2" to-layer="1209" to-port="0" /> <edge from-layer="1208" from-port="0" to-layer="1209" to-port="1" /> <edge from-layer="1209" from-port="2" to-layer="1210" to-port="1" /> <edge from-layer="1210" from-port="2" to-layer="1211" to-port="0" /> <edge from-layer="1211" from-port="2" to-layer="1213" to-port="0" /> <edge from-layer="1211" from-port="2" to-layer="1442" to-port="1" /> <edge from-layer="1211" from-port="2" to-layer="1240" to-port="0" /> <edge from-layer="1212" from-port="0" to-layer="1213" to-port="1" /> <edge from-layer="1213" from-port="2" to-layer="1215" to-port="0" /> <edge from-layer="1214" from-port="0" to-layer="1215" to-port="1" /> <edge from-layer="1215" from-port="2" to-layer="1217" to-port="0" /> <edge from-layer="1216" from-port="0" to-layer="1217" to-port="1" /> <edge from-layer="1217" from-port="2" to-layer="1225" to-port="0" /> <edge from-layer="1218" from-port="0" to-layer="1219" to-port="0" /> <edge from-layer="1219" from-port="1" to-layer="1222" to-port="0" /> <edge from-layer="1220" from-port="0" to-layer="1221" to-port="0" /> <edge from-layer="1221" from-port="1" to-layer="1222" to-port="1" /> <edge from-layer="1222" from-port="2" to-layer="1224" to-port="0" /> <edge from-layer="1223" from-port="0" to-layer="1224" to-port="1" /> <edge from-layer="1224" from-port="2" to-layer="1225" to-port="1" /> <edge from-layer="1225" from-port="2" to-layer="1227" to-port="0" /> <edge from-layer="1226" from-port="0" to-layer="1227" to-port="1" /> <edge from-layer="1227" from-port="2" to-layer="1228" to-port="0" /> <edge from-layer="1228" from-port="1" to-layer="1236" to-port="0" /> <edge from-layer="1229" from-port="0" to-layer="1230" to-port="0" /> <edge from-layer="1230" from-port="1" to-layer="1233" to-port="0" /> <edge from-layer="1231" from-port="0" to-layer="1232" to-port="0" /> <edge from-layer="1232" from-port="1" to-layer="1233" to-port="1" /> <edge from-layer="1233" from-port="2" to-layer="1235" to-port="0" /> <edge from-layer="1234" from-port="0" to-layer="1235" to-port="1" /> <edge from-layer="1235" from-port="2" to-layer="1236" to-port="1" /> <edge from-layer="1236" from-port="2" to-layer="1238" to-port="0" /> <edge from-layer="1237" from-port="0" to-layer="1238" to-port="1" /> <edge from-layer="1238" from-port="2" to-layer="1441" to-port="0" /> <edge from-layer="1239" from-port="0" to-layer="1240" to-port="1" /> <edge from-layer="1240" from-port="2" to-layer="1242" to-port="0" /> <edge from-layer="1241" from-port="0" to-layer="1242" to-port="1" /> <edge from-layer="1242" from-port="2" to-layer="1244" to-port="0" /> <edge from-layer="1243" from-port="0" to-layer="1244" to-port="1" /> <edge from-layer="1244" from-port="2" to-layer="1252" to-port="0" /> <edge from-layer="1245" from-port="0" to-layer="1246" to-port="0" /> <edge from-layer="1246" from-port="1" to-layer="1249" to-port="0" /> <edge from-layer="1247" from-port="0" to-layer="1248" to-port="0" /> <edge from-layer="1248" from-port="1" to-layer="1249" to-port="1" /> <edge from-layer="1249" from-port="2" to-layer="1251" to-port="0" /> <edge from-layer="1250" from-port="0" to-layer="1251" to-port="1" /> <edge from-layer="1251" from-port="2" to-layer="1252" to-port="1" /> <edge from-layer="1252" from-port="2" to-layer="1254" to-port="0" /> <edge from-layer="1253" from-port="0" to-layer="1254" to-port="1" /> <edge from-layer="1254" from-port="2" to-layer="1256" to-port="0" /> <edge from-layer="1255" from-port="0" to-layer="1256" to-port="1" /> <edge from-layer="1256" from-port="2" to-layer="1260" to-port="0" /> <edge from-layer="1256" from-port="2" to-layer="1272" to-port="0" /> <edge from-layer="1256" from-port="2" to-layer="1378" to-port="0" /> <edge from-layer="1257" from-port="0" to-layer="1260" to-port="1" /> <edge from-layer="1258" from-port="0" to-layer="1260" to-port="2" /> <edge from-layer="1259" from-port="0" to-layer="1260" to-port="3" /> <edge from-layer="1260" from-port="4" to-layer="1262" to-port="0" /> <edge from-layer="1261" from-port="0" to-layer="1262" to-port="1" /> <edge from-layer="1262" from-port="2" to-layer="1266" to-port="0" /> <edge from-layer="1262" from-port="2" to-layer="1331" to-port="0" /> <edge from-layer="1263" from-port="0" to-layer="1266" to-port="1" /> <edge from-layer="1264" from-port="0" to-layer="1266" to-port="2" /> <edge from-layer="1265" from-port="0" to-layer="1266" to-port="3" /> <edge from-layer="1266" from-port="4" to-layer="1301" to-port="0" /> <edge from-layer="1266" from-port="4" to-layer="1317" to-port="0" /> <edge from-layer="1266" from-port="4" to-layer="1312" to-port="0" /> <edge from-layer="1266" from-port="4" to-layer="1298" to-port="0" /> <edge from-layer="1267" from-port="0" to-layer="1294" to-port="0" /> <edge from-layer="1268" from-port="0" to-layer="1294" to-port="1" /> <edge from-layer="1269" from-port="0" to-layer="1272" to-port="1" /> <edge from-layer="1270" from-port="0" to-layer="1272" to-port="2" /> <edge from-layer="1271" from-port="0" to-layer="1272" to-port="3" /> <edge from-layer="1272" from-port="4" to-layer="1274" to-port="0" /> <edge from-layer="1273" from-port="0" to-layer="1274" to-port="1" /> <edge from-layer="1274" from-port="2" to-layer="1363" to-port="0" /> <edge from-layer="1274" from-port="2" to-layer="1336" to-port="0" /> <edge from-layer="1274" from-port="2" to-layer="1275" to-port="0" /> <edge from-layer="1275" from-port="1" to-layer="1277" to-port="0" /> <edge from-layer="1276" from-port="0" to-layer="1277" to-port="2" /> <edge from-layer="1277" from-port="3" to-layer="1290" to-port="0" /> <edge from-layer="1278" from-port="0" to-layer="1283" to-port="0" /> <edge from-layer="1279" from-port="0" to-layer="1282" to-port="1" /> <edge from-layer="1280" from-port="0" to-layer="1282" to-port="2" /> <edge from-layer="1281" from-port="0" to-layer="1282" to-port="3" /> <edge from-layer="1282" from-port="4" to-layer="1283" to-port="1" /> <edge from-layer="1283" from-port="2" to-layer="1284" to-port="0" /> <edge from-layer="1284" from-port="1" to-layer="1286" to-port="0" /> <edge from-layer="1285" from-port="0" to-layer="1286" to-port="2" /> <edge from-layer="1286" from-port="3" to-layer="1287" to-port="0" /> <edge from-layer="1286" from-port="3" to-layer="1365" to-port="0" /> <edge from-layer="1287" from-port="1" to-layer="1289" to-port="0" /> <edge from-layer="1288" from-port="0" to-layer="1289" to-port="2" /> <edge from-layer="1289" from-port="3" to-layer="1290" to-port="1" /> <edge from-layer="1290" from-port="2" to-layer="1292" to-port="0" /> <edge from-layer="1291" from-port="0" to-layer="1292" to-port="1" /> <edge from-layer="1292" from-port="2" to-layer="1294" to-port="2" /> <edge from-layer="1292" from-port="2" to-layer="1322" to-port="2" /> <edge from-layer="1293" from-port="0" to-layer="1294" to-port="3" /> <edge from-layer="1294" from-port="4" to-layer="1296" to-port="0" /> <edge from-layer="1295" from-port="0" to-layer="1296" to-port="2" /> <edge from-layer="1296" from-port="3" to-layer="1297" to-port="0" /> <edge from-layer="1297" from-port="2" to-layer="1337" to-port="1" /> <edge from-layer="1297" from-port="2" to-layer="1298" to-port="1" /> <edge from-layer="1298" from-port="2" to-layer="1327" to-port="0" /> <edge from-layer="1299" from-port="0" to-layer="1309" to-port="0" /> <edge from-layer="1300" from-port="0" to-layer="1309" to-port="1" /> <edge from-layer="1301" from-port="1" to-layer="1303" to-port="0" /> <edge from-layer="1302" from-port="0" to-layer="1303" to-port="2" /> <edge from-layer="1303" from-port="3" to-layer="1304" to-port="0" /> <edge from-layer="1304" from-port="2" to-layer="1305" to-port="0" /> <edge from-layer="1305" from-port="1" to-layer="1307" to-port="0" /> <edge from-layer="1306" from-port="0" to-layer="1307" to-port="1" /> <edge from-layer="1307" from-port="2" to-layer="1309" to-port="2" /> <edge from-layer="1308" from-port="0" to-layer="1309" to-port="3" /> <edge from-layer="1309" from-port="4" to-layer="1317" to-port="2" /> <edge from-layer="1309" from-port="4" to-layer="1312" to-port="1" /> <edge from-layer="1310" from-port="0" to-layer="1312" to-port="2" /> <edge from-layer="1311" from-port="0" to-layer="1312" to-port="3" /> <edge from-layer="1312" from-port="4" to-layer="1314" to-port="0" /> <edge from-layer="1313" from-port="0" to-layer="1314" to-port="1" /> <edge from-layer="1314" from-port="2" to-layer="1318" to-port="0" /> <edge from-layer="1315" from-port="0" to-layer="1317" to-port="1" /> <edge from-layer="1316" from-port="0" to-layer="1317" to-port="3" /> <edge from-layer="1317" from-port="4" to-layer="1318" to-port="1" /> <edge from-layer="1318" from-port="2" to-layer="1326" to-port="0" /> <edge from-layer="1319" from-port="0" to-layer="1322" to-port="0" /> <edge from-layer="1320" from-port="0" to-layer="1322" to-port="1" /> <edge from-layer="1321" from-port="0" to-layer="1322" to-port="3" /> <edge from-layer="1322" from-port="4" to-layer="1324" to-port="0" /> <edge from-layer="1323" from-port="0" to-layer="1324" to-port="2" /> <edge from-layer="1324" from-port="3" to-layer="1325" to-port="0" /> <edge from-layer="1325" from-port="2" to-layer="1326" to-port="1" /> <edge from-layer="1325" from-port="2" to-layer="1358" to-port="1" /> <edge from-layer="1326" from-port="2" to-layer="1327" to-port="1" /> <edge from-layer="1327" from-port="2" to-layer="1332" to-port="0" /> <edge from-layer="1328" from-port="0" to-layer="1331" to-port="1" /> <edge from-layer="1329" from-port="0" to-layer="1331" to-port="2" /> <edge from-layer="1330" from-port="0" to-layer="1331" to-port="3" /> <edge from-layer="1331" from-port="4" to-layer="1332" to-port="1" /> <edge from-layer="1332" from-port="2" to-layer="1388" to-port="0" /> <edge from-layer="1332" from-port="2" to-layer="1426" to-port="0" /> <edge from-layer="1333" from-port="0" to-layer="1336" to-port="1" /> <edge from-layer="1334" from-port="0" to-layer="1336" to-port="2" /> <edge from-layer="1335" from-port="0" to-layer="1336" to-port="3" /> <edge from-layer="1336" from-port="4" to-layer="1337" to-port="0" /> <edge from-layer="1336" from-port="4" to-layer="1340" to-port="0" /> <edge from-layer="1336" from-port="4" to-layer="1356" to-port="0" /> <edge from-layer="1336" from-port="4" to-layer="1351" to-port="0" /> <edge from-layer="1337" from-port="2" to-layer="1359" to-port="0" /> <edge from-layer="1338" from-port="0" to-layer="1348" to-port="0" /> <edge from-layer="1339" from-port="0" to-layer="1348" to-port="1" /> <edge from-layer="1340" from-port="1" to-layer="1342" to-port="0" /> <edge from-layer="1341" from-port="0" to-layer="1342" to-port="2" /> <edge from-layer="1342" from-port="3" to-layer="1343" to-port="0" /> <edge from-layer="1343" from-port="2" to-layer="1344" to-port="0" /> <edge from-layer="1344" from-port="1" to-layer="1346" to-port="0" /> <edge from-layer="1345" from-port="0" to-layer="1346" to-port="1" /> <edge from-layer="1346" from-port="2" to-layer="1348" to-port="2" /> <edge from-layer="1347" from-port="0" to-layer="1348" to-port="3" /> <edge from-layer="1348" from-port="4" to-layer="1351" to-port="1" /> <edge from-layer="1348" from-port="4" to-layer="1356" to-port="2" /> <edge from-layer="1349" from-port="0" to-layer="1351" to-port="2" /> <edge from-layer="1350" from-port="0" to-layer="1351" to-port="3" /> <edge from-layer="1351" from-port="4" to-layer="1353" to-port="0" /> <edge from-layer="1352" from-port="0" to-layer="1353" to-port="1" /> <edge from-layer="1353" from-port="2" to-layer="1357" to-port="0" /> <edge from-layer="1354" from-port="0" to-layer="1356" to-port="1" /> <edge from-layer="1355" from-port="0" to-layer="1356" to-port="3" /> <edge from-layer="1356" from-port="4" to-layer="1357" to-port="1" /> <edge from-layer="1357" from-port="2" to-layer="1358" to-port="0" /> <edge from-layer="1358" from-port="2" to-layer="1359" to-port="1" /> <edge from-layer="1359" from-port="2" to-layer="1364" to-port="0" /> <edge from-layer="1360" from-port="0" to-layer="1363" to-port="1" /> <edge from-layer="1361" from-port="0" to-layer="1363" to-port="2" /> <edge from-layer="1362" from-port="0" to-layer="1363" to-port="3" /> <edge from-layer="1363" from-port="4" to-layer="1364" to-port="1" /> <edge from-layer="1364" from-port="2" to-layer="1365" to-port="1" /> <edge from-layer="1365" from-port="2" to-layer="2241" to-port="0" /> <edge from-layer="1365" from-port="2" to-layer="1426" to-port="1" /> <edge from-layer="1365" from-port="2" to-layer="1385" to-port="0" /> <edge from-layer="1366" from-port="0" to-layer="1371" to-port="0" /> <edge from-layer="1367" from-port="0" to-layer="1370" to-port="1" /> <edge from-layer="1368" from-port="0" to-layer="1370" to-port="2" /> <edge from-layer="1369" from-port="0" to-layer="1370" to-port="3" /> <edge from-layer="1370" from-port="4" to-layer="1371" to-port="1" /> <edge from-layer="1371" from-port="2" to-layer="1372" to-port="0" /> <edge from-layer="1372" from-port="1" to-layer="1374" to-port="0" /> <edge from-layer="1373" from-port="0" to-layer="1374" to-port="2" /> <edge from-layer="1374" from-port="3" to-layer="1381" to-port="0" /> <edge from-layer="1375" from-port="0" to-layer="1378" to-port="1" /> <edge from-layer="1376" from-port="0" to-layer="1378" to-port="2" /> <edge from-layer="1377" from-port="0" to-layer="1378" to-port="3" /> <edge from-layer="1378" from-port="4" to-layer="1380" to-port="0" /> <edge from-layer="1379" from-port="0" to-layer="1380" to-port="1" /> <edge from-layer="1380" from-port="2" to-layer="1381" to-port="1" /> <edge from-layer="1381" from-port="2" to-layer="2240" to-port="0" /> <edge from-layer="1381" from-port="2" to-layer="1426" to-port="2" /> <edge from-layer="1382" from-port="0" to-layer="1400" to-port="0" /> <edge from-layer="1383" from-port="0" to-layer="1395" to-port="0" /> <edge from-layer="1384" from-port="0" to-layer="1398" to-port="1" /> <edge from-layer="1384" from-port="0" to-layer="1395" to-port="1" /> <edge from-layer="1385" from-port="1" to-layer="1387" to-port="0" /> <edge from-layer="1386" from-port="0" to-layer="1387" to-port="2" /> <edge from-layer="1387" from-port="3" to-layer="1397" to-port="0" /> <edge from-layer="1387" from-port="3" to-layer="1391" to-port="0" /> <edge from-layer="1388" from-port="1" to-layer="1390" to-port="0" /> <edge from-layer="1388" from-port="1" to-layer="1415" to-port="0" /> <edge from-layer="1389" from-port="0" to-layer="1390" to-port="2" /> <edge from-layer="1390" from-port="3" to-layer="1391" to-port="1" /> <edge from-layer="1391" from-port="2" to-layer="1393" to-port="0" /> <edge from-layer="1392" from-port="0" to-layer="1393" to-port="1" /> <edge from-layer="1392" from-port="0" to-layer="1397" to-port="1" /> <edge from-layer="1393" from-port="2" to-layer="1395" to-port="2" /> <edge from-layer="1394" from-port="0" to-layer="1395" to-port="3" /> <edge from-layer="1394" from-port="0" to-layer="1398" to-port="3" /> <edge from-layer="1395" from-port="4" to-layer="1400" to-port="1" /> <edge from-layer="1396" from-port="0" to-layer="1398" to-port="0" /> <edge from-layer="1397" from-port="2" to-layer="1398" to-port="2" /> <edge from-layer="1397" from-port="2" to-layer="1405" to-port="2" /> <edge from-layer="1398" from-port="4" to-layer="1400" to-port="2" /> <edge from-layer="1399" from-port="0" to-layer="1400" to-port="3" /> <edge from-layer="1400" from-port="4" to-layer="1407" to-port="0" /> <edge from-layer="1401" from-port="0" to-layer="1407" to-port="1" /> <edge from-layer="1402" from-port="0" to-layer="1405" to-port="0" /> <edge from-layer="1403" from-port="0" to-layer="1405" to-port="1" /> <edge from-layer="1404" from-port="0" to-layer="1405" to-port="3" /> <edge from-layer="1405" from-port="4" to-layer="1407" to-port="2" /> <edge from-layer="1406" from-port="0" to-layer="1407" to-port="3" /> <edge from-layer="1407" from-port="4" to-layer="1410" to-port="0" /> <edge from-layer="1408" from-port="0" to-layer="1410" to-port="1" /> <edge from-layer="1409" from-port="0" to-layer="1410" to-port="2" /> <edge from-layer="1410" from-port="3" to-layer="1411" to-port="0" /> <edge from-layer="1411" from-port="2" to-layer="1425" to-port="0" /> <edge from-layer="1411" from-port="2" to-layer="1412" to-port="0" /> <edge from-layer="1412" from-port="1" to-layer="1424" to-port="0" /> <edge from-layer="1413" from-port="0" to-layer="1415" to-port="1" /> <edge from-layer="1414" from-port="0" to-layer="1415" to-port="2" /> <edge from-layer="1415" from-port="3" to-layer="1419" to-port="0" /> <edge from-layer="1416" from-port="0" to-layer="1419" to-port="1" /> <edge from-layer="1417" from-port="0" to-layer="1419" to-port="2" /> <edge from-layer="1418" from-port="0" to-layer="1419" to-port="3" /> <edge from-layer="1419" from-port="4" to-layer="1421" to-port="0" /> <edge from-layer="1419" from-port="4" to-layer="1423" to-port="2" /> <edge from-layer="1420" from-port="0" to-layer="1421" to-port="1" /> <edge from-layer="1421" from-port="2" to-layer="1423" to-port="0" /> <edge from-layer="1422" from-port="0" to-layer="1423" to-port="1" /> <edge from-layer="1423" from-port="3" to-layer="1424" to-port="1" /> <edge from-layer="1424" from-port="2" to-layer="1425" to-port="1" /> <edge from-layer="1425" from-port="2" to-layer="1426" to-port="3" /> <edge from-layer="1426" from-port="4" to-layer="1428" to-port="0" /> <edge from-layer="1427" from-port="0" to-layer="1428" to-port="1" /> <edge from-layer="1428" from-port="2" to-layer="1430" to-port="0" /> <edge from-layer="1429" from-port="0" to-layer="1430" to-port="1" /> <edge from-layer="1430" from-port="2" to-layer="1438" to-port="0" /> <edge from-layer="1431" from-port="0" to-layer="1432" to-port="0" /> <edge from-layer="1432" from-port="1" to-layer="1435" to-port="0" /> <edge from-layer="1433" from-port="0" to-layer="1434" to-port="0" /> <edge from-layer="1434" from-port="1" to-layer="1435" to-port="1" /> <edge from-layer="1435" from-port="2" to-layer="1437" to-port="0" /> <edge from-layer="1436" from-port="0" to-layer="1437" to-port="1" /> <edge from-layer="1437" from-port="2" to-layer="1438" to-port="1" /> <edge from-layer="1438" from-port="2" to-layer="1440" to-port="0" /> <edge from-layer="1439" from-port="0" to-layer="1440" to-port="1" /> <edge from-layer="1440" from-port="2" to-layer="1441" to-port="1" /> <edge from-layer="1441" from-port="2" to-layer="1442" to-port="0" /> <edge from-layer="1442" from-port="2" to-layer="1444" to-port="0" /> <edge from-layer="1442" from-port="2" to-layer="1673" to-port="1" /> <edge from-layer="1442" from-port="2" to-layer="1471" to-port="0" /> <edge from-layer="1443" from-port="0" to-layer="1444" to-port="1" /> <edge from-layer="1444" from-port="2" to-layer="1446" to-port="0" /> <edge from-layer="1445" from-port="0" to-layer="1446" to-port="1" /> <edge from-layer="1446" from-port="2" to-layer="1448" to-port="0" /> <edge from-layer="1447" from-port="0" to-layer="1448" to-port="1" /> <edge from-layer="1448" from-port="2" to-layer="1456" to-port="0" /> <edge from-layer="1449" from-port="0" to-layer="1450" to-port="0" /> <edge from-layer="1450" from-port="1" to-layer="1453" to-port="0" /> <edge from-layer="1451" from-port="0" to-layer="1452" to-port="0" /> <edge from-layer="1452" from-port="1" to-layer="1453" to-port="1" /> <edge from-layer="1453" from-port="2" to-layer="1455" to-port="0" /> <edge from-layer="1454" from-port="0" to-layer="1455" to-port="1" /> <edge from-layer="1455" from-port="2" to-layer="1456" to-port="1" /> <edge from-layer="1456" from-port="2" to-layer="1458" to-port="0" /> <edge from-layer="1457" from-port="0" to-layer="1458" to-port="1" /> <edge from-layer="1458" from-port="2" to-layer="1459" to-port="0" /> <edge from-layer="1459" from-port="1" to-layer="1467" to-port="0" /> <edge from-layer="1460" from-port="0" to-layer="1461" to-port="0" /> <edge from-layer="1461" from-port="1" to-layer="1464" to-port="0" /> <edge from-layer="1462" from-port="0" to-layer="1463" to-port="0" /> <edge from-layer="1463" from-port="1" to-layer="1464" to-port="1" /> <edge from-layer="1464" from-port="2" to-layer="1466" to-port="0" /> <edge from-layer="1465" from-port="0" to-layer="1466" to-port="1" /> <edge from-layer="1466" from-port="2" to-layer="1467" to-port="1" /> <edge from-layer="1467" from-port="2" to-layer="1469" to-port="0" /> <edge from-layer="1468" from-port="0" to-layer="1469" to-port="1" /> <edge from-layer="1469" from-port="2" to-layer="1672" to-port="0" /> <edge from-layer="1470" from-port="0" to-layer="1471" to-port="1" /> <edge from-layer="1471" from-port="2" to-layer="1473" to-port="0" /> <edge from-layer="1472" from-port="0" to-layer="1473" to-port="1" /> <edge from-layer="1473" from-port="2" to-layer="1475" to-port="0" /> <edge from-layer="1474" from-port="0" to-layer="1475" to-port="1" /> <edge from-layer="1475" from-port="2" to-layer="1483" to-port="0" /> <edge from-layer="1476" from-port="0" to-layer="1477" to-port="0" /> <edge from-layer="1477" from-port="1" to-layer="1480" to-port="0" /> <edge from-layer="1478" from-port="0" to-layer="1479" to-port="0" /> <edge from-layer="1479" from-port="1" to-layer="1480" to-port="1" /> <edge from-layer="1480" from-port="2" to-layer="1482" to-port="0" /> <edge from-layer="1481" from-port="0" to-layer="1482" to-port="1" /> <edge from-layer="1482" from-port="2" to-layer="1483" to-port="1" /> <edge from-layer="1483" from-port="2" to-layer="1485" to-port="0" /> <edge from-layer="1484" from-port="0" to-layer="1485" to-port="1" /> <edge from-layer="1485" from-port="2" to-layer="1487" to-port="0" /> <edge from-layer="1486" from-port="0" to-layer="1487" to-port="1" /> <edge from-layer="1487" from-port="2" to-layer="1503" to-port="0" /> <edge from-layer="1487" from-port="2" to-layer="1609" to-port="0" /> <edge from-layer="1487" from-port="2" to-layer="1491" to-port="0" /> <edge from-layer="1488" from-port="0" to-layer="1491" to-port="1" /> <edge from-layer="1489" from-port="0" to-layer="1491" to-port="2" /> <edge from-layer="1490" from-port="0" to-layer="1491" to-port="3" /> <edge from-layer="1491" from-port="4" to-layer="1493" to-port="0" /> <edge from-layer="1492" from-port="0" to-layer="1493" to-port="1" /> <edge from-layer="1493" from-port="2" to-layer="1562" to-port="0" /> <edge from-layer="1493" from-port="2" to-layer="1497" to-port="0" /> <edge from-layer="1494" from-port="0" to-layer="1497" to-port="1" /> <edge from-layer="1495" from-port="0" to-layer="1497" to-port="2" /> <edge from-layer="1496" from-port="0" to-layer="1497" to-port="3" /> <edge from-layer="1497" from-port="4" to-layer="1548" to-port="0" /> <edge from-layer="1497" from-port="4" to-layer="1529" to-port="0" /> <edge from-layer="1497" from-port="4" to-layer="1532" to-port="0" /> <edge from-layer="1497" from-port="4" to-layer="1543" to-port="0" /> <edge from-layer="1498" from-port="0" to-layer="1525" to-port="0" /> <edge from-layer="1499" from-port="0" to-layer="1525" to-port="1" /> <edge from-layer="1500" from-port="0" to-layer="1503" to-port="1" /> <edge from-layer="1501" from-port="0" to-layer="1503" to-port="2" /> <edge from-layer="1502" from-port="0" to-layer="1503" to-port="3" /> <edge from-layer="1503" from-port="4" to-layer="1505" to-port="0" /> <edge from-layer="1504" from-port="0" to-layer="1505" to-port="1" /> <edge from-layer="1505" from-port="2" to-layer="1506" to-port="0" /> <edge from-layer="1505" from-port="2" to-layer="1567" to-port="0" /> <edge from-layer="1505" from-port="2" to-layer="1594" to-port="0" /> <edge from-layer="1506" from-port="1" to-layer="1508" to-port="0" /> <edge from-layer="1507" from-port="0" to-layer="1508" to-port="2" /> <edge from-layer="1508" from-port="3" to-layer="1521" to-port="0" /> <edge from-layer="1509" from-port="0" to-layer="1514" to-port="0" /> <edge from-layer="1510" from-port="0" to-layer="1513" to-port="1" /> <edge from-layer="1511" from-port="0" to-layer="1513" to-port="2" /> <edge from-layer="1512" from-port="0" to-layer="1513" to-port="3" /> <edge from-layer="1513" from-port="4" to-layer="1514" to-port="1" /> <edge from-layer="1514" from-port="2" to-layer="1515" to-port="0" /> <edge from-layer="1515" from-port="1" to-layer="1517" to-port="0" /> <edge from-layer="1516" from-port="0" to-layer="1517" to-port="2" /> <edge from-layer="1517" from-port="3" to-layer="1596" to-port="0" /> <edge from-layer="1517" from-port="3" to-layer="1518" to-port="0" /> <edge from-layer="1518" from-port="1" to-layer="1520" to-port="0" /> <edge from-layer="1519" from-port="0" to-layer="1520" to-port="2" /> <edge from-layer="1520" from-port="3" to-layer="1521" to-port="1" /> <edge from-layer="1521" from-port="2" to-layer="1523" to-port="0" /> <edge from-layer="1522" from-port="0" to-layer="1523" to-port="1" /> <edge from-layer="1523" from-port="2" to-layer="1525" to-port="2" /> <edge from-layer="1523" from-port="2" to-layer="1553" to-port="2" /> <edge from-layer="1524" from-port="0" to-layer="1525" to-port="3" /> <edge from-layer="1525" from-port="4" to-layer="1527" to-port="0" /> <edge from-layer="1526" from-port="0" to-layer="1527" to-port="2" /> <edge from-layer="1527" from-port="3" to-layer="1528" to-port="0" /> <edge from-layer="1528" from-port="2" to-layer="1568" to-port="1" /> <edge from-layer="1528" from-port="2" to-layer="1529" to-port="1" /> <edge from-layer="1529" from-port="2" to-layer="1558" to-port="0" /> <edge from-layer="1530" from-port="0" to-layer="1540" to-port="0" /> <edge from-layer="1531" from-port="0" to-layer="1540" to-port="1" /> <edge from-layer="1532" from-port="1" to-layer="1534" to-port="0" /> <edge from-layer="1533" from-port="0" to-layer="1534" to-port="2" /> <edge from-layer="1534" from-port="3" to-layer="1535" to-port="0" /> <edge from-layer="1535" from-port="2" to-layer="1536" to-port="0" /> <edge from-layer="1536" from-port="1" to-layer="1538" to-port="0" /> <edge from-layer="1537" from-port="0" to-layer="1538" to-port="1" /> <edge from-layer="1538" from-port="2" to-layer="1540" to-port="2" /> <edge from-layer="1539" from-port="0" to-layer="1540" to-port="3" /> <edge from-layer="1540" from-port="4" to-layer="1548" to-port="2" /> <edge from-layer="1540" from-port="4" to-layer="1543" to-port="1" /> <edge from-layer="1541" from-port="0" to-layer="1543" to-port="2" /> <edge from-layer="1542" from-port="0" to-layer="1543" to-port="3" /> <edge from-layer="1543" from-port="4" to-layer="1545" to-port="0" /> <edge from-layer="1544" from-port="0" to-layer="1545" to-port="1" /> <edge from-layer="1545" from-port="2" to-layer="1549" to-port="0" /> <edge from-layer="1546" from-port="0" to-layer="1548" to-port="1" /> <edge from-layer="1547" from-port="0" to-layer="1548" to-port="3" /> <edge from-layer="1548" from-port="4" to-layer="1549" to-port="1" /> <edge from-layer="1549" from-port="2" to-layer="1557" to-port="0" /> <edge from-layer="1550" from-port="0" to-layer="1553" to-port="0" /> <edge from-layer="1551" from-port="0" to-layer="1553" to-port="1" /> <edge from-layer="1552" from-port="0" to-layer="1553" to-port="3" /> <edge from-layer="1553" from-port="4" to-layer="1555" to-port="0" /> <edge from-layer="1554" from-port="0" to-layer="1555" to-port="2" /> <edge from-layer="1555" from-port="3" to-layer="1556" to-port="0" /> <edge from-layer="1556" from-port="2" to-layer="1589" to-port="1" /> <edge from-layer="1556" from-port="2" to-layer="1557" to-port="1" /> <edge from-layer="1557" from-port="2" to-layer="1558" to-port="1" /> <edge from-layer="1558" from-port="2" to-layer="1563" to-port="0" /> <edge from-layer="1559" from-port="0" to-layer="1562" to-port="1" /> <edge from-layer="1560" from-port="0" to-layer="1562" to-port="2" /> <edge from-layer="1561" from-port="0" to-layer="1562" to-port="3" /> <edge from-layer="1562" from-port="4" to-layer="1563" to-port="1" /> <edge from-layer="1563" from-port="2" to-layer="1619" to-port="0" /> <edge from-layer="1563" from-port="2" to-layer="1657" to-port="0" /> <edge from-layer="1564" from-port="0" to-layer="1567" to-port="1" /> <edge from-layer="1565" from-port="0" to-layer="1567" to-port="2" /> <edge from-layer="1566" from-port="0" to-layer="1567" to-port="3" /> <edge from-layer="1567" from-port="4" to-layer="1582" to-port="0" /> <edge from-layer="1567" from-port="4" to-layer="1587" to-port="0" /> <edge from-layer="1567" from-port="4" to-layer="1568" to-port="0" /> <edge from-layer="1567" from-port="4" to-layer="1571" to-port="0" /> <edge from-layer="1568" from-port="2" to-layer="1590" to-port="0" /> <edge from-layer="1569" from-port="0" to-layer="1579" to-port="0" /> <edge from-layer="1570" from-port="0" to-layer="1579" to-port="1" /> <edge from-layer="1571" from-port="1" to-layer="1573" to-port="0" /> <edge from-layer="1572" from-port="0" to-layer="1573" to-port="2" /> <edge from-layer="1573" from-port="3" to-layer="1574" to-port="0" /> <edge from-layer="1574" from-port="2" to-layer="1575" to-port="0" /> <edge from-layer="1575" from-port="1" to-layer="1577" to-port="0" /> <edge from-layer="1576" from-port="0" to-layer="1577" to-port="1" /> <edge from-layer="1577" from-port="2" to-layer="1579" to-port="2" /> <edge from-layer="1578" from-port="0" to-layer="1579" to-port="3" /> <edge from-layer="1579" from-port="4" to-layer="1587" to-port="2" /> <edge from-layer="1579" from-port="4" to-layer="1582" to-port="1" /> <edge from-layer="1580" from-port="0" to-layer="1582" to-port="2" /> <edge from-layer="1581" from-port="0" to-layer="1582" to-port="3" /> <edge from-layer="1582" from-port="4" to-layer="1584" to-port="0" /> <edge from-layer="1583" from-port="0" to-layer="1584" to-port="1" /> <edge from-layer="1584" from-port="2" to-layer="1588" to-port="0" /> <edge from-layer="1585" from-port="0" to-layer="1587" to-port="1" /> <edge from-layer="1586" from-port="0" to-layer="1587" to-port="3" /> <edge from-layer="1587" from-port="4" to-layer="1588" to-port="1" /> <edge from-layer="1588" from-port="2" to-layer="1589" to-port="0" /> <edge from-layer="1589" from-port="2" to-layer="1590" to-port="1" /> <edge from-layer="1590" from-port="2" to-layer="1595" to-port="0" /> <edge from-layer="1591" from-port="0" to-layer="1594" to-port="1" /> <edge from-layer="1592" from-port="0" to-layer="1594" to-port="2" /> <edge from-layer="1593" from-port="0" to-layer="1594" to-port="3" /> <edge from-layer="1594" from-port="4" to-layer="1595" to-port="1" /> <edge from-layer="1595" from-port="2" to-layer="1596" to-port="1" /> <edge from-layer="1596" from-port="2" to-layer="1616" to-port="0" /> <edge from-layer="1596" from-port="2" to-layer="2239" to-port="0" /> <edge from-layer="1596" from-port="2" to-layer="1657" to-port="1" /> <edge from-layer="1597" from-port="0" to-layer="1602" to-port="0" /> <edge from-layer="1598" from-port="0" to-layer="1601" to-port="1" /> <edge from-layer="1599" from-port="0" to-layer="1601" to-port="2" /> <edge from-layer="1600" from-port="0" to-layer="1601" to-port="3" /> <edge from-layer="1601" from-port="4" to-layer="1602" to-port="1" /> <edge from-layer="1602" from-port="2" to-layer="1603" to-port="0" /> <edge from-layer="1603" from-port="1" to-layer="1605" to-port="0" /> <edge from-layer="1604" from-port="0" to-layer="1605" to-port="2" /> <edge from-layer="1605" from-port="3" to-layer="1612" to-port="0" /> <edge from-layer="1606" from-port="0" to-layer="1609" to-port="1" /> <edge from-layer="1607" from-port="0" to-layer="1609" to-port="2" /> <edge from-layer="1608" from-port="0" to-layer="1609" to-port="3" /> <edge from-layer="1609" from-port="4" to-layer="1611" to-port="0" /> <edge from-layer="1610" from-port="0" to-layer="1611" to-port="1" /> <edge from-layer="1611" from-port="2" to-layer="1612" to-port="1" /> <edge from-layer="1612" from-port="2" to-layer="1657" to-port="2" /> <edge from-layer="1612" from-port="2" to-layer="2238" to-port="0" /> <edge from-layer="1613" from-port="0" to-layer="1631" to-port="0" /> <edge from-layer="1614" from-port="0" to-layer="1626" to-port="0" /> <edge from-layer="1615" from-port="0" to-layer="1626" to-port="1" /> <edge from-layer="1615" from-port="0" to-layer="1629" to-port="1" /> <edge from-layer="1616" from-port="1" to-layer="1618" to-port="0" /> <edge from-layer="1617" from-port="0" to-layer="1618" to-port="2" /> <edge from-layer="1618" from-port="3" to-layer="1628" to-port="0" /> <edge from-layer="1618" from-port="3" to-layer="1622" to-port="0" /> <edge from-layer="1619" from-port="1" to-layer="1646" to-port="0" /> <edge from-layer="1619" from-port="1" to-layer="1621" to-port="0" /> <edge from-layer="1620" from-port="0" to-layer="1621" to-port="2" /> <edge from-layer="1621" from-port="3" to-layer="1622" to-port="1" /> <edge from-layer="1622" from-port="2" to-layer="1624" to-port="0" /> <edge from-layer="1623" from-port="0" to-layer="1628" to-port="1" /> <edge from-layer="1623" from-port="0" to-layer="1624" to-port="1" /> <edge from-layer="1624" from-port="2" to-layer="1626" to-port="2" /> <edge from-layer="1625" from-port="0" to-layer="1629" to-port="3" /> <edge from-layer="1625" from-port="0" to-layer="1626" to-port="3" /> <edge from-layer="1626" from-port="4" to-layer="1631" to-port="1" /> <edge from-layer="1627" from-port="0" to-layer="1629" to-port="0" /> <edge from-layer="1628" from-port="2" to-layer="1629" to-port="2" /> <edge from-layer="1628" from-port="2" to-layer="1636" to-port="2" /> <edge from-layer="1629" from-port="4" to-layer="1631" to-port="2" /> <edge from-layer="1630" from-port="0" to-layer="1631" to-port="3" /> <edge from-layer="1631" from-port="4" to-layer="1638" to-port="0" /> <edge from-layer="1632" from-port="0" to-layer="1638" to-port="1" /> <edge from-layer="1633" from-port="0" to-layer="1636" to-port="0" /> <edge from-layer="1634" from-port="0" to-layer="1636" to-port="1" /> <edge from-layer="1635" from-port="0" to-layer="1636" to-port="3" /> <edge from-layer="1636" from-port="4" to-layer="1638" to-port="2" /> <edge from-layer="1637" from-port="0" to-layer="1638" to-port="3" /> <edge from-layer="1638" from-port="4" to-layer="1641" to-port="0" /> <edge from-layer="1639" from-port="0" to-layer="1641" to-port="1" /> <edge from-layer="1640" from-port="0" to-layer="1641" to-port="2" /> <edge from-layer="1641" from-port="3" to-layer="1642" to-port="0" /> <edge from-layer="1642" from-port="2" to-layer="1643" to-port="0" /> <edge from-layer="1642" from-port="2" to-layer="1656" to-port="0" /> <edge from-layer="1643" from-port="1" to-layer="1655" to-port="0" /> <edge from-layer="1644" from-port="0" to-layer="1646" to-port="1" /> <edge from-layer="1645" from-port="0" to-layer="1646" to-port="2" /> <edge from-layer="1646" from-port="3" to-layer="1650" to-port="0" /> <edge from-layer="1647" from-port="0" to-layer="1650" to-port="1" /> <edge from-layer="1648" from-port="0" to-layer="1650" to-port="2" /> <edge from-layer="1649" from-port="0" to-layer="1650" to-port="3" /> <edge from-layer="1650" from-port="4" to-layer="1652" to-port="0" /> <edge from-layer="1650" from-port="4" to-layer="1654" to-port="2" /> <edge from-layer="1651" from-port="0" to-layer="1652" to-port="1" /> <edge from-layer="1652" from-port="2" to-layer="1654" to-port="0" /> <edge from-layer="1653" from-port="0" to-layer="1654" to-port="1" /> <edge from-layer="1654" from-port="3" to-layer="1655" to-port="1" /> <edge from-layer="1655" from-port="2" to-layer="1656" to-port="1" /> <edge from-layer="1656" from-port="2" to-layer="1657" to-port="3" /> <edge from-layer="1657" from-port="4" to-layer="1659" to-port="0" /> <edge from-layer="1658" from-port="0" to-layer="1659" to-port="1" /> <edge from-layer="1659" from-port="2" to-layer="1661" to-port="0" /> <edge from-layer="1660" from-port="0" to-layer="1661" to-port="1" /> <edge from-layer="1661" from-port="2" to-layer="1669" to-port="0" /> <edge from-layer="1662" from-port="0" to-layer="1663" to-port="0" /> <edge from-layer="1663" from-port="1" to-layer="1666" to-port="0" /> <edge from-layer="1664" from-port="0" to-layer="1665" to-port="0" /> <edge from-layer="1665" from-port="1" to-layer="1666" to-port="1" /> <edge from-layer="1666" from-port="2" to-layer="1668" to-port="0" /> <edge from-layer="1667" from-port="0" to-layer="1668" to-port="1" /> <edge from-layer="1668" from-port="2" to-layer="1669" to-port="1" /> <edge from-layer="1669" from-port="2" to-layer="1671" to-port="0" /> <edge from-layer="1670" from-port="0" to-layer="1671" to-port="1" /> <edge from-layer="1671" from-port="2" to-layer="1672" to-port="1" /> <edge from-layer="1672" from-port="2" to-layer="1673" to-port="0" /> <edge from-layer="1673" from-port="2" to-layer="1675" to-port="0" /> <edge from-layer="1673" from-port="2" to-layer="1702" to-port="0" /> <edge from-layer="1673" from-port="2" to-layer="1904" to-port="1" /> <edge from-layer="1674" from-port="0" to-layer="1675" to-port="1" /> <edge from-layer="1675" from-port="2" to-layer="1677" to-port="0" /> <edge from-layer="1676" from-port="0" to-layer="1677" to-port="1" /> <edge from-layer="1677" from-port="2" to-layer="1679" to-port="0" /> <edge from-layer="1678" from-port="0" to-layer="1679" to-port="1" /> <edge from-layer="1679" from-port="2" to-layer="1687" to-port="0" /> <edge from-layer="1680" from-port="0" to-layer="1681" to-port="0" /> <edge from-layer="1681" from-port="1" to-layer="1684" to-port="0" /> <edge from-layer="1682" from-port="0" to-layer="1683" to-port="0" /> <edge from-layer="1683" from-port="1" to-layer="1684" to-port="1" /> <edge from-layer="1684" from-port="2" to-layer="1686" to-port="0" /> <edge from-layer="1685" from-port="0" to-layer="1686" to-port="1" /> <edge from-layer="1686" from-port="2" to-layer="1687" to-port="1" /> <edge from-layer="1687" from-port="2" to-layer="1689" to-port="0" /> <edge from-layer="1688" from-port="0" to-layer="1689" to-port="1" /> <edge from-layer="1689" from-port="2" to-layer="1690" to-port="0" /> <edge from-layer="1690" from-port="1" to-layer="1698" to-port="0" /> <edge from-layer="1691" from-port="0" to-layer="1692" to-port="0" /> <edge from-layer="1692" from-port="1" to-layer="1695" to-port="0" /> <edge from-layer="1693" from-port="0" to-layer="1694" to-port="0" /> <edge from-layer="1694" from-port="1" to-layer="1695" to-port="1" /> <edge from-layer="1695" from-port="2" to-layer="1697" to-port="0" /> <edge from-layer="1696" from-port="0" to-layer="1697" to-port="1" /> <edge from-layer="1697" from-port="2" to-layer="1698" to-port="1" /> <edge from-layer="1698" from-port="2" to-layer="1700" to-port="0" /> <edge from-layer="1699" from-port="0" to-layer="1700" to-port="1" /> <edge from-layer="1700" from-port="2" to-layer="1903" to-port="0" /> <edge from-layer="1701" from-port="0" to-layer="1702" to-port="1" /> <edge from-layer="1702" from-port="2" to-layer="1704" to-port="0" /> <edge from-layer="1703" from-port="0" to-layer="1704" to-port="1" /> <edge from-layer="1704" from-port="2" to-layer="1706" to-port="0" /> <edge from-layer="1705" from-port="0" to-layer="1706" to-port="1" /> <edge from-layer="1706" from-port="2" to-layer="1714" to-port="0" /> <edge from-layer="1707" from-port="0" to-layer="1708" to-port="0" /> <edge from-layer="1708" from-port="1" to-layer="1711" to-port="0" /> <edge from-layer="1709" from-port="0" to-layer="1710" to-port="0" /> <edge from-layer="1710" from-port="1" to-layer="1711" to-port="1" /> <edge from-layer="1711" from-port="2" to-layer="1713" to-port="0" /> <edge from-layer="1712" from-port="0" to-layer="1713" to-port="1" /> <edge from-layer="1713" from-port="2" to-layer="1714" to-port="1" /> <edge from-layer="1714" from-port="2" to-layer="1716" to-port="0" /> <edge from-layer="1715" from-port="0" to-layer="1716" to-port="1" /> <edge from-layer="1716" from-port="2" to-layer="1718" to-port="0" /> <edge from-layer="1717" from-port="0" to-layer="1718" to-port="1" /> <edge from-layer="1718" from-port="2" to-layer="1734" to-port="0" /> <edge from-layer="1718" from-port="2" to-layer="1840" to-port="0" /> <edge from-layer="1718" from-port="2" to-layer="1722" to-port="0" /> <edge from-layer="1719" from-port="0" to-layer="1722" to-port="1" /> <edge from-layer="1720" from-port="0" to-layer="1722" to-port="2" /> <edge from-layer="1721" from-port="0" to-layer="1722" to-port="3" /> <edge from-layer="1722" from-port="4" to-layer="1724" to-port="0" /> <edge from-layer="1723" from-port="0" to-layer="1724" to-port="1" /> <edge from-layer="1724" from-port="2" to-layer="1728" to-port="0" /> <edge from-layer="1724" from-port="2" to-layer="1793" to-port="0" /> <edge from-layer="1725" from-port="0" to-layer="1728" to-port="1" /> <edge from-layer="1726" from-port="0" to-layer="1728" to-port="2" /> <edge from-layer="1727" from-port="0" to-layer="1728" to-port="3" /> <edge from-layer="1728" from-port="4" to-layer="1760" to-port="0" /> <edge from-layer="1728" from-port="4" to-layer="1763" to-port="0" /> <edge from-layer="1728" from-port="4" to-layer="1779" to-port="0" /> <edge from-layer="1728" from-port="4" to-layer="1774" to-port="0" /> <edge from-layer="1729" from-port="0" to-layer="1756" to-port="0" /> <edge from-layer="1730" from-port="0" to-layer="1756" to-port="1" /> <edge from-layer="1731" from-port="0" to-layer="1734" to-port="1" /> <edge from-layer="1732" from-port="0" to-layer="1734" to-port="2" /> <edge from-layer="1733" from-port="0" to-layer="1734" to-port="3" /> <edge from-layer="1734" from-port="4" to-layer="1736" to-port="0" /> <edge from-layer="1735" from-port="0" to-layer="1736" to-port="1" /> <edge from-layer="1736" from-port="2" to-layer="1737" to-port="0" /> <edge from-layer="1736" from-port="2" to-layer="1825" to-port="0" /> <edge from-layer="1736" from-port="2" to-layer="1798" to-port="0" /> <edge from-layer="1737" from-port="1" to-layer="1739" to-port="0" /> <edge from-layer="1738" from-port="0" to-layer="1739" to-port="2" /> <edge from-layer="1739" from-port="3" to-layer="1752" to-port="0" /> <edge from-layer="1740" from-port="0" to-layer="1745" to-port="0" /> <edge from-layer="1741" from-port="0" to-layer="1744" to-port="1" /> <edge from-layer="1742" from-port="0" to-layer="1744" to-port="2" /> <edge from-layer="1743" from-port="0" to-layer="1744" to-port="3" /> <edge from-layer="1744" from-port="4" to-layer="1745" to-port="1" /> <edge from-layer="1745" from-port="2" to-layer="1746" to-port="0" /> <edge from-layer="1746" from-port="1" to-layer="1748" to-port="0" /> <edge from-layer="1747" from-port="0" to-layer="1748" to-port="2" /> <edge from-layer="1748" from-port="3" to-layer="1749" to-port="0" /> <edge from-layer="1748" from-port="3" to-layer="1827" to-port="0" /> <edge from-layer="1749" from-port="1" to-layer="1751" to-port="0" /> <edge from-layer="1750" from-port="0" to-layer="1751" to-port="2" /> <edge from-layer="1751" from-port="3" to-layer="1752" to-port="1" /> <edge from-layer="1752" from-port="2" to-layer="1754" to-port="0" /> <edge from-layer="1753" from-port="0" to-layer="1754" to-port="1" /> <edge from-layer="1754" from-port="2" to-layer="1784" to-port="2" /> <edge from-layer="1754" from-port="2" to-layer="1756" to-port="2" /> <edge from-layer="1755" from-port="0" to-layer="1756" to-port="3" /> <edge from-layer="1756" from-port="4" to-layer="1758" to-port="0" /> <edge from-layer="1757" from-port="0" to-layer="1758" to-port="2" /> <edge from-layer="1758" from-port="3" to-layer="1759" to-port="0" /> <edge from-layer="1759" from-port="2" to-layer="1799" to-port="1" /> <edge from-layer="1759" from-port="2" to-layer="1760" to-port="1" /> <edge from-layer="1760" from-port="2" to-layer="1789" to-port="0" /> <edge from-layer="1761" from-port="0" to-layer="1771" to-port="0" /> <edge from-layer="1762" from-port="0" to-layer="1771" to-port="1" /> <edge from-layer="1763" from-port="1" to-layer="1765" to-port="0" /> <edge from-layer="1764" from-port="0" to-layer="1765" to-port="2" /> <edge from-layer="1765" from-port="3" to-layer="1766" to-port="0" /> <edge from-layer="1766" from-port="2" to-layer="1767" to-port="0" /> <edge from-layer="1767" from-port="1" to-layer="1769" to-port="0" /> <edge from-layer="1768" from-port="0" to-layer="1769" to-port="1" /> <edge from-layer="1769" from-port="2" to-layer="1771" to-port="2" /> <edge from-layer="1770" from-port="0" to-layer="1771" to-port="3" /> <edge from-layer="1771" from-port="4" to-layer="1779" to-port="2" /> <edge from-layer="1771" from-port="4" to-layer="1774" to-port="1" /> <edge from-layer="1772" from-port="0" to-layer="1774" to-port="2" /> <edge from-layer="1773" from-port="0" to-layer="1774" to-port="3" /> <edge from-layer="1774" from-port="4" to-layer="1776" to-port="0" /> <edge from-layer="1775" from-port="0" to-layer="1776" to-port="1" /> <edge from-layer="1776" from-port="2" to-layer="1780" to-port="0" /> <edge from-layer="1777" from-port="0" to-layer="1779" to-port="1" /> <edge from-layer="1778" from-port="0" to-layer="1779" to-port="3" /> <edge from-layer="1779" from-port="4" to-layer="1780" to-port="1" /> <edge from-layer="1780" from-port="2" to-layer="1788" to-port="0" /> <edge from-layer="1781" from-port="0" to-layer="1784" to-port="0" /> <edge from-layer="1782" from-port="0" to-layer="1784" to-port="1" /> <edge from-layer="1783" from-port="0" to-layer="1784" to-port="3" /> <edge from-layer="1784" from-port="4" to-layer="1786" to-port="0" /> <edge from-layer="1785" from-port="0" to-layer="1786" to-port="2" /> <edge from-layer="1786" from-port="3" to-layer="1787" to-port="0" /> <edge from-layer="1787" from-port="2" to-layer="1820" to-port="1" /> <edge from-layer="1787" from-port="2" to-layer="1788" to-port="1" /> <edge from-layer="1788" from-port="2" to-layer="1789" to-port="1" /> <edge from-layer="1789" from-port="2" to-layer="1794" to-port="0" /> <edge from-layer="1790" from-port="0" to-layer="1793" to-port="1" /> <edge from-layer="1791" from-port="0" to-layer="1793" to-port="2" /> <edge from-layer="1792" from-port="0" to-layer="1793" to-port="3" /> <edge from-layer="1793" from-port="4" to-layer="1794" to-port="1" /> <edge from-layer="1794" from-port="2" to-layer="1888" to-port="0" /> <edge from-layer="1794" from-port="2" to-layer="1850" to-port="0" /> <edge from-layer="1795" from-port="0" to-layer="1798" to-port="1" /> <edge from-layer="1796" from-port="0" to-layer="1798" to-port="2" /> <edge from-layer="1797" from-port="0" to-layer="1798" to-port="3" /> <edge from-layer="1798" from-port="4" to-layer="1799" to-port="0" /> <edge from-layer="1798" from-port="4" to-layer="1813" to-port="0" /> <edge from-layer="1798" from-port="4" to-layer="1818" to-port="0" /> <edge from-layer="1798" from-port="4" to-layer="1802" to-port="0" /> <edge from-layer="1799" from-port="2" to-layer="1821" to-port="0" /> <edge from-layer="1800" from-port="0" to-layer="1810" to-port="0" /> <edge from-layer="1801" from-port="0" to-layer="1810" to-port="1" /> <edge from-layer="1802" from-port="1" to-layer="1804" to-port="0" /> <edge from-layer="1803" from-port="0" to-layer="1804" to-port="2" /> <edge from-layer="1804" from-port="3" to-layer="1805" to-port="0" /> <edge from-layer="1805" from-port="2" to-layer="1806" to-port="0" /> <edge from-layer="1806" from-port="1" to-layer="1808" to-port="0" /> <edge from-layer="1807" from-port="0" to-layer="1808" to-port="1" /> <edge from-layer="1808" from-port="2" to-layer="1810" to-port="2" /> <edge from-layer="1809" from-port="0" to-layer="1810" to-port="3" /> <edge from-layer="1810" from-port="4" to-layer="1813" to-port="1" /> <edge from-layer="1810" from-port="4" to-layer="1818" to-port="2" /> <edge from-layer="1811" from-port="0" to-layer="1813" to-port="2" /> <edge from-layer="1812" from-port="0" to-layer="1813" to-port="3" /> <edge from-layer="1813" from-port="4" to-layer="1815" to-port="0" /> <edge from-layer="1814" from-port="0" to-layer="1815" to-port="1" /> <edge from-layer="1815" from-port="2" to-layer="1819" to-port="0" /> <edge from-layer="1816" from-port="0" to-layer="1818" to-port="1" /> <edge from-layer="1817" from-port="0" to-layer="1818" to-port="3" /> <edge from-layer="1818" from-port="4" to-layer="1819" to-port="1" /> <edge from-layer="1819" from-port="2" to-layer="1820" to-port="0" /> <edge from-layer="1820" from-port="2" to-layer="1821" to-port="1" /> <edge from-layer="1821" from-port="2" to-layer="1826" to-port="0" /> <edge from-layer="1822" from-port="0" to-layer="1825" to-port="1" /> <edge from-layer="1823" from-port="0" to-layer="1825" to-port="2" /> <edge from-layer="1824" from-port="0" to-layer="1825" to-port="3" /> <edge from-layer="1825" from-port="4" to-layer="1826" to-port="1" /> <edge from-layer="1826" from-port="2" to-layer="1827" to-port="1" /> <edge from-layer="1827" from-port="2" to-layer="1888" to-port="1" /> <edge from-layer="1827" from-port="2" to-layer="2237" to-port="0" /> <edge from-layer="1827" from-port="2" to-layer="1847" to-port="0" /> <edge from-layer="1828" from-port="0" to-layer="1833" to-port="0" /> <edge from-layer="1829" from-port="0" to-layer="1832" to-port="1" /> <edge from-layer="1830" from-port="0" to-layer="1832" to-port="2" /> <edge from-layer="1831" from-port="0" to-layer="1832" to-port="3" /> <edge from-layer="1832" from-port="4" to-layer="1833" to-port="1" /> <edge from-layer="1833" from-port="2" to-layer="1834" to-port="0" /> <edge from-layer="1834" from-port="1" to-layer="1836" to-port="0" /> <edge from-layer="1835" from-port="0" to-layer="1836" to-port="2" /> <edge from-layer="1836" from-port="3" to-layer="1843" to-port="0" /> <edge from-layer="1837" from-port="0" to-layer="1840" to-port="1" /> <edge from-layer="1838" from-port="0" to-layer="1840" to-port="2" /> <edge from-layer="1839" from-port="0" to-layer="1840" to-port="3" /> <edge from-layer="1840" from-port="4" to-layer="1842" to-port="0" /> <edge from-layer="1841" from-port="0" to-layer="1842" to-port="1" /> <edge from-layer="1842" from-port="2" to-layer="1843" to-port="1" /> <edge from-layer="1843" from-port="2" to-layer="2236" to-port="0" /> <edge from-layer="1843" from-port="2" to-layer="1888" to-port="2" /> <edge from-layer="1844" from-port="0" to-layer="1862" to-port="0" /> <edge from-layer="1845" from-port="0" to-layer="1857" to-port="0" /> <edge from-layer="1846" from-port="0" to-layer="1860" to-port="1" /> <edge from-layer="1846" from-port="0" to-layer="1857" to-port="1" /> <edge from-layer="1847" from-port="1" to-layer="1849" to-port="0" /> <edge from-layer="1848" from-port="0" to-layer="1849" to-port="2" /> <edge from-layer="1849" from-port="3" to-layer="1859" to-port="0" /> <edge from-layer="1849" from-port="3" to-layer="1853" to-port="0" /> <edge from-layer="1850" from-port="1" to-layer="1877" to-port="0" /> <edge from-layer="1850" from-port="1" to-layer="1852" to-port="0" /> <edge from-layer="1851" from-port="0" to-layer="1852" to-port="2" /> <edge from-layer="1852" from-port="3" to-layer="1853" to-port="1" /> <edge from-layer="1853" from-port="2" to-layer="1855" to-port="0" /> <edge from-layer="1854" from-port="0" to-layer="1855" to-port="1" /> <edge from-layer="1854" from-port="0" to-layer="1859" to-port="1" /> <edge from-layer="1855" from-port="2" to-layer="1857" to-port="2" /> <edge from-layer="1856" from-port="0" to-layer="1857" to-port="3" /> <edge from-layer="1856" from-port="0" to-layer="1860" to-port="3" /> <edge from-layer="1857" from-port="4" to-layer="1862" to-port="1" /> <edge from-layer="1858" from-port="0" to-layer="1860" to-port="0" /> <edge from-layer="1859" from-port="2" to-layer="1867" to-port="2" /> <edge from-layer="1859" from-port="2" to-layer="1860" to-port="2" /> <edge from-layer="1860" from-port="4" to-layer="1862" to-port="2" /> <edge from-layer="1861" from-port="0" to-layer="1862" to-port="3" /> <edge from-layer="1862" from-port="4" to-layer="1869" to-port="0" /> <edge from-layer="1863" from-port="0" to-layer="1869" to-port="1" /> <edge from-layer="1864" from-port="0" to-layer="1867" to-port="0" /> <edge from-layer="1865" from-port="0" to-layer="1867" to-port="1" /> <edge from-layer="1866" from-port="0" to-layer="1867" to-port="3" /> <edge from-layer="1867" from-port="4" to-layer="1869" to-port="2" /> <edge from-layer="1868" from-port="0" to-layer="1869" to-port="3" /> <edge from-layer="1869" from-port="4" to-layer="1872" to-port="0" /> <edge from-layer="1870" from-port="0" to-layer="1872" to-port="1" /> <edge from-layer="1871" from-port="0" to-layer="1872" to-port="2" /> <edge from-layer="1872" from-port="3" to-layer="1873" to-port="0" /> <edge from-layer="1873" from-port="2" to-layer="1874" to-port="0" /> <edge from-layer="1873" from-port="2" to-layer="1887" to-port="0" /> <edge from-layer="1874" from-port="1" to-layer="1886" to-port="0" /> <edge from-layer="1875" from-port="0" to-layer="1877" to-port="1" /> <edge from-layer="1876" from-port="0" to-layer="1877" to-port="2" /> <edge from-layer="1877" from-port="3" to-layer="1881" to-port="0" /> <edge from-layer="1878" from-port="0" to-layer="1881" to-port="1" /> <edge from-layer="1879" from-port="0" to-layer="1881" to-port="2" /> <edge from-layer="1880" from-port="0" to-layer="1881" to-port="3" /> <edge from-layer="1881" from-port="4" to-layer="1885" to-port="2" /> <edge from-layer="1881" from-port="4" to-layer="1883" to-port="0" /> <edge from-layer="1882" from-port="0" to-layer="1883" to-port="1" /> <edge from-layer="1883" from-port="2" to-layer="1885" to-port="0" /> <edge from-layer="1884" from-port="0" to-layer="1885" to-port="1" /> <edge from-layer="1885" from-port="3" to-layer="1886" to-port="1" /> <edge from-layer="1886" from-port="2" to-layer="1887" to-port="1" /> <edge from-layer="1887" from-port="2" to-layer="1888" to-port="3" /> <edge from-layer="1888" from-port="4" to-layer="1890" to-port="0" /> <edge from-layer="1889" from-port="0" to-layer="1890" to-port="1" /> <edge from-layer="1890" from-port="2" to-layer="1892" to-port="0" /> <edge from-layer="1891" from-port="0" to-layer="1892" to-port="1" /> <edge from-layer="1892" from-port="2" to-layer="1900" to-port="0" /> <edge from-layer="1893" from-port="0" to-layer="1894" to-port="0" /> <edge from-layer="1894" from-port="1" to-layer="1897" to-port="0" /> <edge from-layer="1895" from-port="0" to-layer="1896" to-port="0" /> <edge from-layer="1896" from-port="1" to-layer="1897" to-port="1" /> <edge from-layer="1897" from-port="2" to-layer="1899" to-port="0" /> <edge from-layer="1898" from-port="0" to-layer="1899" to-port="1" /> <edge from-layer="1899" from-port="2" to-layer="1900" to-port="1" /> <edge from-layer="1900" from-port="2" to-layer="1902" to-port="0" /> <edge from-layer="1901" from-port="0" to-layer="1902" to-port="1" /> <edge from-layer="1902" from-port="2" to-layer="1903" to-port="1" /> <edge from-layer="1903" from-port="2" to-layer="1904" to-port="0" /> <edge from-layer="1904" from-port="2" to-layer="1933" to-port="0" /> <edge from-layer="1904" from-port="2" to-layer="2135" to-port="1" /> <edge from-layer="1904" from-port="2" to-layer="1906" to-port="0" /> <edge from-layer="1905" from-port="0" to-layer="1906" to-port="1" /> <edge from-layer="1906" from-port="2" to-layer="1908" to-port="0" /> <edge from-layer="1907" from-port="0" to-layer="1908" to-port="1" /> <edge from-layer="1908" from-port="2" to-layer="1910" to-port="0" /> <edge from-layer="1909" from-port="0" to-layer="1910" to-port="1" /> <edge from-layer="1910" from-port="2" to-layer="1918" to-port="0" /> <edge from-layer="1911" from-port="0" to-layer="1912" to-port="0" /> <edge from-layer="1912" from-port="1" to-layer="1915" to-port="0" /> <edge from-layer="1913" from-port="0" to-layer="1914" to-port="0" /> <edge from-layer="1914" from-port="1" to-layer="1915" to-port="1" /> <edge from-layer="1915" from-port="2" to-layer="1917" to-port="0" /> <edge from-layer="1916" from-port="0" to-layer="1917" to-port="1" /> <edge from-layer="1917" from-port="2" to-layer="1918" to-port="1" /> <edge from-layer="1918" from-port="2" to-layer="1920" to-port="0" /> <edge from-layer="1919" from-port="0" to-layer="1920" to-port="1" /> <edge from-layer="1920" from-port="2" to-layer="1921" to-port="0" /> <edge from-layer="1921" from-port="1" to-layer="1929" to-port="0" /> <edge from-layer="1922" from-port="0" to-layer="1923" to-port="0" /> <edge from-layer="1923" from-port="1" to-layer="1926" to-port="0" /> <edge from-layer="1924" from-port="0" to-layer="1925" to-port="0" /> <edge from-layer="1925" from-port="1" to-layer="1926" to-port="1" /> <edge from-layer="1926" from-port="2" to-layer="1928" to-port="0" /> <edge from-layer="1927" from-port="0" to-layer="1928" to-port="1" /> <edge from-layer="1928" from-port="2" to-layer="1929" to-port="1" /> <edge from-layer="1929" from-port="2" to-layer="1931" to-port="0" /> <edge from-layer="1930" from-port="0" to-layer="1931" to-port="1" /> <edge from-layer="1931" from-port="2" to-layer="2134" to-port="0" /> <edge from-layer="1932" from-port="0" to-layer="1933" to-port="1" /> <edge from-layer="1933" from-port="2" to-layer="1935" to-port="0" /> <edge from-layer="1934" from-port="0" to-layer="1935" to-port="1" /> <edge from-layer="1935" from-port="2" to-layer="1937" to-port="0" /> <edge from-layer="1936" from-port="0" to-layer="1937" to-port="1" /> <edge from-layer="1937" from-port="2" to-layer="1945" to-port="0" /> <edge from-layer="1938" from-port="0" to-layer="1939" to-port="0" /> <edge from-layer="1939" from-port="1" to-layer="1942" to-port="0" /> <edge from-layer="1940" from-port="0" to-layer="1941" to-port="0" /> <edge from-layer="1941" from-port="1" to-layer="1942" to-port="1" /> <edge from-layer="1942" from-port="2" to-layer="1944" to-port="0" /> <edge from-layer="1943" from-port="0" to-layer="1944" to-port="1" /> <edge from-layer="1944" from-port="2" to-layer="1945" to-port="1" /> <edge from-layer="1945" from-port="2" to-layer="1947" to-port="0" /> <edge from-layer="1946" from-port="0" to-layer="1947" to-port="1" /> <edge from-layer="1947" from-port="2" to-layer="1949" to-port="0" /> <edge from-layer="1948" from-port="0" to-layer="1949" to-port="1" /> <edge from-layer="1949" from-port="2" to-layer="1953" to-port="0" /> <edge from-layer="1949" from-port="2" to-layer="2071" to-port="0" /> <edge from-layer="1949" from-port="2" to-layer="1965" to-port="0" /> <edge from-layer="1950" from-port="0" to-layer="1953" to-port="1" /> <edge from-layer="1951" from-port="0" to-layer="1953" to-port="2" /> <edge from-layer="1952" from-port="0" to-layer="1953" to-port="3" /> <edge from-layer="1953" from-port="4" to-layer="1955" to-port="0" /> <edge from-layer="1954" from-port="0" to-layer="1955" to-port="1" /> <edge from-layer="1955" from-port="2" to-layer="1959" to-port="0" /> <edge from-layer="1955" from-port="2" to-layer="2024" to-port="0" /> <edge from-layer="1956" from-port="0" to-layer="1959" to-port="1" /> <edge from-layer="1957" from-port="0" to-layer="1959" to-port="2" /> <edge from-layer="1958" from-port="0" to-layer="1959" to-port="3" /> <edge from-layer="1959" from-port="4" to-layer="1994" to-port="0" /> <edge from-layer="1959" from-port="4" to-layer="2005" to-port="0" /> <edge from-layer="1959" from-port="4" to-layer="2010" to-port="0" /> <edge from-layer="1959" from-port="4" to-layer="1991" to-port="0" /> <edge from-layer="1960" from-port="0" to-layer="1987" to-port="0" /> <edge from-layer="1961" from-port="0" to-layer="1987" to-port="1" /> <edge from-layer="1962" from-port="0" to-layer="1965" to-port="1" /> <edge from-layer="1963" from-port="0" to-layer="1965" to-port="2" /> <edge from-layer="1964" from-port="0" to-layer="1965" to-port="3" /> <edge from-layer="1965" from-port="4" to-layer="1967" to-port="0" /> <edge from-layer="1966" from-port="0" to-layer="1967" to-port="1" /> <edge from-layer="1967" from-port="2" to-layer="2029" to-port="0" /> <edge from-layer="1967" from-port="2" to-layer="1968" to-port="0" /> <edge from-layer="1967" from-port="2" to-layer="2056" to-port="0" /> <edge from-layer="1968" from-port="1" to-layer="1970" to-port="0" /> <edge from-layer="1969" from-port="0" to-layer="1970" to-port="2" /> <edge from-layer="1970" from-port="3" to-layer="1983" to-port="0" /> <edge from-layer="1971" from-port="0" to-layer="1976" to-port="0" /> <edge from-layer="1972" from-port="0" to-layer="1975" to-port="1" /> <edge from-layer="1973" from-port="0" to-layer="1975" to-port="2" /> <edge from-layer="1974" from-port="0" to-layer="1975" to-port="3" /> <edge from-layer="1975" from-port="4" to-layer="1976" to-port="1" /> <edge from-layer="1976" from-port="2" to-layer="1977" to-port="0" /> <edge from-layer="1977" from-port="1" to-layer="1979" to-port="0" /> <edge from-layer="1978" from-port="0" to-layer="1979" to-port="2" /> <edge from-layer="1979" from-port="3" to-layer="1980" to-port="0" /> <edge from-layer="1979" from-port="3" to-layer="2058" to-port="0" /> <edge from-layer="1980" from-port="1" to-layer="1982" to-port="0" /> <edge from-layer="1981" from-port="0" to-layer="1982" to-port="2" /> <edge from-layer="1982" from-port="3" to-layer="1983" to-port="1" /> <edge from-layer="1983" from-port="2" to-layer="1985" to-port="0" /> <edge from-layer="1984" from-port="0" to-layer="1985" to-port="1" /> <edge from-layer="1985" from-port="2" to-layer="1987" to-port="2" /> <edge from-layer="1985" from-port="2" to-layer="2015" to-port="2" /> <edge from-layer="1986" from-port="0" to-layer="1987" to-port="3" /> <edge from-layer="1987" from-port="4" to-layer="1989" to-port="0" /> <edge from-layer="1988" from-port="0" to-layer="1989" to-port="2" /> <edge from-layer="1989" from-port="3" to-layer="1990" to-port="0" /> <edge from-layer="1990" from-port="2" to-layer="1991" to-port="1" /> <edge from-layer="1990" from-port="2" to-layer="2030" to-port="1" /> <edge from-layer="1991" from-port="2" to-layer="2020" to-port="0" /> <edge from-layer="1992" from-port="0" to-layer="2002" to-port="0" /> <edge from-layer="1993" from-port="0" to-layer="2002" to-port="1" /> <edge from-layer="1994" from-port="1" to-layer="1996" to-port="0" /> <edge from-layer="1995" from-port="0" to-layer="1996" to-port="2" /> <edge from-layer="1996" from-port="3" to-layer="1997" to-port="0" /> <edge from-layer="1997" from-port="2" to-layer="1998" to-port="0" /> <edge from-layer="1998" from-port="1" to-layer="2000" to-port="0" /> <edge from-layer="1999" from-port="0" to-layer="2000" to-port="1" /> <edge from-layer="2000" from-port="2" to-layer="2002" to-port="2" /> <edge from-layer="2001" from-port="0" to-layer="2002" to-port="3" /> <edge from-layer="2002" from-port="4" to-layer="2005" to-port="1" /> <edge from-layer="2002" from-port="4" to-layer="2010" to-port="2" /> <edge from-layer="2003" from-port="0" to-layer="2005" to-port="2" /> <edge from-layer="2004" from-port="0" to-layer="2005" to-port="3" /> <edge from-layer="2005" from-port="4" to-layer="2007" to-port="0" /> <edge from-layer="2006" from-port="0" to-layer="2007" to-port="1" /> <edge from-layer="2007" from-port="2" to-layer="2011" to-port="0" /> <edge from-layer="2008" from-port="0" to-layer="2010" to-port="1" /> <edge from-layer="2009" from-port="0" to-layer="2010" to-port="3" /> <edge from-layer="2010" from-port="4" to-layer="2011" to-port="1" /> <edge from-layer="2011" from-port="2" to-layer="2019" to-port="0" /> <edge from-layer="2012" from-port="0" to-layer="2015" to-port="0" /> <edge from-layer="2013" from-port="0" to-layer="2015" to-port="1" /> <edge from-layer="2014" from-port="0" to-layer="2015" to-port="3" /> <edge from-layer="2015" from-port="4" to-layer="2017" to-port="0" /> <edge from-layer="2016" from-port="0" to-layer="2017" to-port="2" /> <edge from-layer="2017" from-port="3" to-layer="2018" to-port="0" /> <edge from-layer="2018" from-port="2" to-layer="2019" to-port="1" /> <edge from-layer="2018" from-port="2" to-layer="2051" to-port="1" /> <edge from-layer="2019" from-port="2" to-layer="2020" to-port="1" /> <edge from-layer="2020" from-port="2" to-layer="2025" to-port="0" /> <edge from-layer="2021" from-port="0" to-layer="2024" to-port="1" /> <edge from-layer="2022" from-port="0" to-layer="2024" to-port="2" /> <edge from-layer="2023" from-port="0" to-layer="2024" to-port="3" /> <edge from-layer="2024" from-port="4" to-layer="2025" to-port="1" /> <edge from-layer="2025" from-port="2" to-layer="2081" to-port="0" /> <edge from-layer="2025" from-port="2" to-layer="2119" to-port="0" /> <edge from-layer="2026" from-port="0" to-layer="2029" to-port="1" /> <edge from-layer="2027" from-port="0" to-layer="2029" to-port="2" /> <edge from-layer="2028" from-port="0" to-layer="2029" to-port="3" /> <edge from-layer="2029" from-port="4" to-layer="2033" to-port="0" /> <edge from-layer="2029" from-port="4" to-layer="2049" to-port="0" /> <edge from-layer="2029" from-port="4" to-layer="2044" to-port="0" /> <edge from-layer="2029" from-port="4" to-layer="2030" to-port="0" /> <edge from-layer="2030" from-port="2" to-layer="2052" to-port="0" /> <edge from-layer="2031" from-port="0" to-layer="2041" to-port="0" /> <edge from-layer="2032" from-port="0" to-layer="2041" to-port="1" /> <edge from-layer="2033" from-port="1" to-layer="2035" to-port="0" /> <edge from-layer="2034" from-port="0" to-layer="2035" to-port="2" /> <edge from-layer="2035" from-port="3" to-layer="2036" to-port="0" /> <edge from-layer="2036" from-port="2" to-layer="2037" to-port="0" /> <edge from-layer="2037" from-port="1" to-layer="2039" to-port="0" /> <edge from-layer="2038" from-port="0" to-layer="2039" to-port="1" /> <edge from-layer="2039" from-port="2" to-layer="2041" to-port="2" /> <edge from-layer="2040" from-port="0" to-layer="2041" to-port="3" /> <edge from-layer="2041" from-port="4" to-layer="2049" to-port="2" /> <edge from-layer="2041" from-port="4" to-layer="2044" to-port="1" /> <edge from-layer="2042" from-port="0" to-layer="2044" to-port="2" /> <edge from-layer="2043" from-port="0" to-layer="2044" to-port="3" /> <edge from-layer="2044" from-port="4" to-layer="2046" to-port="0" /> <edge from-layer="2045" from-port="0" to-layer="2046" to-port="1" /> <edge from-layer="2046" from-port="2" to-layer="2050" to-port="0" /> <edge from-layer="2047" from-port="0" to-layer="2049" to-port="1" /> <edge from-layer="2048" from-port="0" to-layer="2049" to-port="3" /> <edge from-layer="2049" from-port="4" to-layer="2050" to-port="1" /> <edge from-layer="2050" from-port="2" to-layer="2051" to-port="0" /> <edge from-layer="2051" from-port="2" to-layer="2052" to-port="1" /> <edge from-layer="2052" from-port="2" to-layer="2057" to-port="0" /> <edge from-layer="2053" from-port="0" to-layer="2056" to-port="1" /> <edge from-layer="2054" from-port="0" to-layer="2056" to-port="2" /> <edge from-layer="2055" from-port="0" to-layer="2056" to-port="3" /> <edge from-layer="2056" from-port="4" to-layer="2057" to-port="1" /> <edge from-layer="2057" from-port="2" to-layer="2058" to-port="1" /> <edge from-layer="2058" from-port="2" to-layer="2119" to-port="1" /> <edge from-layer="2058" from-port="2" to-layer="2078" to-port="0" /> <edge from-layer="2058" from-port="2" to-layer="2235" to-port="0" /> <edge from-layer="2059" from-port="0" to-layer="2064" to-port="0" /> <edge from-layer="2060" from-port="0" to-layer="2063" to-port="1" /> <edge from-layer="2061" from-port="0" to-layer="2063" to-port="2" /> <edge from-layer="2062" from-port="0" to-layer="2063" to-port="3" /> <edge from-layer="2063" from-port="4" to-layer="2064" to-port="1" /> <edge from-layer="2064" from-port="2" to-layer="2065" to-port="0" /> <edge from-layer="2065" from-port="1" to-layer="2067" to-port="0" /> <edge from-layer="2066" from-port="0" to-layer="2067" to-port="2" /> <edge from-layer="2067" from-port="3" to-layer="2074" to-port="0" /> <edge from-layer="2068" from-port="0" to-layer="2071" to-port="1" /> <edge from-layer="2069" from-port="0" to-layer="2071" to-port="2" /> <edge from-layer="2070" from-port="0" to-layer="2071" to-port="3" /> <edge from-layer="2071" from-port="4" to-layer="2073" to-port="0" /> <edge from-layer="2072" from-port="0" to-layer="2073" to-port="1" /> <edge from-layer="2073" from-port="2" to-layer="2074" to-port="1" /> <edge from-layer="2074" from-port="2" to-layer="2119" to-port="2" /> <edge from-layer="2074" from-port="2" to-layer="2234" to-port="0" /> <edge from-layer="2075" from-port="0" to-layer="2093" to-port="0" /> <edge from-layer="2076" from-port="0" to-layer="2088" to-port="0" /> <edge from-layer="2077" from-port="0" to-layer="2088" to-port="1" /> <edge from-layer="2077" from-port="0" to-layer="2091" to-port="1" /> <edge from-layer="2078" from-port="1" to-layer="2080" to-port="0" /> <edge from-layer="2079" from-port="0" to-layer="2080" to-port="2" /> <edge from-layer="2080" from-port="3" to-layer="2084" to-port="0" /> <edge from-layer="2080" from-port="3" to-layer="2090" to-port="0" /> <edge from-layer="2081" from-port="1" to-layer="2108" to-port="0" /> <edge from-layer="2081" from-port="1" to-layer="2083" to-port="0" /> <edge from-layer="2082" from-port="0" to-layer="2083" to-port="2" /> <edge from-layer="2083" from-port="3" to-layer="2084" to-port="1" /> <edge from-layer="2084" from-port="2" to-layer="2086" to-port="0" /> <edge from-layer="2085" from-port="0" to-layer="2090" to-port="1" /> <edge from-layer="2085" from-port="0" to-layer="2086" to-port="1" /> <edge from-layer="2086" from-port="2" to-layer="2088" to-port="2" /> <edge from-layer="2087" from-port="0" to-layer="2088" to-port="3" /> <edge from-layer="2087" from-port="0" to-layer="2091" to-port="3" /> <edge from-layer="2088" from-port="4" to-layer="2093" to-port="1" /> <edge from-layer="2089" from-port="0" to-layer="2091" to-port="0" /> <edge from-layer="2090" from-port="2" to-layer="2091" to-port="2" /> <edge from-layer="2090" from-port="2" to-layer="2098" to-port="2" /> <edge from-layer="2091" from-port="4" to-layer="2093" to-port="2" /> <edge from-layer="2092" from-port="0" to-layer="2093" to-port="3" /> <edge from-layer="2093" from-port="4" to-layer="2100" to-port="0" /> <edge from-layer="2094" from-port="0" to-layer="2100" to-port="1" /> <edge from-layer="2095" from-port="0" to-layer="2098" to-port="0" /> <edge from-layer="2096" from-port="0" to-layer="2098" to-port="1" /> <edge from-layer="2097" from-port="0" to-layer="2098" to-port="3" /> <edge from-layer="2098" from-port="4" to-layer="2100" to-port="2" /> <edge from-layer="2099" from-port="0" to-layer="2100" to-port="3" /> <edge from-layer="2100" from-port="4" to-layer="2103" to-port="0" /> <edge from-layer="2101" from-port="0" to-layer="2103" to-port="1" /> <edge from-layer="2102" from-port="0" to-layer="2103" to-port="2" /> <edge from-layer="2103" from-port="3" to-layer="2104" to-port="0" /> <edge from-layer="2104" from-port="2" to-layer="2118" to-port="0" /> <edge from-layer="2104" from-port="2" to-layer="2105" to-port="0" /> <edge from-layer="2105" from-port="1" to-layer="2117" to-port="0" /> <edge from-layer="2106" from-port="0" to-layer="2108" to-port="1" /> <edge from-layer="2107" from-port="0" to-layer="2108" to-port="2" /> <edge from-layer="2108" from-port="3" to-layer="2112" to-port="0" /> <edge from-layer="2109" from-port="0" to-layer="2112" to-port="1" /> <edge from-layer="2110" from-port="0" to-layer="2112" to-port="2" /> <edge from-layer="2111" from-port="0" to-layer="2112" to-port="3" /> <edge from-layer="2112" from-port="4" to-layer="2114" to-port="0" /> <edge from-layer="2112" from-port="4" to-layer="2116" to-port="2" /> <edge from-layer="2113" from-port="0" to-layer="2114" to-port="1" /> <edge from-layer="2114" from-port="2" to-layer="2116" to-port="0" /> <edge from-layer="2115" from-port="0" to-layer="2116" to-port="1" /> <edge from-layer="2116" from-port="3" to-layer="2117" to-port="1" /> <edge from-layer="2117" from-port="2" to-layer="2118" to-port="1" /> <edge from-layer="2118" from-port="2" to-layer="2119" to-port="3" /> <edge from-layer="2119" from-port="4" to-layer="2121" to-port="0" /> <edge from-layer="2120" from-port="0" to-layer="2121" to-port="1" /> <edge from-layer="2121" from-port="2" to-layer="2123" to-port="0" /> <edge from-layer="2122" from-port="0" to-layer="2123" to-port="1" /> <edge from-layer="2123" from-port="2" to-layer="2131" to-port="0" /> <edge from-layer="2124" from-port="0" to-layer="2125" to-port="0" /> <edge from-layer="2125" from-port="1" to-layer="2128" to-port="0" /> <edge from-layer="2126" from-port="0" to-layer="2127" to-port="0" /> <edge from-layer="2127" from-port="1" to-layer="2128" to-port="1" /> <edge from-layer="2128" from-port="2" to-layer="2130" to-port="0" /> <edge from-layer="2129" from-port="0" to-layer="2130" to-port="1" /> <edge from-layer="2130" from-port="2" to-layer="2131" to-port="1" /> <edge from-layer="2131" from-port="2" to-layer="2133" to-port="0" /> <edge from-layer="2132" from-port="0" to-layer="2133" to-port="1" /> <edge from-layer="2133" from-port="2" to-layer="2134" to-port="1" /> <edge from-layer="2134" from-port="2" to-layer="2135" to-port="0" /> <edge from-layer="2135" from-port="2" to-layer="2137" to-port="0" /> <edge from-layer="2135" from-port="2" to-layer="2376" to-port="1" /> <edge from-layer="2135" from-port="2" to-layer="2252" to-port="0" /> <edge from-layer="2136" from-port="0" to-layer="2137" to-port="1" /> <edge from-layer="2137" from-port="2" to-layer="2139" to-port="0" /> <edge from-layer="2138" from-port="0" to-layer="2139" to-port="1" /> <edge from-layer="2139" from-port="2" to-layer="2141" to-port="0" /> <edge from-layer="2140" from-port="0" to-layer="2141" to-port="1" /> <edge from-layer="2141" from-port="2" to-layer="2149" to-port="0" /> <edge from-layer="2142" from-port="0" to-layer="2143" to-port="0" /> <edge from-layer="2143" from-port="1" to-layer="2146" to-port="0" /> <edge from-layer="2144" from-port="0" to-layer="2145" to-port="0" /> <edge from-layer="2145" from-port="1" to-layer="2146" to-port="1" /> <edge from-layer="2146" from-port="2" to-layer="2148" to-port="0" /> <edge from-layer="2147" from-port="0" to-layer="2148" to-port="1" /> <edge from-layer="2148" from-port="2" to-layer="2149" to-port="1" /> <edge from-layer="2149" from-port="2" to-layer="2151" to-port="0" /> <edge from-layer="2150" from-port="0" to-layer="2151" to-port="1" /> <edge from-layer="2151" from-port="2" to-layer="2153" to-port="0" /> <edge from-layer="2152" from-port="0" to-layer="2153" to-port="1" /> <edge from-layer="2153" from-port="2" to-layer="2281" to-port="0" /> <edge from-layer="2153" from-port="2" to-layer="2157" to-port="0" /> <edge from-layer="2153" from-port="2" to-layer="2174" to-port="0" /> <edge from-layer="2154" from-port="0" to-layer="2157" to-port="1" /> <edge from-layer="2155" from-port="0" to-layer="2157" to-port="2" /> <edge from-layer="2156" from-port="0" to-layer="2157" to-port="3" /> <edge from-layer="2157" from-port="4" to-layer="2159" to-port="0" /> <edge from-layer="2158" from-port="0" to-layer="2159" to-port="1" /> <edge from-layer="2159" from-port="2" to-layer="2160" to-port="1" /> <edge from-layer="2160" from-port="2" to-layer="2161" to-port="0" /> <edge from-layer="2160" from-port="2" to-layer="2360" to-port="2" /> <edge from-layer="2162" from-port="0" to-layer="2167" to-port="0" /> <edge from-layer="2163" from-port="0" to-layer="2166" to-port="1" /> <edge from-layer="2164" from-port="0" to-layer="2166" to-port="2" /> <edge from-layer="2165" from-port="0" to-layer="2166" to-port="3" /> <edge from-layer="2166" from-port="4" to-layer="2167" to-port="1" /> <edge from-layer="2167" from-port="2" to-layer="2168" to-port="0" /> <edge from-layer="2168" from-port="1" to-layer="2170" to-port="0" /> <edge from-layer="2169" from-port="0" to-layer="2170" to-port="2" /> <edge from-layer="2170" from-port="3" to-layer="2232" to-port="0" /> <edge from-layer="2170" from-port="3" to-layer="2186" to-port="0" /> <edge from-layer="2171" from-port="0" to-layer="2174" to-port="1" /> <edge from-layer="2172" from-port="0" to-layer="2174" to-port="2" /> <edge from-layer="2173" from-port="0" to-layer="2174" to-port="3" /> <edge from-layer="2174" from-port="4" to-layer="2176" to-port="0" /> <edge from-layer="2175" from-port="0" to-layer="2176" to-port="1" /> <edge from-layer="2176" from-port="2" to-layer="2183" to-port="0" /> <edge from-layer="2176" from-port="2" to-layer="2180" to-port="0" /> <edge from-layer="2176" from-port="2" to-layer="2230" to-port="0" /> <edge from-layer="2177" from-port="0" to-layer="2180" to-port="1" /> <edge from-layer="2178" from-port="0" to-layer="2180" to-port="2" /> <edge from-layer="2179" from-port="0" to-layer="2180" to-port="3" /> <edge from-layer="2180" from-port="4" to-layer="2197" to-port="0" /> <edge from-layer="2180" from-port="4" to-layer="2200" to-port="0" /> <edge from-layer="2180" from-port="4" to-layer="2211" to-port="0" /> <edge from-layer="2180" from-port="4" to-layer="2216" to-port="0" /> <edge from-layer="2181" from-port="0" to-layer="2193" to-port="0" /> <edge from-layer="2182" from-port="0" to-layer="2193" to-port="1" /> <edge from-layer="2183" from-port="1" to-layer="2185" to-port="0" /> <edge from-layer="2184" from-port="0" to-layer="2185" to-port="2" /> <edge from-layer="2185" from-port="3" to-layer="2189" to-port="0" /> <edge from-layer="2186" from-port="1" to-layer="2188" to-port="0" /> <edge from-layer="2187" from-port="0" to-layer="2188" to-port="2" /> <edge from-layer="2188" from-port="3" to-layer="2189" to-port="1" /> <edge from-layer="2189" from-port="2" to-layer="2191" to-port="0" /> <edge from-layer="2190" from-port="0" to-layer="2191" to-port="1" /> <edge from-layer="2191" from-port="2" to-layer="2221" to-port="2" /> <edge from-layer="2191" from-port="2" to-layer="2193" to-port="2" /> <edge from-layer="2192" from-port="0" to-layer="2193" to-port="3" /> <edge from-layer="2193" from-port="4" to-layer="2195" to-port="0" /> <edge from-layer="2194" from-port="0" to-layer="2195" to-port="2" /> <edge from-layer="2195" from-port="3" to-layer="2196" to-port="0" /> <edge from-layer="2196" from-port="2" to-layer="2197" to-port="1" /> <edge from-layer="2196" from-port="2" to-layer="2288" to-port="1" /> <edge from-layer="2197" from-port="2" to-layer="2226" to-port="0" /> <edge from-layer="2198" from-port="0" to-layer="2208" to-port="0" /> <edge from-layer="2199" from-port="0" to-layer="2208" to-port="1" /> <edge from-layer="2200" from-port="1" to-layer="2202" to-port="0" /> <edge from-layer="2201" from-port="0" to-layer="2202" to-port="2" /> <edge from-layer="2202" from-port="3" to-layer="2203" to-port="0" /> <edge from-layer="2203" from-port="2" to-layer="2204" to-port="0" /> <edge from-layer="2204" from-port="1" to-layer="2206" to-port="0" /> <edge from-layer="2205" from-port="0" to-layer="2206" to-port="1" /> <edge from-layer="2206" from-port="2" to-layer="2208" to-port="2" /> <edge from-layer="2207" from-port="0" to-layer="2208" to-port="3" /> <edge from-layer="2208" from-port="4" to-layer="2211" to-port="1" /> <edge from-layer="2208" from-port="4" to-layer="2216" to-port="2" /> <edge from-layer="2209" from-port="0" to-layer="2211" to-port="2" /> <edge from-layer="2210" from-port="0" to-layer="2211" to-port="3" /> <edge from-layer="2211" from-port="4" to-layer="2213" to-port="0" /> <edge from-layer="2212" from-port="0" to-layer="2213" to-port="1" /> <edge from-layer="2213" from-port="2" to-layer="2217" to-port="0" /> <edge from-layer="2214" from-port="0" to-layer="2216" to-port="1" /> <edge from-layer="2215" from-port="0" to-layer="2216" to-port="3" /> <edge from-layer="2216" from-port="4" to-layer="2217" to-port="1" /> <edge from-layer="2217" from-port="2" to-layer="2225" to-port="0" /> <edge from-layer="2218" from-port="0" to-layer="2221" to-port="0" /> <edge from-layer="2219" from-port="0" to-layer="2221" to-port="1" /> <edge from-layer="2220" from-port="0" to-layer="2221" to-port="3" /> <edge from-layer="2221" from-port="4" to-layer="2223" to-port="0" /> <edge from-layer="2222" from-port="0" to-layer="2223" to-port="2" /> <edge from-layer="2223" from-port="3" to-layer="2224" to-port="0" /> <edge from-layer="2224" from-port="2" to-layer="2225" to-port="1" /> <edge from-layer="2224" from-port="2" to-layer="2309" to-port="1" /> <edge from-layer="2225" from-port="2" to-layer="2226" to-port="1" /> <edge from-layer="2226" from-port="2" to-layer="2231" to-port="0" /> <edge from-layer="2227" from-port="0" to-layer="2230" to-port="1" /> <edge from-layer="2228" from-port="0" to-layer="2230" to-port="2" /> <edge from-layer="2229" from-port="0" to-layer="2230" to-port="3" /> <edge from-layer="2230" from-port="4" to-layer="2231" to-port="1" /> <edge from-layer="2231" from-port="2" to-layer="2232" to-port="1" /> <edge from-layer="2232" from-port="2" to-layer="2360" to-port="1" /> <edge from-layer="2232" from-port="2" to-layer="2233" to-port="0" /> <edge from-layer="2232" from-port="2" to-layer="2319" to-port="0" /> <edge from-layer="2242" from-port="0" to-layer="2247" to-port="0" /> <edge from-layer="2243" from-port="0" to-layer="2246" to-port="1" /> <edge from-layer="2244" from-port="0" to-layer="2246" to-port="2" /> <edge from-layer="2245" from-port="0" to-layer="2246" to-port="3" /> <edge from-layer="2246" from-port="4" to-layer="2247" to-port="1" /> <edge from-layer="2247" from-port="2" to-layer="2248" to-port="0" /> <edge from-layer="2248" from-port="1" to-layer="2250" to-port="0" /> <edge from-layer="2249" from-port="0" to-layer="2250" to-port="2" /> <edge from-layer="2250" from-port="3" to-layer="10024" to-port="0" /> <edge from-layer="2251" from-port="0" to-layer="2252" to-port="1" /> <edge from-layer="2252" from-port="2" to-layer="2254" to-port="0" /> <edge from-layer="2253" from-port="0" to-layer="2254" to-port="1" /> <edge from-layer="2254" from-port="2" to-layer="2256" to-port="0" /> <edge from-layer="2255" from-port="0" to-layer="2256" to-port="1" /> <edge from-layer="2256" from-port="2" to-layer="2264" to-port="0" /> <edge from-layer="2257" from-port="0" to-layer="2258" to-port="0" /> <edge from-layer="2258" from-port="1" to-layer="2261" to-port="0" /> <edge from-layer="2259" from-port="0" to-layer="2260" to-port="0" /> <edge from-layer="2260" from-port="1" to-layer="2261" to-port="1" /> <edge from-layer="2261" from-port="2" to-layer="2263" to-port="0" /> <edge from-layer="2262" from-port="0" to-layer="2263" to-port="1" /> <edge from-layer="2263" from-port="2" to-layer="2264" to-port="1" /> <edge from-layer="2264" from-port="2" to-layer="2266" to-port="0" /> <edge from-layer="2265" from-port="0" to-layer="2266" to-port="1" /> <edge from-layer="2266" from-port="2" to-layer="2267" to-port="0" /> <edge from-layer="2267" from-port="1" to-layer="2275" to-port="0" /> <edge from-layer="2268" from-port="0" to-layer="2269" to-port="0" /> <edge from-layer="2269" from-port="1" to-layer="2272" to-port="0" /> <edge from-layer="2270" from-port="0" to-layer="2271" to-port="0" /> <edge from-layer="2271" from-port="1" to-layer="2272" to-port="1" /> <edge from-layer="2272" from-port="2" to-layer="2274" to-port="0" /> <edge from-layer="2273" from-port="0" to-layer="2274" to-port="1" /> <edge from-layer="2274" from-port="2" to-layer="2275" to-port="1" /> <edge from-layer="2275" from-port="2" to-layer="2277" to-port="0" /> <edge from-layer="2276" from-port="0" to-layer="2277" to-port="1" /> <edge from-layer="2277" from-port="2" to-layer="2375" to-port="0" /> <edge from-layer="2278" from-port="0" to-layer="2281" to-port="1" /> <edge from-layer="2279" from-port="0" to-layer="2281" to-port="2" /> <edge from-layer="2280" from-port="0" to-layer="2281" to-port="3" /> <edge from-layer="2281" from-port="4" to-layer="2283" to-port="0" /> <edge from-layer="2282" from-port="0" to-layer="2283" to-port="1" /> <edge from-layer="2283" from-port="2" to-layer="2314" to-port="0" /> <edge from-layer="2283" from-port="2" to-layer="2287" to-port="0" /> <edge from-layer="2284" from-port="0" to-layer="2287" to-port="1" /> <edge from-layer="2285" from-port="0" to-layer="2287" to-port="2" /> <edge from-layer="2286" from-port="0" to-layer="2287" to-port="3" /> <edge from-layer="2287" from-port="4" to-layer="2288" to-port="0" /> <edge from-layer="2287" from-port="4" to-layer="2291" to-port="0" /> <edge from-layer="2287" from-port="4" to-layer="2307" to-port="0" /> <edge from-layer="2287" from-port="4" to-layer="2302" to-port="0" /> <edge from-layer="2288" from-port="2" to-layer="2310" to-port="0" /> <edge from-layer="2289" from-port="0" to-layer="2299" to-port="0" /> <edge from-layer="2290" from-port="0" to-layer="2299" to-port="1" /> <edge from-layer="2291" from-port="1" to-layer="2293" to-port="0" /> <edge from-layer="2292" from-port="0" to-layer="2293" to-port="2" /> <edge from-layer="2293" from-port="3" to-layer="2294" to-port="0" /> <edge from-layer="2294" from-port="2" to-layer="2295" to-port="0" /> <edge from-layer="2295" from-port="1" to-layer="2297" to-port="0" /> <edge from-layer="2296" from-port="0" to-layer="2297" to-port="1" /> <edge from-layer="2297" from-port="2" to-layer="2299" to-port="2" /> <edge from-layer="2298" from-port="0" to-layer="2299" to-port="3" /> <edge from-layer="2299" from-port="4" to-layer="2302" to-port="1" /> <edge from-layer="2299" from-port="4" to-layer="2307" to-port="2" /> <edge from-layer="2300" from-port="0" to-layer="2302" to-port="2" /> <edge from-layer="2301" from-port="0" to-layer="2302" to-port="3" /> <edge from-layer="2302" from-port="4" to-layer="2304" to-port="0" /> <edge from-layer="2303" from-port="0" to-layer="2304" to-port="1" /> <edge from-layer="2304" from-port="2" to-layer="2308" to-port="0" /> <edge from-layer="2305" from-port="0" to-layer="2307" to-port="1" /> <edge from-layer="2306" from-port="0" to-layer="2307" to-port="3" /> <edge from-layer="2307" from-port="4" to-layer="2308" to-port="1" /> <edge from-layer="2308" from-port="2" to-layer="2309" to-port="0" /> <edge from-layer="2309" from-port="2" to-layer="2310" to-port="1" /> <edge from-layer="2310" from-port="2" to-layer="2315" to-port="0" /> <edge from-layer="2311" from-port="0" to-layer="2314" to-port="1" /> <edge from-layer="2312" from-port="0" to-layer="2314" to-port="2" /> <edge from-layer="2313" from-port="0" to-layer="2314" to-port="3" /> <edge from-layer="2314" from-port="4" to-layer="2315" to-port="1" /> <edge from-layer="2315" from-port="2" to-layer="2360" to-port="0" /> <edge from-layer="2315" from-port="2" to-layer="2322" to-port="0" /> <edge from-layer="2316" from-port="0" to-layer="2334" to-port="0" /> <edge from-layer="2317" from-port="0" to-layer="2329" to-port="0" /> <edge from-layer="2318" from-port="0" to-layer="2329" to-port="1" /> <edge from-layer="2318" from-port="0" to-layer="2332" to-port="1" /> <edge from-layer="2319" from-port="1" to-layer="2321" to-port="0" /> <edge from-layer="2320" from-port="0" to-layer="2321" to-port="2" /> <edge from-layer="2321" from-port="3" to-layer="2325" to-port="0" /> <edge from-layer="2321" from-port="3" to-layer="2331" to-port="0" /> <edge from-layer="2322" from-port="1" to-layer="2349" to-port="0" /> <edge from-layer="2322" from-port="1" to-layer="2324" to-port="0" /> <edge from-layer="2323" from-port="0" to-layer="2324" to-port="2" /> <edge from-layer="2324" from-port="3" to-layer="2325" to-port="1" /> <edge from-layer="2325" from-port="2" to-layer="2327" to-port="0" /> <edge from-layer="2326" from-port="0" to-layer="2331" to-port="1" /> <edge from-layer="2326" from-port="0" to-layer="2327" to-port="1" /> <edge from-layer="2327" from-port="2" to-layer="2329" to-port="2" /> <edge from-layer="2328" from-port="0" to-layer="2329" to-port="3" /> <edge from-layer="2328" from-port="0" to-layer="2332" to-port="3" /> <edge from-layer="2329" from-port="4" to-layer="2334" to-port="1" /> <edge from-layer="2330" from-port="0" to-layer="2332" to-port="0" /> <edge from-layer="2331" from-port="2" to-layer="2332" to-port="2" /> <edge from-layer="2331" from-port="2" to-layer="2339" to-port="2" /> <edge from-layer="2332" from-port="4" to-layer="2334" to-port="2" /> <edge from-layer="2333" from-port="0" to-layer="2334" to-port="3" /> <edge from-layer="2334" from-port="4" to-layer="2341" to-port="0" /> <edge from-layer="2335" from-port="0" to-layer="2341" to-port="1" /> <edge from-layer="2336" from-port="0" to-layer="2339" to-port="0" /> <edge from-layer="2337" from-port="0" to-layer="2339" to-port="1" /> <edge from-layer="2338" from-port="0" to-layer="2339" to-port="3" /> <edge from-layer="2339" from-port="4" to-layer="2341" to-port="2" /> <edge from-layer="2340" from-port="0" to-layer="2341" to-port="3" /> <edge from-layer="2341" from-port="4" to-layer="2344" to-port="0" /> <edge from-layer="2342" from-port="0" to-layer="2344" to-port="1" /> <edge from-layer="2343" from-port="0" to-layer="2344" to-port="2" /> <edge from-layer="2344" from-port="3" to-layer="2345" to-port="0" /> <edge from-layer="2345" from-port="2" to-layer="2359" to-port="0" /> <edge from-layer="2345" from-port="2" to-layer="2346" to-port="0" /> <edge from-layer="2346" from-port="1" to-layer="2358" to-port="0" /> <edge from-layer="2347" from-port="0" to-layer="2349" to-port="1" /> <edge from-layer="2348" from-port="0" to-layer="2349" to-port="2" /> <edge from-layer="2349" from-port="3" to-layer="2353" to-port="0" /> <edge from-layer="2350" from-port="0" to-layer="2353" to-port="1" /> <edge from-layer="2351" from-port="0" to-layer="2353" to-port="2" /> <edge from-layer="2352" from-port="0" to-layer="2353" to-port="3" /> <edge from-layer="2353" from-port="4" to-layer="2355" to-port="0" /> <edge from-layer="2353" from-port="4" to-layer="2357" to-port="2" /> <edge from-layer="2354" from-port="0" to-layer="2355" to-port="1" /> <edge from-layer="2355" from-port="2" to-layer="2357" to-port="0" /> <edge from-layer="2356" from-port="0" to-layer="2357" to-port="1" /> <edge from-layer="2357" from-port="3" to-layer="2358" to-port="1" /> <edge from-layer="2358" from-port="2" to-layer="2359" to-port="1" /> <edge from-layer="2359" from-port="2" to-layer="2360" to-port="3" /> <edge from-layer="2360" from-port="4" to-layer="2362" to-port="0" /> <edge from-layer="2361" from-port="0" to-layer="2362" to-port="1" /> <edge from-layer="2362" from-port="2" to-layer="2364" to-port="0" /> <edge from-layer="2363" from-port="0" to-layer="2364" to-port="1" /> <edge from-layer="2364" from-port="2" to-layer="2372" to-port="0" /> <edge from-layer="2365" from-port="0" to-layer="2366" to-port="0" /> <edge from-layer="2366" from-port="1" to-layer="2369" to-port="0" /> <edge from-layer="2367" from-port="0" to-layer="2368" to-port="0" /> <edge from-layer="2368" from-port="1" to-layer="2369" to-port="1" /> <edge from-layer="2369" from-port="2" to-layer="2371" to-port="0" /> <edge from-layer="2370" from-port="0" to-layer="2371" to-port="1" /> <edge from-layer="2371" from-port="2" to-layer="2372" to-port="1" /> <edge from-layer="2372" from-port="2" to-layer="2374" to-port="0" /> <edge from-layer="2373" from-port="0" to-layer="2374" to-port="1" /> <edge from-layer="2374" from-port="2" to-layer="2375" to-port="1" /> <edge from-layer="2375" from-port="2" to-layer="2376" to-port="0" /> <edge from-layer="2376" from-port="2" to-layer="2378" to-port="0" /> <edge from-layer="2376" from-port="2" to-layer="2607" to-port="1" /> <edge from-layer="2376" from-port="2" to-layer="2405" to-port="0" /> <edge from-layer="2377" from-port="0" to-layer="2378" to-port="1" /> <edge from-layer="2378" from-port="2" to-layer="2380" to-port="0" /> <edge from-layer="2379" from-port="0" to-layer="2380" to-port="1" /> <edge from-layer="2380" from-port="2" to-layer="2382" to-port="0" /> <edge from-layer="2381" from-port="0" to-layer="2382" to-port="1" /> <edge from-layer="2382" from-port="2" to-layer="2390" to-port="0" /> <edge from-layer="2383" from-port="0" to-layer="2384" to-port="0" /> <edge from-layer="2384" from-port="1" to-layer="2387" to-port="0" /> <edge from-layer="2385" from-port="0" to-layer="2386" to-port="0" /> <edge from-layer="2386" from-port="1" to-layer="2387" to-port="1" /> <edge from-layer="2387" from-port="2" to-layer="2389" to-port="0" /> <edge from-layer="2388" from-port="0" to-layer="2389" to-port="1" /> <edge from-layer="2389" from-port="2" to-layer="2390" to-port="1" /> <edge from-layer="2390" from-port="2" to-layer="2392" to-port="0" /> <edge from-layer="2391" from-port="0" to-layer="2392" to-port="1" /> <edge from-layer="2392" from-port="2" to-layer="2393" to-port="0" /> <edge from-layer="2393" from-port="1" to-layer="2401" to-port="0" /> <edge from-layer="2394" from-port="0" to-layer="2395" to-port="0" /> <edge from-layer="2395" from-port="1" to-layer="2398" to-port="0" /> <edge from-layer="2396" from-port="0" to-layer="2397" to-port="0" /> <edge from-layer="2397" from-port="1" to-layer="2398" to-port="1" /> <edge from-layer="2398" from-port="2" to-layer="2400" to-port="0" /> <edge from-layer="2399" from-port="0" to-layer="2400" to-port="1" /> <edge from-layer="2400" from-port="2" to-layer="2401" to-port="1" /> <edge from-layer="2401" from-port="2" to-layer="2403" to-port="0" /> <edge from-layer="2402" from-port="0" to-layer="2403" to-port="1" /> <edge from-layer="2403" from-port="2" to-layer="2606" to-port="0" /> <edge from-layer="2404" from-port="0" to-layer="2405" to-port="1" /> <edge from-layer="2405" from-port="2" to-layer="2407" to-port="0" /> <edge from-layer="2406" from-port="0" to-layer="2407" to-port="1" /> <edge from-layer="2407" from-port="2" to-layer="2409" to-port="0" /> <edge from-layer="2408" from-port="0" to-layer="2409" to-port="1" /> <edge from-layer="2409" from-port="2" to-layer="2417" to-port="0" /> <edge from-layer="2410" from-port="0" to-layer="2411" to-port="0" /> <edge from-layer="2411" from-port="1" to-layer="2414" to-port="0" /> <edge from-layer="2412" from-port="0" to-layer="2413" to-port="0" /> <edge from-layer="2413" from-port="1" to-layer="2414" to-port="1" /> <edge from-layer="2414" from-port="2" to-layer="2416" to-port="0" /> <edge from-layer="2415" from-port="0" to-layer="2416" to-port="1" /> <edge from-layer="2416" from-port="2" to-layer="2417" to-port="1" /> <edge from-layer="2417" from-port="2" to-layer="2419" to-port="0" /> <edge from-layer="2418" from-port="0" to-layer="2419" to-port="1" /> <edge from-layer="2419" from-port="2" to-layer="2421" to-port="0" /> <edge from-layer="2420" from-port="0" to-layer="2421" to-port="1" /> <edge from-layer="2421" from-port="2" to-layer="2425" to-port="0" /> <edge from-layer="2421" from-port="2" to-layer="2437" to-port="0" /> <edge from-layer="2421" from-port="2" to-layer="2543" to-port="0" /> <edge from-layer="2422" from-port="0" to-layer="2425" to-port="1" /> <edge from-layer="2423" from-port="0" to-layer="2425" to-port="2" /> <edge from-layer="2424" from-port="0" to-layer="2425" to-port="3" /> <edge from-layer="2425" from-port="4" to-layer="2427" to-port="0" /> <edge from-layer="2426" from-port="0" to-layer="2427" to-port="1" /> <edge from-layer="2427" from-port="2" to-layer="2496" to-port="0" /> <edge from-layer="2427" from-port="2" to-layer="2431" to-port="0" /> <edge from-layer="2428" from-port="0" to-layer="2431" to-port="1" /> <edge from-layer="2429" from-port="0" to-layer="2431" to-port="2" /> <edge from-layer="2430" from-port="0" to-layer="2431" to-port="3" /> <edge from-layer="2431" from-port="4" to-layer="2466" to-port="0" /> <edge from-layer="2431" from-port="4" to-layer="2482" to-port="0" /> <edge from-layer="2431" from-port="4" to-layer="2477" to-port="0" /> <edge from-layer="2431" from-port="4" to-layer="2463" to-port="0" /> <edge from-layer="2432" from-port="0" to-layer="2459" to-port="0" /> <edge from-layer="2433" from-port="0" to-layer="2459" to-port="1" /> <edge from-layer="2434" from-port="0" to-layer="2437" to-port="1" /> <edge from-layer="2435" from-port="0" to-layer="2437" to-port="2" /> <edge from-layer="2436" from-port="0" to-layer="2437" to-port="3" /> <edge from-layer="2437" from-port="4" to-layer="2439" to-port="0" /> <edge from-layer="2438" from-port="0" to-layer="2439" to-port="1" /> <edge from-layer="2439" from-port="2" to-layer="2528" to-port="0" /> <edge from-layer="2439" from-port="2" to-layer="2501" to-port="0" /> <edge from-layer="2439" from-port="2" to-layer="2440" to-port="0" /> <edge from-layer="2440" from-port="1" to-layer="2442" to-port="0" /> <edge from-layer="2441" from-port="0" to-layer="2442" to-port="2" /> <edge from-layer="2442" from-port="3" to-layer="2455" to-port="0" /> <edge from-layer="2443" from-port="0" to-layer="2448" to-port="0" /> <edge from-layer="2444" from-port="0" to-layer="2447" to-port="1" /> <edge from-layer="2445" from-port="0" to-layer="2447" to-port="2" /> <edge from-layer="2446" from-port="0" to-layer="2447" to-port="3" /> <edge from-layer="2447" from-port="4" to-layer="2448" to-port="1" /> <edge from-layer="2448" from-port="2" to-layer="2449" to-port="0" /> <edge from-layer="2449" from-port="1" to-layer="2451" to-port="0" /> <edge from-layer="2450" from-port="0" to-layer="2451" to-port="2" /> <edge from-layer="2451" from-port="3" to-layer="2530" to-port="0" /> <edge from-layer="2451" from-port="3" to-layer="2452" to-port="0" /> <edge from-layer="2452" from-port="1" to-layer="2454" to-port="0" /> <edge from-layer="2453" from-port="0" to-layer="2454" to-port="2" /> <edge from-layer="2454" from-port="3" to-layer="2455" to-port="1" /> <edge from-layer="2455" from-port="2" to-layer="2457" to-port="0" /> <edge from-layer="2456" from-port="0" to-layer="2457" to-port="1" /> <edge from-layer="2457" from-port="2" to-layer="2459" to-port="2" /> <edge from-layer="2457" from-port="2" to-layer="2487" to-port="2" /> <edge from-layer="2458" from-port="0" to-layer="2459" to-port="3" /> <edge from-layer="2459" from-port="4" to-layer="2461" to-port="0" /> <edge from-layer="2460" from-port="0" to-layer="2461" to-port="2" /> <edge from-layer="2461" from-port="3" to-layer="2462" to-port="0" /> <edge from-layer="2462" from-port="2" to-layer="2502" to-port="1" /> <edge from-layer="2462" from-port="2" to-layer="2463" to-port="1" /> <edge from-layer="2463" from-port="2" to-layer="2492" to-port="0" /> <edge from-layer="2464" from-port="0" to-layer="2474" to-port="0" /> <edge from-layer="2465" from-port="0" to-layer="2474" to-port="1" /> <edge from-layer="2466" from-port="1" to-layer="2468" to-port="0" /> <edge from-layer="2467" from-port="0" to-layer="2468" to-port="2" /> <edge from-layer="2468" from-port="3" to-layer="2469" to-port="0" /> <edge from-layer="2469" from-port="2" to-layer="2470" to-port="0" /> <edge from-layer="2470" from-port="1" to-layer="2472" to-port="0" /> <edge from-layer="2471" from-port="0" to-layer="2472" to-port="1" /> <edge from-layer="2472" from-port="2" to-layer="2474" to-port="2" /> <edge from-layer="2473" from-port="0" to-layer="2474" to-port="3" /> <edge from-layer="2474" from-port="4" to-layer="2477" to-port="1" /> <edge from-layer="2474" from-port="4" to-layer="2482" to-port="2" /> <edge from-layer="2475" from-port="0" to-layer="2477" to-port="2" /> <edge from-layer="2476" from-port="0" to-layer="2477" to-port="3" /> <edge from-layer="2477" from-port="4" to-layer="2479" to-port="0" /> <edge from-layer="2478" from-port="0" to-layer="2479" to-port="1" /> <edge from-layer="2479" from-port="2" to-layer="2483" to-port="0" /> <edge from-layer="2480" from-port="0" to-layer="2482" to-port="1" /> <edge from-layer="2481" from-port="0" to-layer="2482" to-port="3" /> <edge from-layer="2482" from-port="4" to-layer="2483" to-port="1" /> <edge from-layer="2483" from-port="2" to-layer="2491" to-port="0" /> <edge from-layer="2484" from-port="0" to-layer="2487" to-port="0" /> <edge from-layer="2485" from-port="0" to-layer="2487" to-port="1" /> <edge from-layer="2486" from-port="0" to-layer="2487" to-port="3" /> <edge from-layer="2487" from-port="4" to-layer="2489" to-port="0" /> <edge from-layer="2488" from-port="0" to-layer="2489" to-port="2" /> <edge from-layer="2489" from-port="3" to-layer="2490" to-port="0" /> <edge from-layer="2490" from-port="2" to-layer="2523" to-port="1" /> <edge from-layer="2490" from-port="2" to-layer="2491" to-port="1" /> <edge from-layer="2491" from-port="2" to-layer="2492" to-port="1" /> <edge from-layer="2492" from-port="2" to-layer="2497" to-port="0" /> <edge from-layer="2493" from-port="0" to-layer="2496" to-port="1" /> <edge from-layer="2494" from-port="0" to-layer="2496" to-port="2" /> <edge from-layer="2495" from-port="0" to-layer="2496" to-port="3" /> <edge from-layer="2496" from-port="4" to-layer="2497" to-port="1" /> <edge from-layer="2497" from-port="2" to-layer="2591" to-port="0" /> <edge from-layer="2497" from-port="2" to-layer="2553" to-port="0" /> <edge from-layer="2498" from-port="0" to-layer="2501" to-port="1" /> <edge from-layer="2499" from-port="0" to-layer="2501" to-port="2" /> <edge from-layer="2500" from-port="0" to-layer="2501" to-port="3" /> <edge from-layer="2501" from-port="4" to-layer="2502" to-port="0" /> <edge from-layer="2501" from-port="4" to-layer="2516" to-port="0" /> <edge from-layer="2501" from-port="4" to-layer="2521" to-port="0" /> <edge from-layer="2501" from-port="4" to-layer="2505" to-port="0" /> <edge from-layer="2502" from-port="2" to-layer="2524" to-port="0" /> <edge from-layer="2503" from-port="0" to-layer="2513" to-port="0" /> <edge from-layer="2504" from-port="0" to-layer="2513" to-port="1" /> <edge from-layer="2505" from-port="1" to-layer="2507" to-port="0" /> <edge from-layer="2506" from-port="0" to-layer="2507" to-port="2" /> <edge from-layer="2507" from-port="3" to-layer="2508" to-port="0" /> <edge from-layer="2508" from-port="2" to-layer="2509" to-port="0" /> <edge from-layer="2509" from-port="1" to-layer="2511" to-port="0" /> <edge from-layer="2510" from-port="0" to-layer="2511" to-port="1" /> <edge from-layer="2511" from-port="2" to-layer="2513" to-port="2" /> <edge from-layer="2512" from-port="0" to-layer="2513" to-port="3" /> <edge from-layer="2513" from-port="4" to-layer="2516" to-port="1" /> <edge from-layer="2513" from-port="4" to-layer="2521" to-port="2" /> <edge from-layer="2514" from-port="0" to-layer="2516" to-port="2" /> <edge from-layer="2515" from-port="0" to-layer="2516" to-port="3" /> <edge from-layer="2516" from-port="4" to-layer="2518" to-port="0" /> <edge from-layer="2517" from-port="0" to-layer="2518" to-port="1" /> <edge from-layer="2518" from-port="2" to-layer="2522" to-port="0" /> <edge from-layer="2519" from-port="0" to-layer="2521" to-port="1" /> <edge from-layer="2520" from-port="0" to-layer="2521" to-port="3" /> <edge from-layer="2521" from-port="4" to-layer="2522" to-port="1" /> <edge from-layer="2522" from-port="2" to-layer="2523" to-port="0" /> <edge from-layer="2523" from-port="2" to-layer="2524" to-port="1" /> <edge from-layer="2524" from-port="2" to-layer="2529" to-port="0" /> <edge from-layer="2525" from-port="0" to-layer="2528" to-port="1" /> <edge from-layer="2526" from-port="0" to-layer="2528" to-port="2" /> <edge from-layer="2527" from-port="0" to-layer="2528" to-port="3" /> <edge from-layer="2528" from-port="4" to-layer="2529" to-port="1" /> <edge from-layer="2529" from-port="2" to-layer="2530" to-port="1" /> <edge from-layer="2530" from-port="2" to-layer="10169" to-port="0" /> <edge from-layer="2530" from-port="2" to-layer="2591" to-port="1" /> <edge from-layer="2530" from-port="2" to-layer="2550" to-port="0" /> <edge from-layer="2531" from-port="0" to-layer="2536" to-port="0" /> <edge from-layer="2532" from-port="0" to-layer="2535" to-port="1" /> <edge from-layer="2533" from-port="0" to-layer="2535" to-port="2" /> <edge from-layer="2534" from-port="0" to-layer="2535" to-port="3" /> <edge from-layer="2535" from-port="4" to-layer="2536" to-port="1" /> <edge from-layer="2536" from-port="2" to-layer="2537" to-port="0" /> <edge from-layer="2537" from-port="1" to-layer="2539" to-port="0" /> <edge from-layer="2538" from-port="0" to-layer="2539" to-port="2" /> <edge from-layer="2539" from-port="3" to-layer="2546" to-port="0" /> <edge from-layer="2540" from-port="0" to-layer="2543" to-port="1" /> <edge from-layer="2541" from-port="0" to-layer="2543" to-port="2" /> <edge from-layer="2542" from-port="0" to-layer="2543" to-port="3" /> <edge from-layer="2543" from-port="4" to-layer="2545" to-port="0" /> <edge from-layer="2544" from-port="0" to-layer="2545" to-port="1" /> <edge from-layer="2545" from-port="2" to-layer="2546" to-port="1" /> <edge from-layer="2546" from-port="2" to-layer="2591" to-port="2" /> <edge from-layer="2546" from-port="2" to-layer="10168" to-port="0" /> <edge from-layer="2547" from-port="0" to-layer="2565" to-port="0" /> <edge from-layer="2548" from-port="0" to-layer="2560" to-port="0" /> <edge from-layer="2549" from-port="0" to-layer="2563" to-port="1" /> <edge from-layer="2549" from-port="0" to-layer="2560" to-port="1" /> <edge from-layer="2550" from-port="1" to-layer="2552" to-port="0" /> <edge from-layer="2551" from-port="0" to-layer="2552" to-port="2" /> <edge from-layer="2552" from-port="3" to-layer="2562" to-port="0" /> <edge from-layer="2552" from-port="3" to-layer="2556" to-port="0" /> <edge from-layer="2553" from-port="1" to-layer="2580" to-port="0" /> <edge from-layer="2553" from-port="1" to-layer="2555" to-port="0" /> <edge from-layer="2554" from-port="0" to-layer="2555" to-port="2" /> <edge from-layer="2555" from-port="3" to-layer="2556" to-port="1" /> <edge from-layer="2556" from-port="2" to-layer="2558" to-port="0" /> <edge from-layer="2557" from-port="0" to-layer="2562" to-port="1" /> <edge from-layer="2557" from-port="0" to-layer="2558" to-port="1" /> <edge from-layer="2558" from-port="2" to-layer="2560" to-port="2" /> <edge from-layer="2559" from-port="0" to-layer="2560" to-port="3" /> <edge from-layer="2559" from-port="0" to-layer="2563" to-port="3" /> <edge from-layer="2560" from-port="4" to-layer="2565" to-port="1" /> <edge from-layer="2561" from-port="0" to-layer="2563" to-port="0" /> <edge from-layer="2562" from-port="2" to-layer="2563" to-port="2" /> <edge from-layer="2562" from-port="2" to-layer="2570" to-port="2" /> <edge from-layer="2563" from-port="4" to-layer="2565" to-port="2" /> <edge from-layer="2564" from-port="0" to-layer="2565" to-port="3" /> <edge from-layer="2565" from-port="4" to-layer="2572" to-port="0" /> <edge from-layer="2566" from-port="0" to-layer="2572" to-port="1" /> <edge from-layer="2567" from-port="0" to-layer="2570" to-port="0" /> <edge from-layer="2568" from-port="0" to-layer="2570" to-port="1" /> <edge from-layer="2569" from-port="0" to-layer="2570" to-port="3" /> <edge from-layer="2570" from-port="4" to-layer="2572" to-port="2" /> <edge from-layer="2571" from-port="0" to-layer="2572" to-port="3" /> <edge from-layer="2572" from-port="4" to-layer="2575" to-port="0" /> <edge from-layer="2573" from-port="0" to-layer="2575" to-port="1" /> <edge from-layer="2574" from-port="0" to-layer="2575" to-port="2" /> <edge from-layer="2575" from-port="3" to-layer="2576" to-port="0" /> <edge from-layer="2576" from-port="2" to-layer="2577" to-port="0" /> <edge from-layer="2576" from-port="2" to-layer="2590" to-port="0" /> <edge from-layer="2577" from-port="1" to-layer="2589" to-port="0" /> <edge from-layer="2578" from-port="0" to-layer="2580" to-port="1" /> <edge from-layer="2579" from-port="0" to-layer="2580" to-port="2" /> <edge from-layer="2580" from-port="3" to-layer="2584" to-port="0" /> <edge from-layer="2581" from-port="0" to-layer="2584" to-port="1" /> <edge from-layer="2582" from-port="0" to-layer="2584" to-port="2" /> <edge from-layer="2583" from-port="0" to-layer="2584" to-port="3" /> <edge from-layer="2584" from-port="4" to-layer="2588" to-port="2" /> <edge from-layer="2584" from-port="4" to-layer="2586" to-port="0" /> <edge from-layer="2585" from-port="0" to-layer="2586" to-port="1" /> <edge from-layer="2586" from-port="2" to-layer="2588" to-port="0" /> <edge from-layer="2587" from-port="0" to-layer="2588" to-port="1" /> <edge from-layer="2588" from-port="3" to-layer="2589" to-port="1" /> <edge from-layer="2589" from-port="2" to-layer="2590" to-port="1" /> <edge from-layer="2590" from-port="2" to-layer="2591" to-port="3" /> <edge from-layer="2591" from-port="4" to-layer="2593" to-port="0" /> <edge from-layer="2592" from-port="0" to-layer="2593" to-port="1" /> <edge from-layer="2593" from-port="2" to-layer="2595" to-port="0" /> <edge from-layer="2594" from-port="0" to-layer="2595" to-port="1" /> <edge from-layer="2595" from-port="2" to-layer="2603" to-port="0" /> <edge from-layer="2596" from-port="0" to-layer="2597" to-port="0" /> <edge from-layer="2597" from-port="1" to-layer="2600" to-port="0" /> <edge from-layer="2598" from-port="0" to-layer="2599" to-port="0" /> <edge from-layer="2599" from-port="1" to-layer="2600" to-port="1" /> <edge from-layer="2600" from-port="2" to-layer="2602" to-port="0" /> <edge from-layer="2601" from-port="0" to-layer="2602" to-port="1" /> <edge from-layer="2602" from-port="2" to-layer="2603" to-port="1" /> <edge from-layer="2603" from-port="2" to-layer="2605" to-port="0" /> <edge from-layer="2604" from-port="0" to-layer="2605" to-port="1" /> <edge from-layer="2605" from-port="2" to-layer="2606" to-port="1" /> <edge from-layer="2606" from-port="2" to-layer="2607" to-port="0" /> <edge from-layer="2607" from-port="2" to-layer="2838" to-port="1" /> <edge from-layer="2607" from-port="2" to-layer="2636" to-port="0" /> <edge from-layer="2607" from-port="2" to-layer="2609" to-port="0" /> <edge from-layer="2608" from-port="0" to-layer="2609" to-port="1" /> <edge from-layer="2609" from-port="2" to-layer="2611" to-port="0" /> <edge from-layer="2610" from-port="0" to-layer="2611" to-port="1" /> <edge from-layer="2611" from-port="2" to-layer="2613" to-port="0" /> <edge from-layer="2612" from-port="0" to-layer="2613" to-port="1" /> <edge from-layer="2613" from-port="2" to-layer="2621" to-port="0" /> <edge from-layer="2614" from-port="0" to-layer="2615" to-port="0" /> <edge from-layer="2615" from-port="1" to-layer="2618" to-port="0" /> <edge from-layer="2616" from-port="0" to-layer="2617" to-port="0" /> <edge from-layer="2617" from-port="1" to-layer="2618" to-port="1" /> <edge from-layer="2618" from-port="2" to-layer="2620" to-port="0" /> <edge from-layer="2619" from-port="0" to-layer="2620" to-port="1" /> <edge from-layer="2620" from-port="2" to-layer="2621" to-port="1" /> <edge from-layer="2621" from-port="2" to-layer="2623" to-port="0" /> <edge from-layer="2622" from-port="0" to-layer="2623" to-port="1" /> <edge from-layer="2623" from-port="2" to-layer="2624" to-port="0" /> <edge from-layer="2624" from-port="1" to-layer="2632" to-port="0" /> <edge from-layer="2625" from-port="0" to-layer="2626" to-port="0" /> <edge from-layer="2626" from-port="1" to-layer="2629" to-port="0" /> <edge from-layer="2627" from-port="0" to-layer="2628" to-port="0" /> <edge from-layer="2628" from-port="1" to-layer="2629" to-port="1" /> <edge from-layer="2629" from-port="2" to-layer="2631" to-port="0" /> <edge from-layer="2630" from-port="0" to-layer="2631" to-port="1" /> <edge from-layer="2631" from-port="2" to-layer="2632" to-port="1" /> <edge from-layer="2632" from-port="2" to-layer="2634" to-port="0" /> <edge from-layer="2633" from-port="0" to-layer="2634" to-port="1" /> <edge from-layer="2634" from-port="2" to-layer="2837" to-port="0" /> <edge from-layer="2635" from-port="0" to-layer="2636" to-port="1" /> <edge from-layer="2636" from-port="2" to-layer="2638" to-port="0" /> <edge from-layer="2637" from-port="0" to-layer="2638" to-port="1" /> <edge from-layer="2638" from-port="2" to-layer="2640" to-port="0" /> <edge from-layer="2639" from-port="0" to-layer="2640" to-port="1" /> <edge from-layer="2640" from-port="2" to-layer="2648" to-port="0" /> <edge from-layer="2641" from-port="0" to-layer="2642" to-port="0" /> <edge from-layer="2642" from-port="1" to-layer="2645" to-port="0" /> <edge from-layer="2643" from-port="0" to-layer="2644" to-port="0" /> <edge from-layer="2644" from-port="1" to-layer="2645" to-port="1" /> <edge from-layer="2645" from-port="2" to-layer="2647" to-port="0" /> <edge from-layer="2646" from-port="0" to-layer="2647" to-port="1" /> <edge from-layer="2647" from-port="2" to-layer="2648" to-port="1" /> <edge from-layer="2648" from-port="2" to-layer="2650" to-port="0" /> <edge from-layer="2649" from-port="0" to-layer="2650" to-port="1" /> <edge from-layer="2650" from-port="2" to-layer="2652" to-port="0" /> <edge from-layer="2651" from-port="0" to-layer="2652" to-port="1" /> <edge from-layer="2652" from-port="2" to-layer="2656" to-port="0" /> <edge from-layer="2652" from-port="2" to-layer="2668" to-port="0" /> <edge from-layer="2652" from-port="2" to-layer="2774" to-port="0" /> <edge from-layer="2653" from-port="0" to-layer="2656" to-port="1" /> <edge from-layer="2654" from-port="0" to-layer="2656" to-port="2" /> <edge from-layer="2655" from-port="0" to-layer="2656" to-port="3" /> <edge from-layer="2656" from-port="4" to-layer="2658" to-port="0" /> <edge from-layer="2657" from-port="0" to-layer="2658" to-port="1" /> <edge from-layer="2658" from-port="2" to-layer="2727" to-port="0" /> <edge from-layer="2658" from-port="2" to-layer="2662" to-port="0" /> <edge from-layer="2659" from-port="0" to-layer="2662" to-port="1" /> <edge from-layer="2660" from-port="0" to-layer="2662" to-port="2" /> <edge from-layer="2661" from-port="0" to-layer="2662" to-port="3" /> <edge from-layer="2662" from-port="4" to-layer="2697" to-port="0" /> <edge from-layer="2662" from-port="4" to-layer="2708" to-port="0" /> <edge from-layer="2662" from-port="4" to-layer="2713" to-port="0" /> <edge from-layer="2662" from-port="4" to-layer="2694" to-port="0" /> <edge from-layer="2663" from-port="0" to-layer="2690" to-port="0" /> <edge from-layer="2664" from-port="0" to-layer="2690" to-port="1" /> <edge from-layer="2665" from-port="0" to-layer="2668" to-port="1" /> <edge from-layer="2666" from-port="0" to-layer="2668" to-port="2" /> <edge from-layer="2667" from-port="0" to-layer="2668" to-port="3" /> <edge from-layer="2668" from-port="4" to-layer="2670" to-port="0" /> <edge from-layer="2669" from-port="0" to-layer="2670" to-port="1" /> <edge from-layer="2670" from-port="2" to-layer="2759" to-port="0" /> <edge from-layer="2670" from-port="2" to-layer="2671" to-port="0" /> <edge from-layer="2670" from-port="2" to-layer="2732" to-port="0" /> <edge from-layer="2671" from-port="1" to-layer="2673" to-port="0" /> <edge from-layer="2672" from-port="0" to-layer="2673" to-port="2" /> <edge from-layer="2673" from-port="3" to-layer="2686" to-port="0" /> <edge from-layer="2674" from-port="0" to-layer="2679" to-port="0" /> <edge from-layer="2675" from-port="0" to-layer="2678" to-port="1" /> <edge from-layer="2676" from-port="0" to-layer="2678" to-port="2" /> <edge from-layer="2677" from-port="0" to-layer="2678" to-port="3" /> <edge from-layer="2678" from-port="4" to-layer="2679" to-port="1" /> <edge from-layer="2679" from-port="2" to-layer="2680" to-port="0" /> <edge from-layer="2680" from-port="1" to-layer="2682" to-port="0" /> <edge from-layer="2681" from-port="0" to-layer="2682" to-port="2" /> <edge from-layer="2682" from-port="3" to-layer="2683" to-port="0" /> <edge from-layer="2682" from-port="3" to-layer="2761" to-port="0" /> <edge from-layer="2683" from-port="1" to-layer="2685" to-port="0" /> <edge from-layer="2684" from-port="0" to-layer="2685" to-port="2" /> <edge from-layer="2685" from-port="3" to-layer="2686" to-port="1" /> <edge from-layer="2686" from-port="2" to-layer="2688" to-port="0" /> <edge from-layer="2687" from-port="0" to-layer="2688" to-port="1" /> <edge from-layer="2688" from-port="2" to-layer="2690" to-port="2" /> <edge from-layer="2688" from-port="2" to-layer="2718" to-port="2" /> <edge from-layer="2689" from-port="0" to-layer="2690" to-port="3" /> <edge from-layer="2690" from-port="4" to-layer="2692" to-port="0" /> <edge from-layer="2691" from-port="0" to-layer="2692" to-port="2" /> <edge from-layer="2692" from-port="3" to-layer="2693" to-port="0" /> <edge from-layer="2693" from-port="2" to-layer="2733" to-port="1" /> <edge from-layer="2693" from-port="2" to-layer="2694" to-port="1" /> <edge from-layer="2694" from-port="2" to-layer="2723" to-port="0" /> <edge from-layer="2695" from-port="0" to-layer="2705" to-port="0" /> <edge from-layer="2696" from-port="0" to-layer="2705" to-port="1" /> <edge from-layer="2697" from-port="1" to-layer="2699" to-port="0" /> <edge from-layer="2698" from-port="0" to-layer="2699" to-port="2" /> <edge from-layer="2699" from-port="3" to-layer="2700" to-port="0" /> <edge from-layer="2700" from-port="2" to-layer="2701" to-port="0" /> <edge from-layer="2701" from-port="1" to-layer="2703" to-port="0" /> <edge from-layer="2702" from-port="0" to-layer="2703" to-port="1" /> <edge from-layer="2703" from-port="2" to-layer="2705" to-port="2" /> <edge from-layer="2704" from-port="0" to-layer="2705" to-port="3" /> <edge from-layer="2705" from-port="4" to-layer="2713" to-port="2" /> <edge from-layer="2705" from-port="4" to-layer="2708" to-port="1" /> <edge from-layer="2706" from-port="0" to-layer="2708" to-port="2" /> <edge from-layer="2707" from-port="0" to-layer="2708" to-port="3" /> <edge from-layer="2708" from-port="4" to-layer="2710" to-port="0" /> <edge from-layer="2709" from-port="0" to-layer="2710" to-port="1" /> <edge from-layer="2710" from-port="2" to-layer="2714" to-port="0" /> <edge from-layer="2711" from-port="0" to-layer="2713" to-port="1" /> <edge from-layer="2712" from-port="0" to-layer="2713" to-port="3" /> <edge from-layer="2713" from-port="4" to-layer="2714" to-port="1" /> <edge from-layer="2714" from-port="2" to-layer="2722" to-port="0" /> <edge from-layer="2715" from-port="0" to-layer="2718" to-port="0" /> <edge from-layer="2716" from-port="0" to-layer="2718" to-port="1" /> <edge from-layer="2717" from-port="0" to-layer="2718" to-port="3" /> <edge from-layer="2718" from-port="4" to-layer="2720" to-port="0" /> <edge from-layer="2719" from-port="0" to-layer="2720" to-port="2" /> <edge from-layer="2720" from-port="3" to-layer="2721" to-port="0" /> <edge from-layer="2721" from-port="2" to-layer="2722" to-port="1" /> <edge from-layer="2721" from-port="2" to-layer="2754" to-port="1" /> <edge from-layer="2722" from-port="2" to-layer="2723" to-port="1" /> <edge from-layer="2723" from-port="2" to-layer="2728" to-port="0" /> <edge from-layer="2724" from-port="0" to-layer="2727" to-port="1" /> <edge from-layer="2725" from-port="0" to-layer="2727" to-port="2" /> <edge from-layer="2726" from-port="0" to-layer="2727" to-port="3" /> <edge from-layer="2727" from-port="4" to-layer="2728" to-port="1" /> <edge from-layer="2728" from-port="2" to-layer="2784" to-port="0" /> <edge from-layer="2728" from-port="2" to-layer="2822" to-port="0" /> <edge from-layer="2729" from-port="0" to-layer="2732" to-port="1" /> <edge from-layer="2730" from-port="0" to-layer="2732" to-port="2" /> <edge from-layer="2731" from-port="0" to-layer="2732" to-port="3" /> <edge from-layer="2732" from-port="4" to-layer="2736" to-port="0" /> <edge from-layer="2732" from-port="4" to-layer="2733" to-port="0" /> <edge from-layer="2732" from-port="4" to-layer="2752" to-port="0" /> <edge from-layer="2732" from-port="4" to-layer="2747" to-port="0" /> <edge from-layer="2733" from-port="2" to-layer="2755" to-port="0" /> <edge from-layer="2734" from-port="0" to-layer="2744" to-port="0" /> <edge from-layer="2735" from-port="0" to-layer="2744" to-port="1" /> <edge from-layer="2736" from-port="1" to-layer="2738" to-port="0" /> <edge from-layer="2737" from-port="0" to-layer="2738" to-port="2" /> <edge from-layer="2738" from-port="3" to-layer="2739" to-port="0" /> <edge from-layer="2739" from-port="2" to-layer="2740" to-port="0" /> <edge from-layer="2740" from-port="1" to-layer="2742" to-port="0" /> <edge from-layer="2741" from-port="0" to-layer="2742" to-port="1" /> <edge from-layer="2742" from-port="2" to-layer="2744" to-port="2" /> <edge from-layer="2743" from-port="0" to-layer="2744" to-port="3" /> <edge from-layer="2744" from-port="4" to-layer="2747" to-port="1" /> <edge from-layer="2744" from-port="4" to-layer="2752" to-port="2" /> <edge from-layer="2745" from-port="0" to-layer="2747" to-port="2" /> <edge from-layer="2746" from-port="0" to-layer="2747" to-port="3" /> <edge from-layer="2747" from-port="4" to-layer="2749" to-port="0" /> <edge from-layer="2748" from-port="0" to-layer="2749" to-port="1" /> <edge from-layer="2749" from-port="2" to-layer="2753" to-port="0" /> <edge from-layer="2750" from-port="0" to-layer="2752" to-port="1" /> <edge from-layer="2751" from-port="0" to-layer="2752" to-port="3" /> <edge from-layer="2752" from-port="4" to-layer="2753" to-port="1" /> <edge from-layer="2753" from-port="2" to-layer="2754" to-port="0" /> <edge from-layer="2754" from-port="2" to-layer="2755" to-port="1" /> <edge from-layer="2755" from-port="2" to-layer="2760" to-port="0" /> <edge from-layer="2756" from-port="0" to-layer="2759" to-port="1" /> <edge from-layer="2757" from-port="0" to-layer="2759" to-port="2" /> <edge from-layer="2758" from-port="0" to-layer="2759" to-port="3" /> <edge from-layer="2759" from-port="4" to-layer="2760" to-port="1" /> <edge from-layer="2760" from-port="2" to-layer="2761" to-port="1" /> <edge from-layer="2761" from-port="2" to-layer="10167" to-port="0" /> <edge from-layer="2761" from-port="2" to-layer="2822" to-port="1" /> <edge from-layer="2761" from-port="2" to-layer="2781" to-port="0" /> <edge from-layer="2762" from-port="0" to-layer="2767" to-port="0" /> <edge from-layer="2763" from-port="0" to-layer="2766" to-port="1" /> <edge from-layer="2764" from-port="0" to-layer="2766" to-port="2" /> <edge from-layer="2765" from-port="0" to-layer="2766" to-port="3" /> <edge from-layer="2766" from-port="4" to-layer="2767" to-port="1" /> <edge from-layer="2767" from-port="2" to-layer="2768" to-port="0" /> <edge from-layer="2768" from-port="1" to-layer="2770" to-port="0" /> <edge from-layer="2769" from-port="0" to-layer="2770" to-port="2" /> <edge from-layer="2770" from-port="3" to-layer="2777" to-port="0" /> <edge from-layer="2771" from-port="0" to-layer="2774" to-port="1" /> <edge from-layer="2772" from-port="0" to-layer="2774" to-port="2" /> <edge from-layer="2773" from-port="0" to-layer="2774" to-port="3" /> <edge from-layer="2774" from-port="4" to-layer="2776" to-port="0" /> <edge from-layer="2775" from-port="0" to-layer="2776" to-port="1" /> <edge from-layer="2776" from-port="2" to-layer="2777" to-port="1" /> <edge from-layer="2777" from-port="2" to-layer="10166" to-port="0" /> <edge from-layer="2777" from-port="2" to-layer="2822" to-port="2" /> <edge from-layer="2778" from-port="0" to-layer="2796" to-port="0" /> <edge from-layer="2779" from-port="0" to-layer="2791" to-port="0" /> <edge from-layer="2780" from-port="0" to-layer="2794" to-port="1" /> <edge from-layer="2780" from-port="0" to-layer="2791" to-port="1" /> <edge from-layer="2781" from-port="1" to-layer="2783" to-port="0" /> <edge from-layer="2782" from-port="0" to-layer="2783" to-port="2" /> <edge from-layer="2783" from-port="3" to-layer="2787" to-port="0" /> <edge from-layer="2783" from-port="3" to-layer="2793" to-port="0" /> <edge from-layer="2784" from-port="1" to-layer="2786" to-port="0" /> <edge from-layer="2784" from-port="1" to-layer="2811" to-port="0" /> <edge from-layer="2785" from-port="0" to-layer="2786" to-port="2" /> <edge from-layer="2786" from-port="3" to-layer="2787" to-port="1" /> <edge from-layer="2787" from-port="2" to-layer="2789" to-port="0" /> <edge from-layer="2788" from-port="0" to-layer="2789" to-port="1" /> <edge from-layer="2788" from-port="0" to-layer="2793" to-port="1" /> <edge from-layer="2789" from-port="2" to-layer="2791" to-port="2" /> <edge from-layer="2790" from-port="0" to-layer="2794" to-port="3" /> <edge from-layer="2790" from-port="0" to-layer="2791" to-port="3" /> <edge from-layer="2791" from-port="4" to-layer="2796" to-port="1" /> <edge from-layer="2792" from-port="0" to-layer="2794" to-port="0" /> <edge from-layer="2793" from-port="2" to-layer="2801" to-port="2" /> <edge from-layer="2793" from-port="2" to-layer="2794" to-port="2" /> <edge from-layer="2794" from-port="4" to-layer="2796" to-port="2" /> <edge from-layer="2795" from-port="0" to-layer="2796" to-port="3" /> <edge from-layer="2796" from-port="4" to-layer="2803" to-port="0" /> <edge from-layer="2797" from-port="0" to-layer="2803" to-port="1" /> <edge from-layer="2798" from-port="0" to-layer="2801" to-port="0" /> <edge from-layer="2799" from-port="0" to-layer="2801" to-port="1" /> <edge from-layer="2800" from-port="0" to-layer="2801" to-port="3" /> <edge from-layer="2801" from-port="4" to-layer="2803" to-port="2" /> <edge from-layer="2802" from-port="0" to-layer="2803" to-port="3" /> <edge from-layer="2803" from-port="4" to-layer="2806" to-port="0" /> <edge from-layer="2804" from-port="0" to-layer="2806" to-port="1" /> <edge from-layer="2805" from-port="0" to-layer="2806" to-port="2" /> <edge from-layer="2806" from-port="3" to-layer="2807" to-port="0" /> <edge from-layer="2807" from-port="2" to-layer="2808" to-port="0" /> <edge from-layer="2807" from-port="2" to-layer="2821" to-port="0" /> <edge from-layer="2808" from-port="1" to-layer="2820" to-port="0" /> <edge from-layer="2809" from-port="0" to-layer="2811" to-port="1" /> <edge from-layer="2810" from-port="0" to-layer="2811" to-port="2" /> <edge from-layer="2811" from-port="3" to-layer="2815" to-port="0" /> <edge from-layer="2812" from-port="0" to-layer="2815" to-port="1" /> <edge from-layer="2813" from-port="0" to-layer="2815" to-port="2" /> <edge from-layer="2814" from-port="0" to-layer="2815" to-port="3" /> <edge from-layer="2815" from-port="4" to-layer="2817" to-port="0" /> <edge from-layer="2815" from-port="4" to-layer="2819" to-port="2" /> <edge from-layer="2816" from-port="0" to-layer="2817" to-port="1" /> <edge from-layer="2817" from-port="2" to-layer="2819" to-port="0" /> <edge from-layer="2818" from-port="0" to-layer="2819" to-port="1" /> <edge from-layer="2819" from-port="3" to-layer="2820" to-port="1" /> <edge from-layer="2820" from-port="2" to-layer="2821" to-port="1" /> <edge from-layer="2821" from-port="2" to-layer="2822" to-port="3" /> <edge from-layer="2822" from-port="4" to-layer="2824" to-port="0" /> <edge from-layer="2823" from-port="0" to-layer="2824" to-port="1" /> <edge from-layer="2824" from-port="2" to-layer="2826" to-port="0" /> <edge from-layer="2825" from-port="0" to-layer="2826" to-port="1" /> <edge from-layer="2826" from-port="2" to-layer="2834" to-port="0" /> <edge from-layer="2827" from-port="0" to-layer="2828" to-port="0" /> <edge from-layer="2828" from-port="1" to-layer="2831" to-port="0" /> <edge from-layer="2829" from-port="0" to-layer="2830" to-port="0" /> <edge from-layer="2830" from-port="1" to-layer="2831" to-port="1" /> <edge from-layer="2831" from-port="2" to-layer="2833" to-port="0" /> <edge from-layer="2832" from-port="0" to-layer="2833" to-port="1" /> <edge from-layer="2833" from-port="2" to-layer="2834" to-port="1" /> <edge from-layer="2834" from-port="2" to-layer="2836" to-port="0" /> <edge from-layer="2835" from-port="0" to-layer="2836" to-port="1" /> <edge from-layer="2836" from-port="2" to-layer="2837" to-port="1" /> <edge from-layer="2837" from-port="2" to-layer="2838" to-port="0" /> <edge from-layer="2838" from-port="2" to-layer="3069" to-port="1" /> <edge from-layer="2838" from-port="2" to-layer="2840" to-port="0" /> <edge from-layer="2838" from-port="2" to-layer="2867" to-port="0" /> <edge from-layer="2839" from-port="0" to-layer="2840" to-port="1" /> <edge from-layer="2840" from-port="2" to-layer="2842" to-port="0" /> <edge from-layer="2841" from-port="0" to-layer="2842" to-port="1" /> <edge from-layer="2842" from-port="2" to-layer="2844" to-port="0" /> <edge from-layer="2843" from-port="0" to-layer="2844" to-port="1" /> <edge from-layer="2844" from-port="2" to-layer="2852" to-port="0" /> <edge from-layer="2845" from-port="0" to-layer="2846" to-port="0" /> <edge from-layer="2846" from-port="1" to-layer="2849" to-port="0" /> <edge from-layer="2847" from-port="0" to-layer="2848" to-port="0" /> <edge from-layer="2848" from-port="1" to-layer="2849" to-port="1" /> <edge from-layer="2849" from-port="2" to-layer="2851" to-port="0" /> <edge from-layer="2850" from-port="0" to-layer="2851" to-port="1" /> <edge from-layer="2851" from-port="2" to-layer="2852" to-port="1" /> <edge from-layer="2852" from-port="2" to-layer="2854" to-port="0" /> <edge from-layer="2853" from-port="0" to-layer="2854" to-port="1" /> <edge from-layer="2854" from-port="2" to-layer="2855" to-port="0" /> <edge from-layer="2855" from-port="1" to-layer="2863" to-port="0" /> <edge from-layer="2856" from-port="0" to-layer="2857" to-port="0" /> <edge from-layer="2857" from-port="1" to-layer="2860" to-port="0" /> <edge from-layer="2858" from-port="0" to-layer="2859" to-port="0" /> <edge from-layer="2859" from-port="1" to-layer="2860" to-port="1" /> <edge from-layer="2860" from-port="2" to-layer="2862" to-port="0" /> <edge from-layer="2861" from-port="0" to-layer="2862" to-port="1" /> <edge from-layer="2862" from-port="2" to-layer="2863" to-port="1" /> <edge from-layer="2863" from-port="2" to-layer="2865" to-port="0" /> <edge from-layer="2864" from-port="0" to-layer="2865" to-port="1" /> <edge from-layer="2865" from-port="2" to-layer="3068" to-port="0" /> <edge from-layer="2866" from-port="0" to-layer="2867" to-port="1" /> <edge from-layer="2867" from-port="2" to-layer="2869" to-port="0" /> <edge from-layer="2868" from-port="0" to-layer="2869" to-port="1" /> <edge from-layer="2869" from-port="2" to-layer="2871" to-port="0" /> <edge from-layer="2870" from-port="0" to-layer="2871" to-port="1" /> <edge from-layer="2871" from-port="2" to-layer="2879" to-port="0" /> <edge from-layer="2872" from-port="0" to-layer="2873" to-port="0" /> <edge from-layer="2873" from-port="1" to-layer="2876" to-port="0" /> <edge from-layer="2874" from-port="0" to-layer="2875" to-port="0" /> <edge from-layer="2875" from-port="1" to-layer="2876" to-port="1" /> <edge from-layer="2876" from-port="2" to-layer="2878" to-port="0" /> <edge from-layer="2877" from-port="0" to-layer="2878" to-port="1" /> <edge from-layer="2878" from-port="2" to-layer="2879" to-port="1" /> <edge from-layer="2879" from-port="2" to-layer="2881" to-port="0" /> <edge from-layer="2880" from-port="0" to-layer="2881" to-port="1" /> <edge from-layer="2881" from-port="2" to-layer="2883" to-port="0" /> <edge from-layer="2882" from-port="0" to-layer="2883" to-port="1" /> <edge from-layer="2883" from-port="2" to-layer="3005" to-port="0" /> <edge from-layer="2883" from-port="2" to-layer="2899" to-port="0" /> <edge from-layer="2883" from-port="2" to-layer="2887" to-port="0" /> <edge from-layer="2884" from-port="0" to-layer="2887" to-port="1" /> <edge from-layer="2885" from-port="0" to-layer="2887" to-port="2" /> <edge from-layer="2886" from-port="0" to-layer="2887" to-port="3" /> <edge from-layer="2887" from-port="4" to-layer="2889" to-port="0" /> <edge from-layer="2888" from-port="0" to-layer="2889" to-port="1" /> <edge from-layer="2889" from-port="2" to-layer="2893" to-port="0" /> <edge from-layer="2889" from-port="2" to-layer="2958" to-port="0" /> <edge from-layer="2890" from-port="0" to-layer="2893" to-port="1" /> <edge from-layer="2891" from-port="0" to-layer="2893" to-port="2" /> <edge from-layer="2892" from-port="0" to-layer="2893" to-port="3" /> <edge from-layer="2893" from-port="4" to-layer="2939" to-port="0" /> <edge from-layer="2893" from-port="4" to-layer="2928" to-port="0" /> <edge from-layer="2893" from-port="4" to-layer="2925" to-port="0" /> <edge from-layer="2893" from-port="4" to-layer="2944" to-port="0" /> <edge from-layer="2894" from-port="0" to-layer="2921" to-port="0" /> <edge from-layer="2895" from-port="0" to-layer="2921" to-port="1" /> <edge from-layer="2896" from-port="0" to-layer="2899" to-port="1" /> <edge from-layer="2897" from-port="0" to-layer="2899" to-port="2" /> <edge from-layer="2898" from-port="0" to-layer="2899" to-port="3" /> <edge from-layer="2899" from-port="4" to-layer="2901" to-port="0" /> <edge from-layer="2900" from-port="0" to-layer="2901" to-port="1" /> <edge from-layer="2901" from-port="2" to-layer="2990" to-port="0" /> <edge from-layer="2901" from-port="2" to-layer="2963" to-port="0" /> <edge from-layer="2901" from-port="2" to-layer="2902" to-port="0" /> <edge from-layer="2902" from-port="1" to-layer="2904" to-port="0" /> <edge from-layer="2903" from-port="0" to-layer="2904" to-port="2" /> <edge from-layer="2904" from-port="3" to-layer="2917" to-port="0" /> <edge from-layer="2905" from-port="0" to-layer="2910" to-port="0" /> <edge from-layer="2906" from-port="0" to-layer="2909" to-port="1" /> <edge from-layer="2907" from-port="0" to-layer="2909" to-port="2" /> <edge from-layer="2908" from-port="0" to-layer="2909" to-port="3" /> <edge from-layer="2909" from-port="4" to-layer="2910" to-port="1" /> <edge from-layer="2910" from-port="2" to-layer="2911" to-port="0" /> <edge from-layer="2911" from-port="1" to-layer="2913" to-port="0" /> <edge from-layer="2912" from-port="0" to-layer="2913" to-port="2" /> <edge from-layer="2913" from-port="3" to-layer="2992" to-port="0" /> <edge from-layer="2913" from-port="3" to-layer="2914" to-port="0" /> <edge from-layer="2914" from-port="1" to-layer="2916" to-port="0" /> <edge from-layer="2915" from-port="0" to-layer="2916" to-port="2" /> <edge from-layer="2916" from-port="3" to-layer="2917" to-port="1" /> <edge from-layer="2917" from-port="2" to-layer="2919" to-port="0" /> <edge from-layer="2918" from-port="0" to-layer="2919" to-port="1" /> <edge from-layer="2919" from-port="2" to-layer="2949" to-port="2" /> <edge from-layer="2919" from-port="2" to-layer="2921" to-port="2" /> <edge from-layer="2920" from-port="0" to-layer="2921" to-port="3" /> <edge from-layer="2921" from-port="4" to-layer="2923" to-port="0" /> <edge from-layer="2922" from-port="0" to-layer="2923" to-port="2" /> <edge from-layer="2923" from-port="3" to-layer="2924" to-port="0" /> <edge from-layer="2924" from-port="2" to-layer="2925" to-port="1" /> <edge from-layer="2924" from-port="2" to-layer="2964" to-port="1" /> <edge from-layer="2925" from-port="2" to-layer="2954" to-port="0" /> <edge from-layer="2926" from-port="0" to-layer="2936" to-port="0" /> <edge from-layer="2927" from-port="0" to-layer="2936" to-port="1" /> <edge from-layer="2928" from-port="1" to-layer="2930" to-port="0" /> <edge from-layer="2929" from-port="0" to-layer="2930" to-port="2" /> <edge from-layer="2930" from-port="3" to-layer="2931" to-port="0" /> <edge from-layer="2931" from-port="2" to-layer="2932" to-port="0" /> <edge from-layer="2932" from-port="1" to-layer="2934" to-port="0" /> <edge from-layer="2933" from-port="0" to-layer="2934" to-port="1" /> <edge from-layer="2934" from-port="2" to-layer="2936" to-port="2" /> <edge from-layer="2935" from-port="0" to-layer="2936" to-port="3" /> <edge from-layer="2936" from-port="4" to-layer="2944" to-port="2" /> <edge from-layer="2936" from-port="4" to-layer="2939" to-port="1" /> <edge from-layer="2937" from-port="0" to-layer="2939" to-port="2" /> <edge from-layer="2938" from-port="0" to-layer="2939" to-port="3" /> <edge from-layer="2939" from-port="4" to-layer="2941" to-port="0" /> <edge from-layer="2940" from-port="0" to-layer="2941" to-port="1" /> <edge from-layer="2941" from-port="2" to-layer="2945" to-port="0" /> <edge from-layer="2942" from-port="0" to-layer="2944" to-port="1" /> <edge from-layer="2943" from-port="0" to-layer="2944" to-port="3" /> <edge from-layer="2944" from-port="4" to-layer="2945" to-port="1" /> <edge from-layer="2945" from-port="2" to-layer="2953" to-port="0" /> <edge from-layer="2946" from-port="0" to-layer="2949" to-port="0" /> <edge from-layer="2947" from-port="0" to-layer="2949" to-port="1" /> <edge from-layer="2948" from-port="0" to-layer="2949" to-port="3" /> <edge from-layer="2949" from-port="4" to-layer="2951" to-port="0" /> <edge from-layer="2950" from-port="0" to-layer="2951" to-port="2" /> <edge from-layer="2951" from-port="3" to-layer="2952" to-port="0" /> <edge from-layer="2952" from-port="2" to-layer="2953" to-port="1" /> <edge from-layer="2952" from-port="2" to-layer="2985" to-port="1" /> <edge from-layer="2953" from-port="2" to-layer="2954" to-port="1" /> <edge from-layer="2954" from-port="2" to-layer="2959" to-port="0" /> <edge from-layer="2955" from-port="0" to-layer="2958" to-port="1" /> <edge from-layer="2956" from-port="0" to-layer="2958" to-port="2" /> <edge from-layer="2957" from-port="0" to-layer="2958" to-port="3" /> <edge from-layer="2958" from-port="4" to-layer="2959" to-port="1" /> <edge from-layer="2959" from-port="2" to-layer="3015" to-port="0" /> <edge from-layer="2959" from-port="2" to-layer="3053" to-port="0" /> <edge from-layer="2960" from-port="0" to-layer="2963" to-port="1" /> <edge from-layer="2961" from-port="0" to-layer="2963" to-port="2" /> <edge from-layer="2962" from-port="0" to-layer="2963" to-port="3" /> <edge from-layer="2963" from-port="4" to-layer="2964" to-port="0" /> <edge from-layer="2963" from-port="4" to-layer="2967" to-port="0" /> <edge from-layer="2963" from-port="4" to-layer="2983" to-port="0" /> <edge from-layer="2963" from-port="4" to-layer="2978" to-port="0" /> <edge from-layer="2964" from-port="2" to-layer="2986" to-port="0" /> <edge from-layer="2965" from-port="0" to-layer="2975" to-port="0" /> <edge from-layer="2966" from-port="0" to-layer="2975" to-port="1" /> <edge from-layer="2967" from-port="1" to-layer="2969" to-port="0" /> <edge from-layer="2968" from-port="0" to-layer="2969" to-port="2" /> <edge from-layer="2969" from-port="3" to-layer="2970" to-port="0" /> <edge from-layer="2970" from-port="2" to-layer="2971" to-port="0" /> <edge from-layer="2971" from-port="1" to-layer="2973" to-port="0" /> <edge from-layer="2972" from-port="0" to-layer="2973" to-port="1" /> <edge from-layer="2973" from-port="2" to-layer="2975" to-port="2" /> <edge from-layer="2974" from-port="0" to-layer="2975" to-port="3" /> <edge from-layer="2975" from-port="4" to-layer="2978" to-port="1" /> <edge from-layer="2975" from-port="4" to-layer="2983" to-port="2" /> <edge from-layer="2976" from-port="0" to-layer="2978" to-port="2" /> <edge from-layer="2977" from-port="0" to-layer="2978" to-port="3" /> <edge from-layer="2978" from-port="4" to-layer="2980" to-port="0" /> <edge from-layer="2979" from-port="0" to-layer="2980" to-port="1" /> <edge from-layer="2980" from-port="2" to-layer="2984" to-port="0" /> <edge from-layer="2981" from-port="0" to-layer="2983" to-port="1" /> <edge from-layer="2982" from-port="0" to-layer="2983" to-port="3" /> <edge from-layer="2983" from-port="4" to-layer="2984" to-port="1" /> <edge from-layer="2984" from-port="2" to-layer="2985" to-port="0" /> <edge from-layer="2985" from-port="2" to-layer="2986" to-port="1" /> <edge from-layer="2986" from-port="2" to-layer="2991" to-port="0" /> <edge from-layer="2987" from-port="0" to-layer="2990" to-port="1" /> <edge from-layer="2988" from-port="0" to-layer="2990" to-port="2" /> <edge from-layer="2989" from-port="0" to-layer="2990" to-port="3" /> <edge from-layer="2990" from-port="4" to-layer="2991" to-port="1" /> <edge from-layer="2991" from-port="2" to-layer="2992" to-port="1" /> <edge from-layer="2992" from-port="2" to-layer="3012" to-port="0" /> <edge from-layer="2992" from-port="2" to-layer="3053" to-port="1" /> <edge from-layer="2992" from-port="2" to-layer="10165" to-port="0" /> <edge from-layer="2993" from-port="0" to-layer="2998" to-port="0" /> <edge from-layer="2994" from-port="0" to-layer="2997" to-port="1" /> <edge from-layer="2995" from-port="0" to-layer="2997" to-port="2" /> <edge from-layer="2996" from-port="0" to-layer="2997" to-port="3" /> <edge from-layer="2997" from-port="4" to-layer="2998" to-port="1" /> <edge from-layer="2998" from-port="2" to-layer="2999" to-port="0" /> <edge from-layer="2999" from-port="1" to-layer="3001" to-port="0" /> <edge from-layer="3000" from-port="0" to-layer="3001" to-port="2" /> <edge from-layer="3001" from-port="3" to-layer="3008" to-port="0" /> <edge from-layer="3002" from-port="0" to-layer="3005" to-port="1" /> <edge from-layer="3003" from-port="0" to-layer="3005" to-port="2" /> <edge from-layer="3004" from-port="0" to-layer="3005" to-port="3" /> <edge from-layer="3005" from-port="4" to-layer="3007" to-port="0" /> <edge from-layer="3006" from-port="0" to-layer="3007" to-port="1" /> <edge from-layer="3007" from-port="2" to-layer="3008" to-port="1" /> <edge from-layer="3008" from-port="2" to-layer="3053" to-port="2" /> <edge from-layer="3008" from-port="2" to-layer="10164" to-port="0" /> <edge from-layer="3009" from-port="0" to-layer="3027" to-port="0" /> <edge from-layer="3010" from-port="0" to-layer="3022" to-port="0" /> <edge from-layer="3011" from-port="0" to-layer="3022" to-port="1" /> <edge from-layer="3011" from-port="0" to-layer="3025" to-port="1" /> <edge from-layer="3012" from-port="1" to-layer="3014" to-port="0" /> <edge from-layer="3013" from-port="0" to-layer="3014" to-port="2" /> <edge from-layer="3014" from-port="3" to-layer="3018" to-port="0" /> <edge from-layer="3014" from-port="3" to-layer="3024" to-port="0" /> <edge from-layer="3015" from-port="1" to-layer="3017" to-port="0" /> <edge from-layer="3015" from-port="1" to-layer="3042" to-port="0" /> <edge from-layer="3016" from-port="0" to-layer="3017" to-port="2" /> <edge from-layer="3017" from-port="3" to-layer="3018" to-port="1" /> <edge from-layer="3018" from-port="2" to-layer="3020" to-port="0" /> <edge from-layer="3019" from-port="0" to-layer="3020" to-port="1" /> <edge from-layer="3019" from-port="0" to-layer="3024" to-port="1" /> <edge from-layer="3020" from-port="2" to-layer="3022" to-port="2" /> <edge from-layer="3021" from-port="0" to-layer="3025" to-port="3" /> <edge from-layer="3021" from-port="0" to-layer="3022" to-port="3" /> <edge from-layer="3022" from-port="4" to-layer="3027" to-port="1" /> <edge from-layer="3023" from-port="0" to-layer="3025" to-port="0" /> <edge from-layer="3024" from-port="2" to-layer="3032" to-port="2" /> <edge from-layer="3024" from-port="2" to-layer="3025" to-port="2" /> <edge from-layer="3025" from-port="4" to-layer="3027" to-port="2" /> <edge from-layer="3026" from-port="0" to-layer="3027" to-port="3" /> <edge from-layer="3027" from-port="4" to-layer="3034" to-port="0" /> <edge from-layer="3028" from-port="0" to-layer="3034" to-port="1" /> <edge from-layer="3029" from-port="0" to-layer="3032" to-port="0" /> <edge from-layer="3030" from-port="0" to-layer="3032" to-port="1" /> <edge from-layer="3031" from-port="0" to-layer="3032" to-port="3" /> <edge from-layer="3032" from-port="4" to-layer="3034" to-port="2" /> <edge from-layer="3033" from-port="0" to-layer="3034" to-port="3" /> <edge from-layer="3034" from-port="4" to-layer="3037" to-port="0" /> <edge from-layer="3035" from-port="0" to-layer="3037" to-port="1" /> <edge from-layer="3036" from-port="0" to-layer="3037" to-port="2" /> <edge from-layer="3037" from-port="3" to-layer="3038" to-port="0" /> <edge from-layer="3038" from-port="2" to-layer="3039" to-port="0" /> <edge from-layer="3038" from-port="2" to-layer="3052" to-port="0" /> <edge from-layer="3039" from-port="1" to-layer="3051" to-port="0" /> <edge from-layer="3040" from-port="0" to-layer="3042" to-port="1" /> <edge from-layer="3041" from-port="0" to-layer="3042" to-port="2" /> <edge from-layer="3042" from-port="3" to-layer="3046" to-port="0" /> <edge from-layer="3043" from-port="0" to-layer="3046" to-port="1" /> <edge from-layer="3044" from-port="0" to-layer="3046" to-port="2" /> <edge from-layer="3045" from-port="0" to-layer="3046" to-port="3" /> <edge from-layer="3046" from-port="4" to-layer="3048" to-port="0" /> <edge from-layer="3046" from-port="4" to-layer="3050" to-port="2" /> <edge from-layer="3047" from-port="0" to-layer="3048" to-port="1" /> <edge from-layer="3048" from-port="2" to-layer="3050" to-port="0" /> <edge from-layer="3049" from-port="0" to-layer="3050" to-port="1" /> <edge from-layer="3050" from-port="3" to-layer="3051" to-port="1" /> <edge from-layer="3051" from-port="2" to-layer="3052" to-port="1" /> <edge from-layer="3052" from-port="2" to-layer="3053" to-port="3" /> <edge from-layer="3053" from-port="4" to-layer="3055" to-port="0" /> <edge from-layer="3054" from-port="0" to-layer="3055" to-port="1" /> <edge from-layer="3055" from-port="2" to-layer="3057" to-port="0" /> <edge from-layer="3056" from-port="0" to-layer="3057" to-port="1" /> <edge from-layer="3057" from-port="2" to-layer="3065" to-port="0" /> <edge from-layer="3058" from-port="0" to-layer="3059" to-port="0" /> <edge from-layer="3059" from-port="1" to-layer="3062" to-port="0" /> <edge from-layer="3060" from-port="0" to-layer="3061" to-port="0" /> <edge from-layer="3061" from-port="1" to-layer="3062" to-port="1" /> <edge from-layer="3062" from-port="2" to-layer="3064" to-port="0" /> <edge from-layer="3063" from-port="0" to-layer="3064" to-port="1" /> <edge from-layer="3064" from-port="2" to-layer="3065" to-port="1" /> <edge from-layer="3065" from-port="2" to-layer="3067" to-port="0" /> <edge from-layer="3066" from-port="0" to-layer="3067" to-port="1" /> <edge from-layer="3067" from-port="2" to-layer="3068" to-port="1" /> <edge from-layer="3068" from-port="2" to-layer="3069" to-port="0" /> <edge from-layer="3069" from-port="2" to-layer="3071" to-port="0" /> <edge from-layer="3069" from-port="2" to-layer="3300" to-port="1" /> <edge from-layer="3069" from-port="2" to-layer="3098" to-port="0" /> <edge from-layer="3070" from-port="0" to-layer="3071" to-port="1" /> <edge from-layer="3071" from-port="2" to-layer="3073" to-port="0" /> <edge from-layer="3072" from-port="0" to-layer="3073" to-port="1" /> <edge from-layer="3073" from-port="2" to-layer="3075" to-port="0" /> <edge from-layer="3074" from-port="0" to-layer="3075" to-port="1" /> <edge from-layer="3075" from-port="2" to-layer="3083" to-port="0" /> <edge from-layer="3076" from-port="0" to-layer="3077" to-port="0" /> <edge from-layer="3077" from-port="1" to-layer="3080" to-port="0" /> <edge from-layer="3078" from-port="0" to-layer="3079" to-port="0" /> <edge from-layer="3079" from-port="1" to-layer="3080" to-port="1" /> <edge from-layer="3080" from-port="2" to-layer="3082" to-port="0" /> <edge from-layer="3081" from-port="0" to-layer="3082" to-port="1" /> <edge from-layer="3082" from-port="2" to-layer="3083" to-port="1" /> <edge from-layer="3083" from-port="2" to-layer="3085" to-port="0" /> <edge from-layer="3084" from-port="0" to-layer="3085" to-port="1" /> <edge from-layer="3085" from-port="2" to-layer="3086" to-port="0" /> <edge from-layer="3086" from-port="1" to-layer="3094" to-port="0" /> <edge from-layer="3087" from-port="0" to-layer="3088" to-port="0" /> <edge from-layer="3088" from-port="1" to-layer="3091" to-port="0" /> <edge from-layer="3089" from-port="0" to-layer="3090" to-port="0" /> <edge from-layer="3090" from-port="1" to-layer="3091" to-port="1" /> <edge from-layer="3091" from-port="2" to-layer="3093" to-port="0" /> <edge from-layer="3092" from-port="0" to-layer="3093" to-port="1" /> <edge from-layer="3093" from-port="2" to-layer="3094" to-port="1" /> <edge from-layer="3094" from-port="2" to-layer="3096" to-port="0" /> <edge from-layer="3095" from-port="0" to-layer="3096" to-port="1" /> <edge from-layer="3096" from-port="2" to-layer="3299" to-port="0" /> <edge from-layer="3097" from-port="0" to-layer="3098" to-port="1" /> <edge from-layer="3098" from-port="2" to-layer="3100" to-port="0" /> <edge from-layer="3099" from-port="0" to-layer="3100" to-port="1" /> <edge from-layer="3100" from-port="2" to-layer="3102" to-port="0" /> <edge from-layer="3101" from-port="0" to-layer="3102" to-port="1" /> <edge from-layer="3102" from-port="2" to-layer="3110" to-port="0" /> <edge from-layer="3103" from-port="0" to-layer="3104" to-port="0" /> <edge from-layer="3104" from-port="1" to-layer="3107" to-port="0" /> <edge from-layer="3105" from-port="0" to-layer="3106" to-port="0" /> <edge from-layer="3106" from-port="1" to-layer="3107" to-port="1" /> <edge from-layer="3107" from-port="2" to-layer="3109" to-port="0" /> <edge from-layer="3108" from-port="0" to-layer="3109" to-port="1" /> <edge from-layer="3109" from-port="2" to-layer="3110" to-port="1" /> <edge from-layer="3110" from-port="2" to-layer="3112" to-port="0" /> <edge from-layer="3111" from-port="0" to-layer="3112" to-port="1" /> <edge from-layer="3112" from-port="2" to-layer="3114" to-port="0" /> <edge from-layer="3113" from-port="0" to-layer="3114" to-port="1" /> <edge from-layer="3114" from-port="2" to-layer="3130" to-port="0" /> <edge from-layer="3114" from-port="2" to-layer="3236" to-port="0" /> <edge from-layer="3114" from-port="2" to-layer="3118" to-port="0" /> <edge from-layer="3115" from-port="0" to-layer="3118" to-port="1" /> <edge from-layer="3116" from-port="0" to-layer="3118" to-port="2" /> <edge from-layer="3117" from-port="0" to-layer="3118" to-port="3" /> <edge from-layer="3118" from-port="4" to-layer="3120" to-port="0" /> <edge from-layer="3119" from-port="0" to-layer="3120" to-port="1" /> <edge from-layer="3120" from-port="2" to-layer="3189" to-port="0" /> <edge from-layer="3120" from-port="2" to-layer="3124" to-port="0" /> <edge from-layer="3121" from-port="0" to-layer="3124" to-port="1" /> <edge from-layer="3122" from-port="0" to-layer="3124" to-port="2" /> <edge from-layer="3123" from-port="0" to-layer="3124" to-port="3" /> <edge from-layer="3124" from-port="4" to-layer="3170" to-port="0" /> <edge from-layer="3124" from-port="4" to-layer="3175" to-port="0" /> <edge from-layer="3124" from-port="4" to-layer="3159" to-port="0" /> <edge from-layer="3124" from-port="4" to-layer="3156" to-port="0" /> <edge from-layer="3125" from-port="0" to-layer="3152" to-port="0" /> <edge from-layer="3126" from-port="0" to-layer="3152" to-port="1" /> <edge from-layer="3127" from-port="0" to-layer="3130" to-port="1" /> <edge from-layer="3128" from-port="0" to-layer="3130" to-port="2" /> <edge from-layer="3129" from-port="0" to-layer="3130" to-port="3" /> <edge from-layer="3130" from-port="4" to-layer="3132" to-port="0" /> <edge from-layer="3131" from-port="0" to-layer="3132" to-port="1" /> <edge from-layer="3132" from-port="2" to-layer="3221" to-port="0" /> <edge from-layer="3132" from-port="2" to-layer="3133" to-port="0" /> <edge from-layer="3132" from-port="2" to-layer="3194" to-port="0" /> <edge from-layer="3133" from-port="1" to-layer="3135" to-port="0" /> <edge from-layer="3134" from-port="0" to-layer="3135" to-port="2" /> <edge from-layer="3135" from-port="3" to-layer="3148" to-port="0" /> <edge from-layer="3136" from-port="0" to-layer="3141" to-port="0" /> <edge from-layer="3137" from-port="0" to-layer="3140" to-port="1" /> <edge from-layer="3138" from-port="0" to-layer="3140" to-port="2" /> <edge from-layer="3139" from-port="0" to-layer="3140" to-port="3" /> <edge from-layer="3140" from-port="4" to-layer="3141" to-port="1" /> <edge from-layer="3141" from-port="2" to-layer="3142" to-port="0" /> <edge from-layer="3142" from-port="1" to-layer="3144" to-port="0" /> <edge from-layer="3143" from-port="0" to-layer="3144" to-port="2" /> <edge from-layer="3144" from-port="3" to-layer="3145" to-port="0" /> <edge from-layer="3144" from-port="3" to-layer="3223" to-port="0" /> <edge from-layer="3145" from-port="1" to-layer="3147" to-port="0" /> <edge from-layer="3146" from-port="0" to-layer="3147" to-port="2" /> <edge from-layer="3147" from-port="3" to-layer="3148" to-port="1" /> <edge from-layer="3148" from-port="2" to-layer="3150" to-port="0" /> <edge from-layer="3149" from-port="0" to-layer="3150" to-port="1" /> <edge from-layer="3150" from-port="2" to-layer="3152" to-port="2" /> <edge from-layer="3150" from-port="2" to-layer="3180" to-port="2" /> <edge from-layer="3151" from-port="0" to-layer="3152" to-port="3" /> <edge from-layer="3152" from-port="4" to-layer="3154" to-port="0" /> <edge from-layer="3153" from-port="0" to-layer="3154" to-port="2" /> <edge from-layer="3154" from-port="3" to-layer="3155" to-port="0" /> <edge from-layer="3155" from-port="2" to-layer="3156" to-port="1" /> <edge from-layer="3155" from-port="2" to-layer="3195" to-port="1" /> <edge from-layer="3156" from-port="2" to-layer="3185" to-port="0" /> <edge from-layer="3157" from-port="0" to-layer="3167" to-port="0" /> <edge from-layer="3158" from-port="0" to-layer="3167" to-port="1" /> <edge from-layer="3159" from-port="1" to-layer="3161" to-port="0" /> <edge from-layer="3160" from-port="0" to-layer="3161" to-port="2" /> <edge from-layer="3161" from-port="3" to-layer="3162" to-port="0" /> <edge from-layer="3162" from-port="2" to-layer="3163" to-port="0" /> <edge from-layer="3163" from-port="1" to-layer="3165" to-port="0" /> <edge from-layer="3164" from-port="0" to-layer="3165" to-port="1" /> <edge from-layer="3165" from-port="2" to-layer="3167" to-port="2" /> <edge from-layer="3166" from-port="0" to-layer="3167" to-port="3" /> <edge from-layer="3167" from-port="4" to-layer="3175" to-port="2" /> <edge from-layer="3167" from-port="4" to-layer="3170" to-port="1" /> <edge from-layer="3168" from-port="0" to-layer="3170" to-port="2" /> <edge from-layer="3169" from-port="0" to-layer="3170" to-port="3" /> <edge from-layer="3170" from-port="4" to-layer="3172" to-port="0" /> <edge from-layer="3171" from-port="0" to-layer="3172" to-port="1" /> <edge from-layer="3172" from-port="2" to-layer="3176" to-port="0" /> <edge from-layer="3173" from-port="0" to-layer="3175" to-port="1" /> <edge from-layer="3174" from-port="0" to-layer="3175" to-port="3" /> <edge from-layer="3175" from-port="4" to-layer="3176" to-port="1" /> <edge from-layer="3176" from-port="2" to-layer="3184" to-port="0" /> <edge from-layer="3177" from-port="0" to-layer="3180" to-port="0" /> <edge from-layer="3178" from-port="0" to-layer="3180" to-port="1" /> <edge from-layer="3179" from-port="0" to-layer="3180" to-port="3" /> <edge from-layer="3180" from-port="4" to-layer="3182" to-port="0" /> <edge from-layer="3181" from-port="0" to-layer="3182" to-port="2" /> <edge from-layer="3182" from-port="3" to-layer="3183" to-port="0" /> <edge from-layer="3183" from-port="2" to-layer="3216" to-port="1" /> <edge from-layer="3183" from-port="2" to-layer="3184" to-port="1" /> <edge from-layer="3184" from-port="2" to-layer="3185" to-port="1" /> <edge from-layer="3185" from-port="2" to-layer="3190" to-port="0" /> <edge from-layer="3186" from-port="0" to-layer="3189" to-port="1" /> <edge from-layer="3187" from-port="0" to-layer="3189" to-port="2" /> <edge from-layer="3188" from-port="0" to-layer="3189" to-port="3" /> <edge from-layer="3189" from-port="4" to-layer="3190" to-port="1" /> <edge from-layer="3190" from-port="2" to-layer="3284" to-port="0" /> <edge from-layer="3190" from-port="2" to-layer="3246" to-port="0" /> <edge from-layer="3191" from-port="0" to-layer="3194" to-port="1" /> <edge from-layer="3192" from-port="0" to-layer="3194" to-port="2" /> <edge from-layer="3193" from-port="0" to-layer="3194" to-port="3" /> <edge from-layer="3194" from-port="4" to-layer="3195" to-port="0" /> <edge from-layer="3194" from-port="4" to-layer="3198" to-port="0" /> <edge from-layer="3194" from-port="4" to-layer="3214" to-port="0" /> <edge from-layer="3194" from-port="4" to-layer="3209" to-port="0" /> <edge from-layer="3195" from-port="2" to-layer="3217" to-port="0" /> <edge from-layer="3196" from-port="0" to-layer="3206" to-port="0" /> <edge from-layer="3197" from-port="0" to-layer="3206" to-port="1" /> <edge from-layer="3198" from-port="1" to-layer="3200" to-port="0" /> <edge from-layer="3199" from-port="0" to-layer="3200" to-port="2" /> <edge from-layer="3200" from-port="3" to-layer="3201" to-port="0" /> <edge from-layer="3201" from-port="2" to-layer="3202" to-port="0" /> <edge from-layer="3202" from-port="1" to-layer="3204" to-port="0" /> <edge from-layer="3203" from-port="0" to-layer="3204" to-port="1" /> <edge from-layer="3204" from-port="2" to-layer="3206" to-port="2" /> <edge from-layer="3205" from-port="0" to-layer="3206" to-port="3" /> <edge from-layer="3206" from-port="4" to-layer="3214" to-port="2" /> <edge from-layer="3206" from-port="4" to-layer="3209" to-port="1" /> <edge from-layer="3207" from-port="0" to-layer="3209" to-port="2" /> <edge from-layer="3208" from-port="0" to-layer="3209" to-port="3" /> <edge from-layer="3209" from-port="4" to-layer="3211" to-port="0" /> <edge from-layer="3210" from-port="0" to-layer="3211" to-port="1" /> <edge from-layer="3211" from-port="2" to-layer="3215" to-port="0" /> <edge from-layer="3212" from-port="0" to-layer="3214" to-port="1" /> <edge from-layer="3213" from-port="0" to-layer="3214" to-port="3" /> <edge from-layer="3214" from-port="4" to-layer="3215" to-port="1" /> <edge from-layer="3215" from-port="2" to-layer="3216" to-port="0" /> <edge from-layer="3216" from-port="2" to-layer="3217" to-port="1" /> <edge from-layer="3217" from-port="2" to-layer="3222" to-port="0" /> <edge from-layer="3218" from-port="0" to-layer="3221" to-port="1" /> <edge from-layer="3219" from-port="0" to-layer="3221" to-port="2" /> <edge from-layer="3220" from-port="0" to-layer="3221" to-port="3" /> <edge from-layer="3221" from-port="4" to-layer="3222" to-port="1" /> <edge from-layer="3222" from-port="2" to-layer="3223" to-port="1" /> <edge from-layer="3223" from-port="2" to-layer="3284" to-port="1" /> <edge from-layer="3223" from-port="2" to-layer="10163" to-port="0" /> <edge from-layer="3223" from-port="2" to-layer="3243" to-port="0" /> <edge from-layer="3224" from-port="0" to-layer="3229" to-port="0" /> <edge from-layer="3225" from-port="0" to-layer="3228" to-port="1" /> <edge from-layer="3226" from-port="0" to-layer="3228" to-port="2" /> <edge from-layer="3227" from-port="0" to-layer="3228" to-port="3" /> <edge from-layer="3228" from-port="4" to-layer="3229" to-port="1" /> <edge from-layer="3229" from-port="2" to-layer="3230" to-port="0" /> <edge from-layer="3230" from-port="1" to-layer="3232" to-port="0" /> <edge from-layer="3231" from-port="0" to-layer="3232" to-port="2" /> <edge from-layer="3232" from-port="3" to-layer="3239" to-port="0" /> <edge from-layer="3233" from-port="0" to-layer="3236" to-port="1" /> <edge from-layer="3234" from-port="0" to-layer="3236" to-port="2" /> <edge from-layer="3235" from-port="0" to-layer="3236" to-port="3" /> <edge from-layer="3236" from-port="4" to-layer="3238" to-port="0" /> <edge from-layer="3237" from-port="0" to-layer="3238" to-port="1" /> <edge from-layer="3238" from-port="2" to-layer="3239" to-port="1" /> <edge from-layer="3239" from-port="2" to-layer="10162" to-port="0" /> <edge from-layer="3239" from-port="2" to-layer="3284" to-port="2" /> <edge from-layer="3240" from-port="0" to-layer="3258" to-port="0" /> <edge from-layer="3241" from-port="0" to-layer="3253" to-port="0" /> <edge from-layer="3242" from-port="0" to-layer="3256" to-port="1" /> <edge from-layer="3242" from-port="0" to-layer="3253" to-port="1" /> <edge from-layer="3243" from-port="1" to-layer="3245" to-port="0" /> <edge from-layer="3244" from-port="0" to-layer="3245" to-port="2" /> <edge from-layer="3245" from-port="3" to-layer="3255" to-port="0" /> <edge from-layer="3245" from-port="3" to-layer="3249" to-port="0" /> <edge from-layer="3246" from-port="1" to-layer="3273" to-port="0" /> <edge from-layer="3246" from-port="1" to-layer="3248" to-port="0" /> <edge from-layer="3247" from-port="0" to-layer="3248" to-port="2" /> <edge from-layer="3248" from-port="3" to-layer="3249" to-port="1" /> <edge from-layer="3249" from-port="2" to-layer="3251" to-port="0" /> <edge from-layer="3250" from-port="0" to-layer="3255" to-port="1" /> <edge from-layer="3250" from-port="0" to-layer="3251" to-port="1" /> <edge from-layer="3251" from-port="2" to-layer="3253" to-port="2" /> <edge from-layer="3252" from-port="0" to-layer="3253" to-port="3" /> <edge from-layer="3252" from-port="0" to-layer="3256" to-port="3" /> <edge from-layer="3253" from-port="4" to-layer="3258" to-port="1" /> <edge from-layer="3254" from-port="0" to-layer="3256" to-port="0" /> <edge from-layer="3255" from-port="2" to-layer="3263" to-port="2" /> <edge from-layer="3255" from-port="2" to-layer="3256" to-port="2" /> <edge from-layer="3256" from-port="4" to-layer="3258" to-port="2" /> <edge from-layer="3257" from-port="0" to-layer="3258" to-port="3" /> <edge from-layer="3258" from-port="4" to-layer="3265" to-port="0" /> <edge from-layer="3259" from-port="0" to-layer="3265" to-port="1" /> <edge from-layer="3260" from-port="0" to-layer="3263" to-port="0" /> <edge from-layer="3261" from-port="0" to-layer="3263" to-port="1" /> <edge from-layer="3262" from-port="0" to-layer="3263" to-port="3" /> <edge from-layer="3263" from-port="4" to-layer="3265" to-port="2" /> <edge from-layer="3264" from-port="0" to-layer="3265" to-port="3" /> <edge from-layer="3265" from-port="4" to-layer="3268" to-port="0" /> <edge from-layer="3266" from-port="0" to-layer="3268" to-port="1" /> <edge from-layer="3267" from-port="0" to-layer="3268" to-port="2" /> <edge from-layer="3268" from-port="3" to-layer="3269" to-port="0" /> <edge from-layer="3269" from-port="2" to-layer="3270" to-port="0" /> <edge from-layer="3269" from-port="2" to-layer="3283" to-port="0" /> <edge from-layer="3270" from-port="1" to-layer="3282" to-port="0" /> <edge from-layer="3271" from-port="0" to-layer="3273" to-port="1" /> <edge from-layer="3272" from-port="0" to-layer="3273" to-port="2" /> <edge from-layer="3273" from-port="3" to-layer="3277" to-port="0" /> <edge from-layer="3274" from-port="0" to-layer="3277" to-port="1" /> <edge from-layer="3275" from-port="0" to-layer="3277" to-port="2" /> <edge from-layer="3276" from-port="0" to-layer="3277" to-port="3" /> <edge from-layer="3277" from-port="4" to-layer="3279" to-port="0" /> <edge from-layer="3277" from-port="4" to-layer="3281" to-port="2" /> <edge from-layer="3278" from-port="0" to-layer="3279" to-port="1" /> <edge from-layer="3279" from-port="2" to-layer="3281" to-port="0" /> <edge from-layer="3280" from-port="0" to-layer="3281" to-port="1" /> <edge from-layer="3281" from-port="3" to-layer="3282" to-port="1" /> <edge from-layer="3282" from-port="2" to-layer="3283" to-port="1" /> <edge from-layer="3283" from-port="2" to-layer="3284" to-port="3" /> <edge from-layer="3284" from-port="4" to-layer="3286" to-port="0" /> <edge from-layer="3285" from-port="0" to-layer="3286" to-port="1" /> <edge from-layer="3286" from-port="2" to-layer="3288" to-port="0" /> <edge from-layer="3287" from-port="0" to-layer="3288" to-port="1" /> <edge from-layer="3288" from-port="2" to-layer="3296" to-port="0" /> <edge from-layer="3289" from-port="0" to-layer="3290" to-port="0" /> <edge from-layer="3290" from-port="1" to-layer="3293" to-port="0" /> <edge from-layer="3291" from-port="0" to-layer="3292" to-port="0" /> <edge from-layer="3292" from-port="1" to-layer="3293" to-port="1" /> <edge from-layer="3293" from-port="2" to-layer="3295" to-port="0" /> <edge from-layer="3294" from-port="0" to-layer="3295" to-port="1" /> <edge from-layer="3295" from-port="2" to-layer="3296" to-port="1" /> <edge from-layer="3296" from-port="2" to-layer="3298" to-port="0" /> <edge from-layer="3297" from-port="0" to-layer="3298" to-port="1" /> <edge from-layer="3298" from-port="2" to-layer="3299" to-port="1" /> <edge from-layer="3299" from-port="2" to-layer="3300" to-port="0" /> <edge from-layer="3300" from-port="2" to-layer="3329" to-port="0" /> <edge from-layer="3300" from-port="2" to-layer="3302" to-port="0" /> <edge from-layer="3300" from-port="2" to-layer="3531" to-port="1" /> <edge from-layer="3301" from-port="0" to-layer="3302" to-port="1" /> <edge from-layer="3302" from-port="2" to-layer="3304" to-port="0" /> <edge from-layer="3303" from-port="0" to-layer="3304" to-port="1" /> <edge from-layer="3304" from-port="2" to-layer="3306" to-port="0" /> <edge from-layer="3305" from-port="0" to-layer="3306" to-port="1" /> <edge from-layer="3306" from-port="2" to-layer="3314" to-port="0" /> <edge from-layer="3307" from-port="0" to-layer="3308" to-port="0" /> <edge from-layer="3308" from-port="1" to-layer="3311" to-port="0" /> <edge from-layer="3309" from-port="0" to-layer="3310" to-port="0" /> <edge from-layer="3310" from-port="1" to-layer="3311" to-port="1" /> <edge from-layer="3311" from-port="2" to-layer="3313" to-port="0" /> <edge from-layer="3312" from-port="0" to-layer="3313" to-port="1" /> <edge from-layer="3313" from-port="2" to-layer="3314" to-port="1" /> <edge from-layer="3314" from-port="2" to-layer="3316" to-port="0" /> <edge from-layer="3315" from-port="0" to-layer="3316" to-port="1" /> <edge from-layer="3316" from-port="2" to-layer="3317" to-port="0" /> <edge from-layer="3317" from-port="1" to-layer="3325" to-port="0" /> <edge from-layer="3318" from-port="0" to-layer="3319" to-port="0" /> <edge from-layer="3319" from-port="1" to-layer="3322" to-port="0" /> <edge from-layer="3320" from-port="0" to-layer="3321" to-port="0" /> <edge from-layer="3321" from-port="1" to-layer="3322" to-port="1" /> <edge from-layer="3322" from-port="2" to-layer="3324" to-port="0" /> <edge from-layer="3323" from-port="0" to-layer="3324" to-port="1" /> <edge from-layer="3324" from-port="2" to-layer="3325" to-port="1" /> <edge from-layer="3325" from-port="2" to-layer="3327" to-port="0" /> <edge from-layer="3326" from-port="0" to-layer="3327" to-port="1" /> <edge from-layer="3327" from-port="2" to-layer="3530" to-port="0" /> <edge from-layer="3328" from-port="0" to-layer="3329" to-port="1" /> <edge from-layer="3329" from-port="2" to-layer="3331" to-port="0" /> <edge from-layer="3330" from-port="0" to-layer="3331" to-port="1" /> <edge from-layer="3331" from-port="2" to-layer="3333" to-port="0" /> <edge from-layer="3332" from-port="0" to-layer="3333" to-port="1" /> <edge from-layer="3333" from-port="2" to-layer="3341" to-port="0" /> <edge from-layer="3334" from-port="0" to-layer="3335" to-port="0" /> <edge from-layer="3335" from-port="1" to-layer="3338" to-port="0" /> <edge from-layer="3336" from-port="0" to-layer="3337" to-port="0" /> <edge from-layer="3337" from-port="1" to-layer="3338" to-port="1" /> <edge from-layer="3338" from-port="2" to-layer="3340" to-port="0" /> <edge from-layer="3339" from-port="0" to-layer="3340" to-port="1" /> <edge from-layer="3340" from-port="2" to-layer="3341" to-port="1" /> <edge from-layer="3341" from-port="2" to-layer="3343" to-port="0" /> <edge from-layer="3342" from-port="0" to-layer="3343" to-port="1" /> <edge from-layer="3343" from-port="2" to-layer="3345" to-port="0" /> <edge from-layer="3344" from-port="0" to-layer="3345" to-port="1" /> <edge from-layer="3345" from-port="2" to-layer="3467" to-port="0" /> <edge from-layer="3345" from-port="2" to-layer="3361" to-port="0" /> <edge from-layer="3345" from-port="2" to-layer="3349" to-port="0" /> <edge from-layer="3346" from-port="0" to-layer="3349" to-port="1" /> <edge from-layer="3347" from-port="0" to-layer="3349" to-port="2" /> <edge from-layer="3348" from-port="0" to-layer="3349" to-port="3" /> <edge from-layer="3349" from-port="4" to-layer="3351" to-port="0" /> <edge from-layer="3350" from-port="0" to-layer="3351" to-port="1" /> <edge from-layer="3351" from-port="2" to-layer="3355" to-port="0" /> <edge from-layer="3351" from-port="2" to-layer="3420" to-port="0" /> <edge from-layer="3352" from-port="0" to-layer="3355" to-port="1" /> <edge from-layer="3353" from-port="0" to-layer="3355" to-port="2" /> <edge from-layer="3354" from-port="0" to-layer="3355" to-port="3" /> <edge from-layer="3355" from-port="4" to-layer="3406" to-port="0" /> <edge from-layer="3355" from-port="4" to-layer="3390" to-port="0" /> <edge from-layer="3355" from-port="4" to-layer="3387" to-port="0" /> <edge from-layer="3355" from-port="4" to-layer="3401" to-port="0" /> <edge from-layer="3356" from-port="0" to-layer="3383" to-port="0" /> <edge from-layer="3357" from-port="0" to-layer="3383" to-port="1" /> <edge from-layer="3358" from-port="0" to-layer="3361" to-port="1" /> <edge from-layer="3359" from-port="0" to-layer="3361" to-port="2" /> <edge from-layer="3360" from-port="0" to-layer="3361" to-port="3" /> <edge from-layer="3361" from-port="4" to-layer="3363" to-port="0" /> <edge from-layer="3362" from-port="0" to-layer="3363" to-port="1" /> <edge from-layer="3363" from-port="2" to-layer="3364" to-port="0" /> <edge from-layer="3363" from-port="2" to-layer="3425" to-port="0" /> <edge from-layer="3363" from-port="2" to-layer="3452" to-port="0" /> <edge from-layer="3364" from-port="1" to-layer="3366" to-port="0" /> <edge from-layer="3365" from-port="0" to-layer="3366" to-port="2" /> <edge from-layer="3366" from-port="3" to-layer="3379" to-port="0" /> <edge from-layer="3367" from-port="0" to-layer="3372" to-port="0" /> <edge from-layer="3368" from-port="0" to-layer="3371" to-port="1" /> <edge from-layer="3369" from-port="0" to-layer="3371" to-port="2" /> <edge from-layer="3370" from-port="0" to-layer="3371" to-port="3" /> <edge from-layer="3371" from-port="4" to-layer="3372" to-port="1" /> <edge from-layer="3372" from-port="2" to-layer="3373" to-port="0" /> <edge from-layer="3373" from-port="1" to-layer="3375" to-port="0" /> <edge from-layer="3374" from-port="0" to-layer="3375" to-port="2" /> <edge from-layer="3375" from-port="3" to-layer="3454" to-port="0" /> <edge from-layer="3375" from-port="3" to-layer="3376" to-port="0" /> <edge from-layer="3376" from-port="1" to-layer="3378" to-port="0" /> <edge from-layer="3377" from-port="0" to-layer="3378" to-port="2" /> <edge from-layer="3378" from-port="3" to-layer="3379" to-port="1" /> <edge from-layer="3379" from-port="2" to-layer="3381" to-port="0" /> <edge from-layer="3380" from-port="0" to-layer="3381" to-port="1" /> <edge from-layer="3381" from-port="2" to-layer="3383" to-port="2" /> <edge from-layer="3381" from-port="2" to-layer="3411" to-port="2" /> <edge from-layer="3382" from-port="0" to-layer="3383" to-port="3" /> <edge from-layer="3383" from-port="4" to-layer="3385" to-port="0" /> <edge from-layer="3384" from-port="0" to-layer="3385" to-port="2" /> <edge from-layer="3385" from-port="3" to-layer="3386" to-port="0" /> <edge from-layer="3386" from-port="2" to-layer="3387" to-port="1" /> <edge from-layer="3386" from-port="2" to-layer="3426" to-port="1" /> <edge from-layer="3387" from-port="2" to-layer="3416" to-port="0" /> <edge from-layer="3388" from-port="0" to-layer="3398" to-port="0" /> <edge from-layer="3389" from-port="0" to-layer="3398" to-port="1" /> <edge from-layer="3390" from-port="1" to-layer="3392" to-port="0" /> <edge from-layer="3391" from-port="0" to-layer="3392" to-port="2" /> <edge from-layer="3392" from-port="3" to-layer="3393" to-port="0" /> <edge from-layer="3393" from-port="2" to-layer="3394" to-port="0" /> <edge from-layer="3394" from-port="1" to-layer="3396" to-port="0" /> <edge from-layer="3395" from-port="0" to-layer="3396" to-port="1" /> <edge from-layer="3396" from-port="2" to-layer="3398" to-port="2" /> <edge from-layer="3397" from-port="0" to-layer="3398" to-port="3" /> <edge from-layer="3398" from-port="4" to-layer="3401" to-port="1" /> <edge from-layer="3398" from-port="4" to-layer="3406" to-port="2" /> <edge from-layer="3399" from-port="0" to-layer="3401" to-port="2" /> <edge from-layer="3400" from-port="0" to-layer="3401" to-port="3" /> <edge from-layer="3401" from-port="4" to-layer="3403" to-port="0" /> <edge from-layer="3402" from-port="0" to-layer="3403" to-port="1" /> <edge from-layer="3403" from-port="2" to-layer="3407" to-port="0" /> <edge from-layer="3404" from-port="0" to-layer="3406" to-port="1" /> <edge from-layer="3405" from-port="0" to-layer="3406" to-port="3" /> <edge from-layer="3406" from-port="4" to-layer="3407" to-port="1" /> <edge from-layer="3407" from-port="2" to-layer="3415" to-port="0" /> <edge from-layer="3408" from-port="0" to-layer="3411" to-port="0" /> <edge from-layer="3409" from-port="0" to-layer="3411" to-port="1" /> <edge from-layer="3410" from-port="0" to-layer="3411" to-port="3" /> <edge from-layer="3411" from-port="4" to-layer="3413" to-port="0" /> <edge from-layer="3412" from-port="0" to-layer="3413" to-port="2" /> <edge from-layer="3413" from-port="3" to-layer="3414" to-port="0" /> <edge from-layer="3414" from-port="2" to-layer="3447" to-port="1" /> <edge from-layer="3414" from-port="2" to-layer="3415" to-port="1" /> <edge from-layer="3415" from-port="2" to-layer="3416" to-port="1" /> <edge from-layer="3416" from-port="2" to-layer="3421" to-port="0" /> <edge from-layer="3417" from-port="0" to-layer="3420" to-port="1" /> <edge from-layer="3418" from-port="0" to-layer="3420" to-port="2" /> <edge from-layer="3419" from-port="0" to-layer="3420" to-port="3" /> <edge from-layer="3420" from-port="4" to-layer="3421" to-port="1" /> <edge from-layer="3421" from-port="2" to-layer="3477" to-port="0" /> <edge from-layer="3421" from-port="2" to-layer="3515" to-port="0" /> <edge from-layer="3422" from-port="0" to-layer="3425" to-port="1" /> <edge from-layer="3423" from-port="0" to-layer="3425" to-port="2" /> <edge from-layer="3424" from-port="0" to-layer="3425" to-port="3" /> <edge from-layer="3425" from-port="4" to-layer="3426" to-port="0" /> <edge from-layer="3425" from-port="4" to-layer="3445" to-port="0" /> <edge from-layer="3425" from-port="4" to-layer="3440" to-port="0" /> <edge from-layer="3425" from-port="4" to-layer="3429" to-port="0" /> <edge from-layer="3426" from-port="2" to-layer="3448" to-port="0" /> <edge from-layer="3427" from-port="0" to-layer="3437" to-port="0" /> <edge from-layer="3428" from-port="0" to-layer="3437" to-port="1" /> <edge from-layer="3429" from-port="1" to-layer="3431" to-port="0" /> <edge from-layer="3430" from-port="0" to-layer="3431" to-port="2" /> <edge from-layer="3431" from-port="3" to-layer="3432" to-port="0" /> <edge from-layer="3432" from-port="2" to-layer="3433" to-port="0" /> <edge from-layer="3433" from-port="1" to-layer="3435" to-port="0" /> <edge from-layer="3434" from-port="0" to-layer="3435" to-port="1" /> <edge from-layer="3435" from-port="2" to-layer="3437" to-port="2" /> <edge from-layer="3436" from-port="0" to-layer="3437" to-port="3" /> <edge from-layer="3437" from-port="4" to-layer="3440" to-port="1" /> <edge from-layer="3437" from-port="4" to-layer="3445" to-port="2" /> <edge from-layer="3438" from-port="0" to-layer="3440" to-port="2" /> <edge from-layer="3439" from-port="0" to-layer="3440" to-port="3" /> <edge from-layer="3440" from-port="4" to-layer="3442" to-port="0" /> <edge from-layer="3441" from-port="0" to-layer="3442" to-port="1" /> <edge from-layer="3442" from-port="2" to-layer="3446" to-port="0" /> <edge from-layer="3443" from-port="0" to-layer="3445" to-port="1" /> <edge from-layer="3444" from-port="0" to-layer="3445" to-port="3" /> <edge from-layer="3445" from-port="4" to-layer="3446" to-port="1" /> <edge from-layer="3446" from-port="2" to-layer="3447" to-port="0" /> <edge from-layer="3447" from-port="2" to-layer="3448" to-port="1" /> <edge from-layer="3448" from-port="2" to-layer="3453" to-port="0" /> <edge from-layer="3449" from-port="0" to-layer="3452" to-port="1" /> <edge from-layer="3450" from-port="0" to-layer="3452" to-port="2" /> <edge from-layer="3451" from-port="0" to-layer="3452" to-port="3" /> <edge from-layer="3452" from-port="4" to-layer="3453" to-port="1" /> <edge from-layer="3453" from-port="2" to-layer="3454" to-port="1" /> <edge from-layer="3454" from-port="2" to-layer="3474" to-port="0" /> <edge from-layer="3454" from-port="2" to-layer="3515" to-port="1" /> <edge from-layer="3454" from-port="2" to-layer="10161" to-port="0" /> <edge from-layer="3455" from-port="0" to-layer="3460" to-port="0" /> <edge from-layer="3456" from-port="0" to-layer="3459" to-port="1" /> <edge from-layer="3457" from-port="0" to-layer="3459" to-port="2" /> <edge from-layer="3458" from-port="0" to-layer="3459" to-port="3" /> <edge from-layer="3459" from-port="4" to-layer="3460" to-port="1" /> <edge from-layer="3460" from-port="2" to-layer="3461" to-port="0" /> <edge from-layer="3461" from-port="1" to-layer="3463" to-port="0" /> <edge from-layer="3462" from-port="0" to-layer="3463" to-port="2" /> <edge from-layer="3463" from-port="3" to-layer="3470" to-port="0" /> <edge from-layer="3464" from-port="0" to-layer="3467" to-port="1" /> <edge from-layer="3465" from-port="0" to-layer="3467" to-port="2" /> <edge from-layer="3466" from-port="0" to-layer="3467" to-port="3" /> <edge from-layer="3467" from-port="4" to-layer="3469" to-port="0" /> <edge from-layer="3468" from-port="0" to-layer="3469" to-port="1" /> <edge from-layer="3469" from-port="2" to-layer="3470" to-port="1" /> <edge from-layer="3470" from-port="2" to-layer="3515" to-port="2" /> <edge from-layer="3470" from-port="2" to-layer="10160" to-port="0" /> <edge from-layer="3471" from-port="0" to-layer="3489" to-port="0" /> <edge from-layer="3472" from-port="0" to-layer="3484" to-port="0" /> <edge from-layer="3473" from-port="0" to-layer="3484" to-port="1" /> <edge from-layer="3473" from-port="0" to-layer="3487" to-port="1" /> <edge from-layer="3474" from-port="1" to-layer="3476" to-port="0" /> <edge from-layer="3475" from-port="0" to-layer="3476" to-port="2" /> <edge from-layer="3476" from-port="3" to-layer="3486" to-port="0" /> <edge from-layer="3476" from-port="3" to-layer="3480" to-port="0" /> <edge from-layer="3477" from-port="1" to-layer="3504" to-port="0" /> <edge from-layer="3477" from-port="1" to-layer="3479" to-port="0" /> <edge from-layer="3478" from-port="0" to-layer="3479" to-port="2" /> <edge from-layer="3479" from-port="3" to-layer="3480" to-port="1" /> <edge from-layer="3480" from-port="2" to-layer="3482" to-port="0" /> <edge from-layer="3481" from-port="0" to-layer="3486" to-port="1" /> <edge from-layer="3481" from-port="0" to-layer="3482" to-port="1" /> <edge from-layer="3482" from-port="2" to-layer="3484" to-port="2" /> <edge from-layer="3483" from-port="0" to-layer="3484" to-port="3" /> <edge from-layer="3483" from-port="0" to-layer="3487" to-port="3" /> <edge from-layer="3484" from-port="4" to-layer="3489" to-port="1" /> <edge from-layer="3485" from-port="0" to-layer="3487" to-port="0" /> <edge from-layer="3486" from-port="2" to-layer="3487" to-port="2" /> <edge from-layer="3486" from-port="2" to-layer="3494" to-port="2" /> <edge from-layer="3487" from-port="4" to-layer="3489" to-port="2" /> <edge from-layer="3488" from-port="0" to-layer="3489" to-port="3" /> <edge from-layer="3489" from-port="4" to-layer="3496" to-port="0" /> <edge from-layer="3490" from-port="0" to-layer="3496" to-port="1" /> <edge from-layer="3491" from-port="0" to-layer="3494" to-port="0" /> <edge from-layer="3492" from-port="0" to-layer="3494" to-port="1" /> <edge from-layer="3493" from-port="0" to-layer="3494" to-port="3" /> <edge from-layer="3494" from-port="4" to-layer="3496" to-port="2" /> <edge from-layer="3495" from-port="0" to-layer="3496" to-port="3" /> <edge from-layer="3496" from-port="4" to-layer="3499" to-port="0" /> <edge from-layer="3497" from-port="0" to-layer="3499" to-port="1" /> <edge from-layer="3498" from-port="0" to-layer="3499" to-port="2" /> <edge from-layer="3499" from-port="3" to-layer="3500" to-port="0" /> <edge from-layer="3500" from-port="2" to-layer="3501" to-port="0" /> <edge from-layer="3500" from-port="2" to-layer="3514" to-port="0" /> <edge from-layer="3501" from-port="1" to-layer="3513" to-port="0" /> <edge from-layer="3502" from-port="0" to-layer="3504" to-port="1" /> <edge from-layer="3503" from-port="0" to-layer="3504" to-port="2" /> <edge from-layer="3504" from-port="3" to-layer="3508" to-port="0" /> <edge from-layer="3505" from-port="0" to-layer="3508" to-port="1" /> <edge from-layer="3506" from-port="0" to-layer="3508" to-port="2" /> <edge from-layer="3507" from-port="0" to-layer="3508" to-port="3" /> <edge from-layer="3508" from-port="4" to-layer="3510" to-port="0" /> <edge from-layer="3508" from-port="4" to-layer="3512" to-port="2" /> <edge from-layer="3509" from-port="0" to-layer="3510" to-port="1" /> <edge from-layer="3510" from-port="2" to-layer="3512" to-port="0" /> <edge from-layer="3511" from-port="0" to-layer="3512" to-port="1" /> <edge from-layer="3512" from-port="3" to-layer="3513" to-port="1" /> <edge from-layer="3513" from-port="2" to-layer="3514" to-port="1" /> <edge from-layer="3514" from-port="2" to-layer="3515" to-port="3" /> <edge from-layer="3515" from-port="4" to-layer="3517" to-port="0" /> <edge from-layer="3516" from-port="0" to-layer="3517" to-port="1" /> <edge from-layer="3517" from-port="2" to-layer="3519" to-port="0" /> <edge from-layer="3518" from-port="0" to-layer="3519" to-port="1" /> <edge from-layer="3519" from-port="2" to-layer="3527" to-port="0" /> <edge from-layer="3520" from-port="0" to-layer="3521" to-port="0" /> <edge from-layer="3521" from-port="1" to-layer="3524" to-port="0" /> <edge from-layer="3522" from-port="0" to-layer="3523" to-port="0" /> <edge from-layer="3523" from-port="1" to-layer="3524" to-port="1" /> <edge from-layer="3524" from-port="2" to-layer="3526" to-port="0" /> <edge from-layer="3525" from-port="0" to-layer="3526" to-port="1" /> <edge from-layer="3526" from-port="2" to-layer="3527" to-port="1" /> <edge from-layer="3527" from-port="2" to-layer="3529" to-port="0" /> <edge from-layer="3528" from-port="0" to-layer="3529" to-port="1" /> <edge from-layer="3529" from-port="2" to-layer="3530" to-port="1" /> <edge from-layer="3530" from-port="2" to-layer="3531" to-port="0" /> <edge from-layer="3531" from-port="2" to-layer="3560" to-port="0" /> <edge from-layer="3531" from-port="2" to-layer="3533" to-port="0" /> <edge from-layer="3531" from-port="2" to-layer="3762" to-port="1" /> <edge from-layer="3532" from-port="0" to-layer="3533" to-port="1" /> <edge from-layer="3533" from-port="2" to-layer="3535" to-port="0" /> <edge from-layer="3534" from-port="0" to-layer="3535" to-port="1" /> <edge from-layer="3535" from-port="2" to-layer="3537" to-port="0" /> <edge from-layer="3536" from-port="0" to-layer="3537" to-port="1" /> <edge from-layer="3537" from-port="2" to-layer="3545" to-port="0" /> <edge from-layer="3538" from-port="0" to-layer="3539" to-port="0" /> <edge from-layer="3539" from-port="1" to-layer="3542" to-port="0" /> <edge from-layer="3540" from-port="0" to-layer="3541" to-port="0" /> <edge from-layer="3541" from-port="1" to-layer="3542" to-port="1" /> <edge from-layer="3542" from-port="2" to-layer="3544" to-port="0" /> <edge from-layer="3543" from-port="0" to-layer="3544" to-port="1" /> <edge from-layer="3544" from-port="2" to-layer="3545" to-port="1" /> <edge from-layer="3545" from-port="2" to-layer="3547" to-port="0" /> <edge from-layer="3546" from-port="0" to-layer="3547" to-port="1" /> <edge from-layer="3547" from-port="2" to-layer="3548" to-port="0" /> <edge from-layer="3548" from-port="1" to-layer="3556" to-port="0" /> <edge from-layer="3549" from-port="0" to-layer="3550" to-port="0" /> <edge from-layer="3550" from-port="1" to-layer="3553" to-port="0" /> <edge from-layer="3551" from-port="0" to-layer="3552" to-port="0" /> <edge from-layer="3552" from-port="1" to-layer="3553" to-port="1" /> <edge from-layer="3553" from-port="2" to-layer="3555" to-port="0" /> <edge from-layer="3554" from-port="0" to-layer="3555" to-port="1" /> <edge from-layer="3555" from-port="2" to-layer="3556" to-port="1" /> <edge from-layer="3556" from-port="2" to-layer="3558" to-port="0" /> <edge from-layer="3557" from-port="0" to-layer="3558" to-port="1" /> <edge from-layer="3558" from-port="2" to-layer="3761" to-port="0" /> <edge from-layer="3559" from-port="0" to-layer="3560" to-port="1" /> <edge from-layer="3560" from-port="2" to-layer="3562" to-port="0" /> <edge from-layer="3561" from-port="0" to-layer="3562" to-port="1" /> <edge from-layer="3562" from-port="2" to-layer="3564" to-port="0" /> <edge from-layer="3563" from-port="0" to-layer="3564" to-port="1" /> <edge from-layer="3564" from-port="2" to-layer="3572" to-port="0" /> <edge from-layer="3565" from-port="0" to-layer="3566" to-port="0" /> <edge from-layer="3566" from-port="1" to-layer="3569" to-port="0" /> <edge from-layer="3567" from-port="0" to-layer="3568" to-port="0" /> <edge from-layer="3568" from-port="1" to-layer="3569" to-port="1" /> <edge from-layer="3569" from-port="2" to-layer="3571" to-port="0" /> <edge from-layer="3570" from-port="0" to-layer="3571" to-port="1" /> <edge from-layer="3571" from-port="2" to-layer="3572" to-port="1" /> <edge from-layer="3572" from-port="2" to-layer="3574" to-port="0" /> <edge from-layer="3573" from-port="0" to-layer="3574" to-port="1" /> <edge from-layer="3574" from-port="2" to-layer="3576" to-port="0" /> <edge from-layer="3575" from-port="0" to-layer="3576" to-port="1" /> <edge from-layer="3576" from-port="2" to-layer="3698" to-port="0" /> <edge from-layer="3576" from-port="2" to-layer="3592" to-port="0" /> <edge from-layer="3576" from-port="2" to-layer="3580" to-port="0" /> <edge from-layer="3577" from-port="0" to-layer="3580" to-port="1" /> <edge from-layer="3578" from-port="0" to-layer="3580" to-port="2" /> <edge from-layer="3579" from-port="0" to-layer="3580" to-port="3" /> <edge from-layer="3580" from-port="4" to-layer="3582" to-port="0" /> <edge from-layer="3581" from-port="0" to-layer="3582" to-port="1" /> <edge from-layer="3582" from-port="2" to-layer="3651" to-port="0" /> <edge from-layer="3582" from-port="2" to-layer="3586" to-port="0" /> <edge from-layer="3583" from-port="0" to-layer="3586" to-port="1" /> <edge from-layer="3584" from-port="0" to-layer="3586" to-port="2" /> <edge from-layer="3585" from-port="0" to-layer="3586" to-port="3" /> <edge from-layer="3586" from-port="4" to-layer="3618" to-port="0" /> <edge from-layer="3586" from-port="4" to-layer="3621" to-port="0" /> <edge from-layer="3586" from-port="4" to-layer="3637" to-port="0" /> <edge from-layer="3586" from-port="4" to-layer="3632" to-port="0" /> <edge from-layer="3587" from-port="0" to-layer="3614" to-port="0" /> <edge from-layer="3588" from-port="0" to-layer="3614" to-port="1" /> <edge from-layer="3589" from-port="0" to-layer="3592" to-port="1" /> <edge from-layer="3590" from-port="0" to-layer="3592" to-port="2" /> <edge from-layer="3591" from-port="0" to-layer="3592" to-port="3" /> <edge from-layer="3592" from-port="4" to-layer="3594" to-port="0" /> <edge from-layer="3593" from-port="0" to-layer="3594" to-port="1" /> <edge from-layer="3594" from-port="2" to-layer="3683" to-port="0" /> <edge from-layer="3594" from-port="2" to-layer="3656" to-port="0" /> <edge from-layer="3594" from-port="2" to-layer="3595" to-port="0" /> <edge from-layer="3595" from-port="1" to-layer="3597" to-port="0" /> <edge from-layer="3596" from-port="0" to-layer="3597" to-port="2" /> <edge from-layer="3597" from-port="3" to-layer="3610" to-port="0" /> <edge from-layer="3598" from-port="0" to-layer="3603" to-port="0" /> <edge from-layer="3599" from-port="0" to-layer="3602" to-port="1" /> <edge from-layer="3600" from-port="0" to-layer="3602" to-port="2" /> <edge from-layer="3601" from-port="0" to-layer="3602" to-port="3" /> <edge from-layer="3602" from-port="4" to-layer="3603" to-port="1" /> <edge from-layer="3603" from-port="2" to-layer="3604" to-port="0" /> <edge from-layer="3604" from-port="1" to-layer="3606" to-port="0" /> <edge from-layer="3605" from-port="0" to-layer="3606" to-port="2" /> <edge from-layer="3606" from-port="3" to-layer="3685" to-port="0" /> <edge from-layer="3606" from-port="3" to-layer="3607" to-port="0" /> <edge from-layer="3607" from-port="1" to-layer="3609" to-port="0" /> <edge from-layer="3608" from-port="0" to-layer="3609" to-port="2" /> <edge from-layer="3609" from-port="3" to-layer="3610" to-port="1" /> <edge from-layer="3610" from-port="2" to-layer="3612" to-port="0" /> <edge from-layer="3611" from-port="0" to-layer="3612" to-port="1" /> <edge from-layer="3612" from-port="2" to-layer="3614" to-port="2" /> <edge from-layer="3612" from-port="2" to-layer="3642" to-port="2" /> <edge from-layer="3613" from-port="0" to-layer="3614" to-port="3" /> <edge from-layer="3614" from-port="4" to-layer="3616" to-port="0" /> <edge from-layer="3615" from-port="0" to-layer="3616" to-port="2" /> <edge from-layer="3616" from-port="3" to-layer="3617" to-port="0" /> <edge from-layer="3617" from-port="2" to-layer="3618" to-port="1" /> <edge from-layer="3617" from-port="2" to-layer="3657" to-port="1" /> <edge from-layer="3618" from-port="2" to-layer="3647" to-port="0" /> <edge from-layer="3619" from-port="0" to-layer="3629" to-port="0" /> <edge from-layer="3620" from-port="0" to-layer="3629" to-port="1" /> <edge from-layer="3621" from-port="1" to-layer="3623" to-port="0" /> <edge from-layer="3622" from-port="0" to-layer="3623" to-port="2" /> <edge from-layer="3623" from-port="3" to-layer="3624" to-port="0" /> <edge from-layer="3624" from-port="2" to-layer="3625" to-port="0" /> <edge from-layer="3625" from-port="1" to-layer="3627" to-port="0" /> <edge from-layer="3626" from-port="0" to-layer="3627" to-port="1" /> <edge from-layer="3627" from-port="2" to-layer="3629" to-port="2" /> <edge from-layer="3628" from-port="0" to-layer="3629" to-port="3" /> <edge from-layer="3629" from-port="4" to-layer="3637" to-port="2" /> <edge from-layer="3629" from-port="4" to-layer="3632" to-port="1" /> <edge from-layer="3630" from-port="0" to-layer="3632" to-port="2" /> <edge from-layer="3631" from-port="0" to-layer="3632" to-port="3" /> <edge from-layer="3632" from-port="4" to-layer="3634" to-port="0" /> <edge from-layer="3633" from-port="0" to-layer="3634" to-port="1" /> <edge from-layer="3634" from-port="2" to-layer="3638" to-port="0" /> <edge from-layer="3635" from-port="0" to-layer="3637" to-port="1" /> <edge from-layer="3636" from-port="0" to-layer="3637" to-port="3" /> <edge from-layer="3637" from-port="4" to-layer="3638" to-port="1" /> <edge from-layer="3638" from-port="2" to-layer="3646" to-port="0" /> <edge from-layer="3639" from-port="0" to-layer="3642" to-port="0" /> <edge from-layer="3640" from-port="0" to-layer="3642" to-port="1" /> <edge from-layer="3641" from-port="0" to-layer="3642" to-port="3" /> <edge from-layer="3642" from-port="4" to-layer="3644" to-port="0" /> <edge from-layer="3643" from-port="0" to-layer="3644" to-port="2" /> <edge from-layer="3644" from-port="3" to-layer="3645" to-port="0" /> <edge from-layer="3645" from-port="2" to-layer="3646" to-port="1" /> <edge from-layer="3645" from-port="2" to-layer="3678" to-port="1" /> <edge from-layer="3646" from-port="2" to-layer="3647" to-port="1" /> <edge from-layer="3647" from-port="2" to-layer="3652" to-port="0" /> <edge from-layer="3648" from-port="0" to-layer="3651" to-port="1" /> <edge from-layer="3649" from-port="0" to-layer="3651" to-port="2" /> <edge from-layer="3650" from-port="0" to-layer="3651" to-port="3" /> <edge from-layer="3651" from-port="4" to-layer="3652" to-port="1" /> <edge from-layer="3652" from-port="2" to-layer="3746" to-port="0" /> <edge from-layer="3652" from-port="2" to-layer="3708" to-port="0" /> <edge from-layer="3653" from-port="0" to-layer="3656" to-port="1" /> <edge from-layer="3654" from-port="0" to-layer="3656" to-port="2" /> <edge from-layer="3655" from-port="0" to-layer="3656" to-port="3" /> <edge from-layer="3656" from-port="4" to-layer="3671" to-port="0" /> <edge from-layer="3656" from-port="4" to-layer="3660" to-port="0" /> <edge from-layer="3656" from-port="4" to-layer="3657" to-port="0" /> <edge from-layer="3656" from-port="4" to-layer="3676" to-port="0" /> <edge from-layer="3657" from-port="2" to-layer="3679" to-port="0" /> <edge from-layer="3658" from-port="0" to-layer="3668" to-port="0" /> <edge from-layer="3659" from-port="0" to-layer="3668" to-port="1" /> <edge from-layer="3660" from-port="1" to-layer="3662" to-port="0" /> <edge from-layer="3661" from-port="0" to-layer="3662" to-port="2" /> <edge from-layer="3662" from-port="3" to-layer="3663" to-port="0" /> <edge from-layer="3663" from-port="2" to-layer="3664" to-port="0" /> <edge from-layer="3664" from-port="1" to-layer="3666" to-port="0" /> <edge from-layer="3665" from-port="0" to-layer="3666" to-port="1" /> <edge from-layer="3666" from-port="2" to-layer="3668" to-port="2" /> <edge from-layer="3667" from-port="0" to-layer="3668" to-port="3" /> <edge from-layer="3668" from-port="4" to-layer="3676" to-port="2" /> <edge from-layer="3668" from-port="4" to-layer="3671" to-port="1" /> <edge from-layer="3669" from-port="0" to-layer="3671" to-port="2" /> <edge from-layer="3670" from-port="0" to-layer="3671" to-port="3" /> <edge from-layer="3671" from-port="4" to-layer="3673" to-port="0" /> <edge from-layer="3672" from-port="0" to-layer="3673" to-port="1" /> <edge from-layer="3673" from-port="2" to-layer="3677" to-port="0" /> <edge from-layer="3674" from-port="0" to-layer="3676" to-port="1" /> <edge from-layer="3675" from-port="0" to-layer="3676" to-port="3" /> <edge from-layer="3676" from-port="4" to-layer="3677" to-port="1" /> <edge from-layer="3677" from-port="2" to-layer="3678" to-port="0" /> <edge from-layer="3678" from-port="2" to-layer="3679" to-port="1" /> <edge from-layer="3679" from-port="2" to-layer="3684" to-port="0" /> <edge from-layer="3680" from-port="0" to-layer="3683" to-port="1" /> <edge from-layer="3681" from-port="0" to-layer="3683" to-port="2" /> <edge from-layer="3682" from-port="0" to-layer="3683" to-port="3" /> <edge from-layer="3683" from-port="4" to-layer="3684" to-port="1" /> <edge from-layer="3684" from-port="2" to-layer="3685" to-port="1" /> <edge from-layer="3685" from-port="2" to-layer="3705" to-port="0" /> <edge from-layer="3685" from-port="2" to-layer="10159" to-port="0" /> <edge from-layer="3685" from-port="2" to-layer="3746" to-port="1" /> <edge from-layer="3686" from-port="0" to-layer="3691" to-port="0" /> <edge from-layer="3687" from-port="0" to-layer="3690" to-port="1" /> <edge from-layer="3688" from-port="0" to-layer="3690" to-port="2" /> <edge from-layer="3689" from-port="0" to-layer="3690" to-port="3" /> <edge from-layer="3690" from-port="4" to-layer="3691" to-port="1" /> <edge from-layer="3691" from-port="2" to-layer="3692" to-port="0" /> <edge from-layer="3692" from-port="1" to-layer="3694" to-port="0" /> <edge from-layer="3693" from-port="0" to-layer="3694" to-port="2" /> <edge from-layer="3694" from-port="3" to-layer="3701" to-port="0" /> <edge from-layer="3695" from-port="0" to-layer="3698" to-port="1" /> <edge from-layer="3696" from-port="0" to-layer="3698" to-port="2" /> <edge from-layer="3697" from-port="0" to-layer="3698" to-port="3" /> <edge from-layer="3698" from-port="4" to-layer="3700" to-port="0" /> <edge from-layer="3699" from-port="0" to-layer="3700" to-port="1" /> <edge from-layer="3700" from-port="2" to-layer="3701" to-port="1" /> <edge from-layer="3701" from-port="2" to-layer="10158" to-port="0" /> <edge from-layer="3701" from-port="2" to-layer="3746" to-port="2" /> <edge from-layer="3702" from-port="0" to-layer="3720" to-port="0" /> <edge from-layer="3703" from-port="0" to-layer="3715" to-port="0" /> <edge from-layer="3704" from-port="0" to-layer="3715" to-port="1" /> <edge from-layer="3704" from-port="0" to-layer="3718" to-port="1" /> <edge from-layer="3705" from-port="1" to-layer="3707" to-port="0" /> <edge from-layer="3706" from-port="0" to-layer="3707" to-port="2" /> <edge from-layer="3707" from-port="3" to-layer="3711" to-port="0" /> <edge from-layer="3707" from-port="3" to-layer="3717" to-port="0" /> <edge from-layer="3708" from-port="1" to-layer="3735" to-port="0" /> <edge from-layer="3708" from-port="1" to-layer="3710" to-port="0" /> <edge from-layer="3709" from-port="0" to-layer="3710" to-port="2" /> <edge from-layer="3710" from-port="3" to-layer="3711" to-port="1" /> <edge from-layer="3711" from-port="2" to-layer="3713" to-port="0" /> <edge from-layer="3712" from-port="0" to-layer="3717" to-port="1" /> <edge from-layer="3712" from-port="0" to-layer="3713" to-port="1" /> <edge from-layer="3713" from-port="2" to-layer="3715" to-port="2" /> <edge from-layer="3714" from-port="0" to-layer="3718" to-port="3" /> <edge from-layer="3714" from-port="0" to-layer="3715" to-port="3" /> <edge from-layer="3715" from-port="4" to-layer="3720" to-port="1" /> <edge from-layer="3716" from-port="0" to-layer="3718" to-port="0" /> <edge from-layer="3717" from-port="2" to-layer="3718" to-port="2" /> <edge from-layer="3717" from-port="2" to-layer="3725" to-port="2" /> <edge from-layer="3718" from-port="4" to-layer="3720" to-port="2" /> <edge from-layer="3719" from-port="0" to-layer="3720" to-port="3" /> <edge from-layer="3720" from-port="4" to-layer="3727" to-port="0" /> <edge from-layer="3721" from-port="0" to-layer="3727" to-port="1" /> <edge from-layer="3722" from-port="0" to-layer="3725" to-port="0" /> <edge from-layer="3723" from-port="0" to-layer="3725" to-port="1" /> <edge from-layer="3724" from-port="0" to-layer="3725" to-port="3" /> <edge from-layer="3725" from-port="4" to-layer="3727" to-port="2" /> <edge from-layer="3726" from-port="0" to-layer="3727" to-port="3" /> <edge from-layer="3727" from-port="4" to-layer="3730" to-port="0" /> <edge from-layer="3728" from-port="0" to-layer="3730" to-port="1" /> <edge from-layer="3729" from-port="0" to-layer="3730" to-port="2" /> <edge from-layer="3730" from-port="3" to-layer="3731" to-port="0" /> <edge from-layer="3731" from-port="2" to-layer="3732" to-port="0" /> <edge from-layer="3731" from-port="2" to-layer="3745" to-port="0" /> <edge from-layer="3732" from-port="1" to-layer="3744" to-port="0" /> <edge from-layer="3733" from-port="0" to-layer="3735" to-port="1" /> <edge from-layer="3734" from-port="0" to-layer="3735" to-port="2" /> <edge from-layer="3735" from-port="3" to-layer="3739" to-port="0" /> <edge from-layer="3736" from-port="0" to-layer="3739" to-port="1" /> <edge from-layer="3737" from-port="0" to-layer="3739" to-port="2" /> <edge from-layer="3738" from-port="0" to-layer="3739" to-port="3" /> <edge from-layer="3739" from-port="4" to-layer="3743" to-port="2" /> <edge from-layer="3739" from-port="4" to-layer="3741" to-port="0" /> <edge from-layer="3740" from-port="0" to-layer="3741" to-port="1" /> <edge from-layer="3741" from-port="2" to-layer="3743" to-port="0" /> <edge from-layer="3742" from-port="0" to-layer="3743" to-port="1" /> <edge from-layer="3743" from-port="3" to-layer="3744" to-port="1" /> <edge from-layer="3744" from-port="2" to-layer="3745" to-port="1" /> <edge from-layer="3745" from-port="2" to-layer="3746" to-port="3" /> <edge from-layer="3746" from-port="4" to-layer="3748" to-port="0" /> <edge from-layer="3747" from-port="0" to-layer="3748" to-port="1" /> <edge from-layer="3748" from-port="2" to-layer="3750" to-port="0" /> <edge from-layer="3749" from-port="0" to-layer="3750" to-port="1" /> <edge from-layer="3750" from-port="2" to-layer="3758" to-port="0" /> <edge from-layer="3751" from-port="0" to-layer="3752" to-port="0" /> <edge from-layer="3752" from-port="1" to-layer="3755" to-port="0" /> <edge from-layer="3753" from-port="0" to-layer="3754" to-port="0" /> <edge from-layer="3754" from-port="1" to-layer="3755" to-port="1" /> <edge from-layer="3755" from-port="2" to-layer="3757" to-port="0" /> <edge from-layer="3756" from-port="0" to-layer="3757" to-port="1" /> <edge from-layer="3757" from-port="2" to-layer="3758" to-port="1" /> <edge from-layer="3758" from-port="2" to-layer="3760" to-port="0" /> <edge from-layer="3759" from-port="0" to-layer="3760" to-port="1" /> <edge from-layer="3760" from-port="2" to-layer="3761" to-port="1" /> <edge from-layer="3761" from-port="2" to-layer="3762" to-port="0" /> <edge from-layer="3762" from-port="2" to-layer="3791" to-port="0" /> <edge from-layer="3762" from-port="2" to-layer="3764" to-port="0" /> <edge from-layer="3762" from-port="2" to-layer="3993" to-port="1" /> <edge from-layer="3763" from-port="0" to-layer="3764" to-port="1" /> <edge from-layer="3764" from-port="2" to-layer="3766" to-port="0" /> <edge from-layer="3765" from-port="0" to-layer="3766" to-port="1" /> <edge from-layer="3766" from-port="2" to-layer="3768" to-port="0" /> <edge from-layer="3767" from-port="0" to-layer="3768" to-port="1" /> <edge from-layer="3768" from-port="2" to-layer="3776" to-port="0" /> <edge from-layer="3769" from-port="0" to-layer="3770" to-port="0" /> <edge from-layer="3770" from-port="1" to-layer="3773" to-port="0" /> <edge from-layer="3771" from-port="0" to-layer="3772" to-port="0" /> <edge from-layer="3772" from-port="1" to-layer="3773" to-port="1" /> <edge from-layer="3773" from-port="2" to-layer="3775" to-port="0" /> <edge from-layer="3774" from-port="0" to-layer="3775" to-port="1" /> <edge from-layer="3775" from-port="2" to-layer="3776" to-port="1" /> <edge from-layer="3776" from-port="2" to-layer="3778" to-port="0" /> <edge from-layer="3777" from-port="0" to-layer="3778" to-port="1" /> <edge from-layer="3778" from-port="2" to-layer="3779" to-port="0" /> <edge from-layer="3779" from-port="1" to-layer="3787" to-port="0" /> <edge from-layer="3780" from-port="0" to-layer="3781" to-port="0" /> <edge from-layer="3781" from-port="1" to-layer="3784" to-port="0" /> <edge from-layer="3782" from-port="0" to-layer="3783" to-port="0" /> <edge from-layer="3783" from-port="1" to-layer="3784" to-port="1" /> <edge from-layer="3784" from-port="2" to-layer="3786" to-port="0" /> <edge from-layer="3785" from-port="0" to-layer="3786" to-port="1" /> <edge from-layer="3786" from-port="2" to-layer="3787" to-port="1" /> <edge from-layer="3787" from-port="2" to-layer="3789" to-port="0" /> <edge from-layer="3788" from-port="0" to-layer="3789" to-port="1" /> <edge from-layer="3789" from-port="2" to-layer="3992" to-port="0" /> <edge from-layer="3790" from-port="0" to-layer="3791" to-port="1" /> <edge from-layer="3791" from-port="2" to-layer="3793" to-port="0" /> <edge from-layer="3792" from-port="0" to-layer="3793" to-port="1" /> <edge from-layer="3793" from-port="2" to-layer="3795" to-port="0" /> <edge from-layer="3794" from-port="0" to-layer="3795" to-port="1" /> <edge from-layer="3795" from-port="2" to-layer="3803" to-port="0" /> <edge from-layer="3796" from-port="0" to-layer="3797" to-port="0" /> <edge from-layer="3797" from-port="1" to-layer="3800" to-port="0" /> <edge from-layer="3798" from-port="0" to-layer="3799" to-port="0" /> <edge from-layer="3799" from-port="1" to-layer="3800" to-port="1" /> <edge from-layer="3800" from-port="2" to-layer="3802" to-port="0" /> <edge from-layer="3801" from-port="0" to-layer="3802" to-port="1" /> <edge from-layer="3802" from-port="2" to-layer="3803" to-port="1" /> <edge from-layer="3803" from-port="2" to-layer="3805" to-port="0" /> <edge from-layer="3804" from-port="0" to-layer="3805" to-port="1" /> <edge from-layer="3805" from-port="2" to-layer="3807" to-port="0" /> <edge from-layer="3806" from-port="0" to-layer="3807" to-port="1" /> <edge from-layer="3807" from-port="2" to-layer="3811" to-port="0" /> <edge from-layer="3807" from-port="2" to-layer="3823" to-port="0" /> <edge from-layer="3807" from-port="2" to-layer="3929" to-port="0" /> <edge from-layer="3808" from-port="0" to-layer="3811" to-port="1" /> <edge from-layer="3809" from-port="0" to-layer="3811" to-port="2" /> <edge from-layer="3810" from-port="0" to-layer="3811" to-port="3" /> <edge from-layer="3811" from-port="4" to-layer="3813" to-port="0" /> <edge from-layer="3812" from-port="0" to-layer="3813" to-port="1" /> <edge from-layer="3813" from-port="2" to-layer="3882" to-port="0" /> <edge from-layer="3813" from-port="2" to-layer="3817" to-port="0" /> <edge from-layer="3814" from-port="0" to-layer="3817" to-port="1" /> <edge from-layer="3815" from-port="0" to-layer="3817" to-port="2" /> <edge from-layer="3816" from-port="0" to-layer="3817" to-port="3" /> <edge from-layer="3817" from-port="4" to-layer="3868" to-port="0" /> <edge from-layer="3817" from-port="4" to-layer="3849" to-port="0" /> <edge from-layer="3817" from-port="4" to-layer="3852" to-port="0" /> <edge from-layer="3817" from-port="4" to-layer="3863" to-port="0" /> <edge from-layer="3818" from-port="0" to-layer="3845" to-port="0" /> <edge from-layer="3819" from-port="0" to-layer="3845" to-port="1" /> <edge from-layer="3820" from-port="0" to-layer="3823" to-port="1" /> <edge from-layer="3821" from-port="0" to-layer="3823" to-port="2" /> <edge from-layer="3822" from-port="0" to-layer="3823" to-port="3" /> <edge from-layer="3823" from-port="4" to-layer="3825" to-port="0" /> <edge from-layer="3824" from-port="0" to-layer="3825" to-port="1" /> <edge from-layer="3825" from-port="2" to-layer="3887" to-port="0" /> <edge from-layer="3825" from-port="2" to-layer="3826" to-port="0" /> <edge from-layer="3825" from-port="2" to-layer="3914" to-port="0" /> <edge from-layer="3826" from-port="1" to-layer="3828" to-port="0" /> <edge from-layer="3827" from-port="0" to-layer="3828" to-port="2" /> <edge from-layer="3828" from-port="3" to-layer="3841" to-port="0" /> <edge from-layer="3829" from-port="0" to-layer="3834" to-port="0" /> <edge from-layer="3830" from-port="0" to-layer="3833" to-port="1" /> <edge from-layer="3831" from-port="0" to-layer="3833" to-port="2" /> <edge from-layer="3832" from-port="0" to-layer="3833" to-port="3" /> <edge from-layer="3833" from-port="4" to-layer="3834" to-port="1" /> <edge from-layer="3834" from-port="2" to-layer="3835" to-port="0" /> <edge from-layer="3835" from-port="1" to-layer="3837" to-port="0" /> <edge from-layer="3836" from-port="0" to-layer="3837" to-port="2" /> <edge from-layer="3837" from-port="3" to-layer="3916" to-port="0" /> <edge from-layer="3837" from-port="3" to-layer="3838" to-port="0" /> <edge from-layer="3838" from-port="1" to-layer="3840" to-port="0" /> <edge from-layer="3839" from-port="0" to-layer="3840" to-port="2" /> <edge from-layer="3840" from-port="3" to-layer="3841" to-port="1" /> <edge from-layer="3841" from-port="2" to-layer="3843" to-port="0" /> <edge from-layer="3842" from-port="0" to-layer="3843" to-port="1" /> <edge from-layer="3843" from-port="2" to-layer="3873" to-port="2" /> <edge from-layer="3843" from-port="2" to-layer="3845" to-port="2" /> <edge from-layer="3844" from-port="0" to-layer="3845" to-port="3" /> <edge from-layer="3845" from-port="4" to-layer="3847" to-port="0" /> <edge from-layer="3846" from-port="0" to-layer="3847" to-port="2" /> <edge from-layer="3847" from-port="3" to-layer="3848" to-port="0" /> <edge from-layer="3848" from-port="2" to-layer="3849" to-port="1" /> <edge from-layer="3848" from-port="2" to-layer="3888" to-port="1" /> <edge from-layer="3849" from-port="2" to-layer="3878" to-port="0" /> <edge from-layer="3850" from-port="0" to-layer="3860" to-port="0" /> <edge from-layer="3851" from-port="0" to-layer="3860" to-port="1" /> <edge from-layer="3852" from-port="1" to-layer="3854" to-port="0" /> <edge from-layer="3853" from-port="0" to-layer="3854" to-port="2" /> <edge from-layer="3854" from-port="3" to-layer="3855" to-port="0" /> <edge from-layer="3855" from-port="2" to-layer="3856" to-port="0" /> <edge from-layer="3856" from-port="1" to-layer="3858" to-port="0" /> <edge from-layer="3857" from-port="0" to-layer="3858" to-port="1" /> <edge from-layer="3858" from-port="2" to-layer="3860" to-port="2" /> <edge from-layer="3859" from-port="0" to-layer="3860" to-port="3" /> <edge from-layer="3860" from-port="4" to-layer="3868" to-port="2" /> <edge from-layer="3860" from-port="4" to-layer="3863" to-port="1" /> <edge from-layer="3861" from-port="0" to-layer="3863" to-port="2" /> <edge from-layer="3862" from-port="0" to-layer="3863" to-port="3" /> <edge from-layer="3863" from-port="4" to-layer="3865" to-port="0" /> <edge from-layer="3864" from-port="0" to-layer="3865" to-port="1" /> <edge from-layer="3865" from-port="2" to-layer="3869" to-port="0" /> <edge from-layer="3866" from-port="0" to-layer="3868" to-port="1" /> <edge from-layer="3867" from-port="0" to-layer="3868" to-port="3" /> <edge from-layer="3868" from-port="4" to-layer="3869" to-port="1" /> <edge from-layer="3869" from-port="2" to-layer="3877" to-port="0" /> <edge from-layer="3870" from-port="0" to-layer="3873" to-port="0" /> <edge from-layer="3871" from-port="0" to-layer="3873" to-port="1" /> <edge from-layer="3872" from-port="0" to-layer="3873" to-port="3" /> <edge from-layer="3873" from-port="4" to-layer="3875" to-port="0" /> <edge from-layer="3874" from-port="0" to-layer="3875" to-port="2" /> <edge from-layer="3875" from-port="3" to-layer="3876" to-port="0" /> <edge from-layer="3876" from-port="2" to-layer="3877" to-port="1" /> <edge from-layer="3876" from-port="2" to-layer="3909" to-port="1" /> <edge from-layer="3877" from-port="2" to-layer="3878" to-port="1" /> <edge from-layer="3878" from-port="2" to-layer="3883" to-port="0" /> <edge from-layer="3879" from-port="0" to-layer="3882" to-port="1" /> <edge from-layer="3880" from-port="0" to-layer="3882" to-port="2" /> <edge from-layer="3881" from-port="0" to-layer="3882" to-port="3" /> <edge from-layer="3882" from-port="4" to-layer="3883" to-port="1" /> <edge from-layer="3883" from-port="2" to-layer="3977" to-port="0" /> <edge from-layer="3883" from-port="2" to-layer="3939" to-port="0" /> <edge from-layer="3884" from-port="0" to-layer="3887" to-port="1" /> <edge from-layer="3885" from-port="0" to-layer="3887" to-port="2" /> <edge from-layer="3886" from-port="0" to-layer="3887" to-port="3" /> <edge from-layer="3887" from-port="4" to-layer="3902" to-port="0" /> <edge from-layer="3887" from-port="4" to-layer="3907" to-port="0" /> <edge from-layer="3887" from-port="4" to-layer="3891" to-port="0" /> <edge from-layer="3887" from-port="4" to-layer="3888" to-port="0" /> <edge from-layer="3888" from-port="2" to-layer="3910" to-port="0" /> <edge from-layer="3889" from-port="0" to-layer="3899" to-port="0" /> <edge from-layer="3890" from-port="0" to-layer="3899" to-port="1" /> <edge from-layer="3891" from-port="1" to-layer="3893" to-port="0" /> <edge from-layer="3892" from-port="0" to-layer="3893" to-port="2" /> <edge from-layer="3893" from-port="3" to-layer="3894" to-port="0" /> <edge from-layer="3894" from-port="2" to-layer="3895" to-port="0" /> <edge from-layer="3895" from-port="1" to-layer="3897" to-port="0" /> <edge from-layer="3896" from-port="0" to-layer="3897" to-port="1" /> <edge from-layer="3897" from-port="2" to-layer="3899" to-port="2" /> <edge from-layer="3898" from-port="0" to-layer="3899" to-port="3" /> <edge from-layer="3899" from-port="4" to-layer="3902" to-port="1" /> <edge from-layer="3899" from-port="4" to-layer="3907" to-port="2" /> <edge from-layer="3900" from-port="0" to-layer="3902" to-port="2" /> <edge from-layer="3901" from-port="0" to-layer="3902" to-port="3" /> <edge from-layer="3902" from-port="4" to-layer="3904" to-port="0" /> <edge from-layer="3903" from-port="0" to-layer="3904" to-port="1" /> <edge from-layer="3904" from-port="2" to-layer="3908" to-port="0" /> <edge from-layer="3905" from-port="0" to-layer="3907" to-port="1" /> <edge from-layer="3906" from-port="0" to-layer="3907" to-port="3" /> <edge from-layer="3907" from-port="4" to-layer="3908" to-port="1" /> <edge from-layer="3908" from-port="2" to-layer="3909" to-port="0" /> <edge from-layer="3909" from-port="2" to-layer="3910" to-port="1" /> <edge from-layer="3910" from-port="2" to-layer="3915" to-port="0" /> <edge from-layer="3911" from-port="0" to-layer="3914" to-port="1" /> <edge from-layer="3912" from-port="0" to-layer="3914" to-port="2" /> <edge from-layer="3913" from-port="0" to-layer="3914" to-port="3" /> <edge from-layer="3914" from-port="4" to-layer="3915" to-port="1" /> <edge from-layer="3915" from-port="2" to-layer="3916" to-port="1" /> <edge from-layer="3916" from-port="2" to-layer="10157" to-port="0" /> <edge from-layer="3916" from-port="2" to-layer="3936" to-port="0" /> <edge from-layer="3916" from-port="2" to-layer="3977" to-port="1" /> <edge from-layer="3917" from-port="0" to-layer="3922" to-port="0" /> <edge from-layer="3918" from-port="0" to-layer="3921" to-port="1" /> <edge from-layer="3919" from-port="0" to-layer="3921" to-port="2" /> <edge from-layer="3920" from-port="0" to-layer="3921" to-port="3" /> <edge from-layer="3921" from-port="4" to-layer="3922" to-port="1" /> <edge from-layer="3922" from-port="2" to-layer="3923" to-port="0" /> <edge from-layer="3923" from-port="1" to-layer="3925" to-port="0" /> <edge from-layer="3924" from-port="0" to-layer="3925" to-port="2" /> <edge from-layer="3925" from-port="3" to-layer="3932" to-port="0" /> <edge from-layer="3926" from-port="0" to-layer="3929" to-port="1" /> <edge from-layer="3927" from-port="0" to-layer="3929" to-port="2" /> <edge from-layer="3928" from-port="0" to-layer="3929" to-port="3" /> <edge from-layer="3929" from-port="4" to-layer="3931" to-port="0" /> <edge from-layer="3930" from-port="0" to-layer="3931" to-port="1" /> <edge from-layer="3931" from-port="2" to-layer="3932" to-port="1" /> <edge from-layer="3932" from-port="2" to-layer="3977" to-port="2" /> <edge from-layer="3932" from-port="2" to-layer="10156" to-port="0" /> <edge from-layer="3933" from-port="0" to-layer="3951" to-port="0" /> <edge from-layer="3934" from-port="0" to-layer="3946" to-port="0" /> <edge from-layer="3935" from-port="0" to-layer="3946" to-port="1" /> <edge from-layer="3935" from-port="0" to-layer="3949" to-port="1" /> <edge from-layer="3936" from-port="1" to-layer="3938" to-port="0" /> <edge from-layer="3937" from-port="0" to-layer="3938" to-port="2" /> <edge from-layer="3938" from-port="3" to-layer="3948" to-port="0" /> <edge from-layer="3938" from-port="3" to-layer="3942" to-port="0" /> <edge from-layer="3939" from-port="1" to-layer="3941" to-port="0" /> <edge from-layer="3939" from-port="1" to-layer="3966" to-port="0" /> <edge from-layer="3940" from-port="0" to-layer="3941" to-port="2" /> <edge from-layer="3941" from-port="3" to-layer="3942" to-port="1" /> <edge from-layer="3942" from-port="2" to-layer="3944" to-port="0" /> <edge from-layer="3943" from-port="0" to-layer="3948" to-port="1" /> <edge from-layer="3943" from-port="0" to-layer="3944" to-port="1" /> <edge from-layer="3944" from-port="2" to-layer="3946" to-port="2" /> <edge from-layer="3945" from-port="0" to-layer="3949" to-port="3" /> <edge from-layer="3945" from-port="0" to-layer="3946" to-port="3" /> <edge from-layer="3946" from-port="4" to-layer="3951" to-port="1" /> <edge from-layer="3947" from-port="0" to-layer="3949" to-port="0" /> <edge from-layer="3948" from-port="2" to-layer="3949" to-port="2" /> <edge from-layer="3948" from-port="2" to-layer="3956" to-port="2" /> <edge from-layer="3949" from-port="4" to-layer="3951" to-port="2" /> <edge from-layer="3950" from-port="0" to-layer="3951" to-port="3" /> <edge from-layer="3951" from-port="4" to-layer="3958" to-port="0" /> <edge from-layer="3952" from-port="0" to-layer="3958" to-port="1" /> <edge from-layer="3953" from-port="0" to-layer="3956" to-port="0" /> <edge from-layer="3954" from-port="0" to-layer="3956" to-port="1" /> <edge from-layer="3955" from-port="0" to-layer="3956" to-port="3" /> <edge from-layer="3956" from-port="4" to-layer="3958" to-port="2" /> <edge from-layer="3957" from-port="0" to-layer="3958" to-port="3" /> <edge from-layer="3958" from-port="4" to-layer="3961" to-port="0" /> <edge from-layer="3959" from-port="0" to-layer="3961" to-port="1" /> <edge from-layer="3960" from-port="0" to-layer="3961" to-port="2" /> <edge from-layer="3961" from-port="3" to-layer="3962" to-port="0" /> <edge from-layer="3962" from-port="2" to-layer="3976" to-port="0" /> <edge from-layer="3962" from-port="2" to-layer="3963" to-port="0" /> <edge from-layer="3963" from-port="1" to-layer="3975" to-port="0" /> <edge from-layer="3964" from-port="0" to-layer="3966" to-port="1" /> <edge from-layer="3965" from-port="0" to-layer="3966" to-port="2" /> <edge from-layer="3966" from-port="3" to-layer="3970" to-port="0" /> <edge from-layer="3967" from-port="0" to-layer="3970" to-port="1" /> <edge from-layer="3968" from-port="0" to-layer="3970" to-port="2" /> <edge from-layer="3969" from-port="0" to-layer="3970" to-port="3" /> <edge from-layer="3970" from-port="4" to-layer="3974" to-port="2" /> <edge from-layer="3970" from-port="4" to-layer="3972" to-port="0" /> <edge from-layer="3971" from-port="0" to-layer="3972" to-port="1" /> <edge from-layer="3972" from-port="2" to-layer="3974" to-port="0" /> <edge from-layer="3973" from-port="0" to-layer="3974" to-port="1" /> <edge from-layer="3974" from-port="3" to-layer="3975" to-port="1" /> <edge from-layer="3975" from-port="2" to-layer="3976" to-port="1" /> <edge from-layer="3976" from-port="2" to-layer="3977" to-port="3" /> <edge from-layer="3977" from-port="4" to-layer="3979" to-port="0" /> <edge from-layer="3978" from-port="0" to-layer="3979" to-port="1" /> <edge from-layer="3979" from-port="2" to-layer="3981" to-port="0" /> <edge from-layer="3980" from-port="0" to-layer="3981" to-port="1" /> <edge from-layer="3981" from-port="2" to-layer="3989" to-port="0" /> <edge from-layer="3982" from-port="0" to-layer="3983" to-port="0" /> <edge from-layer="3983" from-port="1" to-layer="3986" to-port="0" /> <edge from-layer="3984" from-port="0" to-layer="3985" to-port="0" /> <edge from-layer="3985" from-port="1" to-layer="3986" to-port="1" /> <edge from-layer="3986" from-port="2" to-layer="3988" to-port="0" /> <edge from-layer="3987" from-port="0" to-layer="3988" to-port="1" /> <edge from-layer="3988" from-port="2" to-layer="3989" to-port="1" /> <edge from-layer="3989" from-port="2" to-layer="3991" to-port="0" /> <edge from-layer="3990" from-port="0" to-layer="3991" to-port="1" /> <edge from-layer="3991" from-port="2" to-layer="3992" to-port="1" /> <edge from-layer="3992" from-port="2" to-layer="3993" to-port="0" /> <edge from-layer="3993" from-port="2" to-layer="4022" to-port="0" /> <edge from-layer="3993" from-port="2" to-layer="3995" to-port="0" /> <edge from-layer="3993" from-port="2" to-layer="4224" to-port="1" /> <edge from-layer="3994" from-port="0" to-layer="3995" to-port="1" /> <edge from-layer="3995" from-port="2" to-layer="3997" to-port="0" /> <edge from-layer="3996" from-port="0" to-layer="3997" to-port="1" /> <edge from-layer="3997" from-port="2" to-layer="3999" to-port="0" /> <edge from-layer="3998" from-port="0" to-layer="3999" to-port="1" /> <edge from-layer="3999" from-port="2" to-layer="4007" to-port="0" /> <edge from-layer="4000" from-port="0" to-layer="4001" to-port="0" /> <edge from-layer="4001" from-port="1" to-layer="4004" to-port="0" /> <edge from-layer="4002" from-port="0" to-layer="4003" to-port="0" /> <edge from-layer="4003" from-port="1" to-layer="4004" to-port="1" /> <edge from-layer="4004" from-port="2" to-layer="4006" to-port="0" /> <edge from-layer="4005" from-port="0" to-layer="4006" to-port="1" /> <edge from-layer="4006" from-port="2" to-layer="4007" to-port="1" /> <edge from-layer="4007" from-port="2" to-layer="4009" to-port="0" /> <edge from-layer="4008" from-port="0" to-layer="4009" to-port="1" /> <edge from-layer="4009" from-port="2" to-layer="4010" to-port="0" /> <edge from-layer="4010" from-port="1" to-layer="4018" to-port="0" /> <edge from-layer="4011" from-port="0" to-layer="4012" to-port="0" /> <edge from-layer="4012" from-port="1" to-layer="4015" to-port="0" /> <edge from-layer="4013" from-port="0" to-layer="4014" to-port="0" /> <edge from-layer="4014" from-port="1" to-layer="4015" to-port="1" /> <edge from-layer="4015" from-port="2" to-layer="4017" to-port="0" /> <edge from-layer="4016" from-port="0" to-layer="4017" to-port="1" /> <edge from-layer="4017" from-port="2" to-layer="4018" to-port="1" /> <edge from-layer="4018" from-port="2" to-layer="4020" to-port="0" /> <edge from-layer="4019" from-port="0" to-layer="4020" to-port="1" /> <edge from-layer="4020" from-port="2" to-layer="4223" to-port="0" /> <edge from-layer="4021" from-port="0" to-layer="4022" to-port="1" /> <edge from-layer="4022" from-port="2" to-layer="4024" to-port="0" /> <edge from-layer="4023" from-port="0" to-layer="4024" to-port="1" /> <edge from-layer="4024" from-port="2" to-layer="4026" to-port="0" /> <edge from-layer="4025" from-port="0" to-layer="4026" to-port="1" /> <edge from-layer="4026" from-port="2" to-layer="4034" to-port="0" /> <edge from-layer="4027" from-port="0" to-layer="4028" to-port="0" /> <edge from-layer="4028" from-port="1" to-layer="4031" to-port="0" /> <edge from-layer="4029" from-port="0" to-layer="4030" to-port="0" /> <edge from-layer="4030" from-port="1" to-layer="4031" to-port="1" /> <edge from-layer="4031" from-port="2" to-layer="4033" to-port="0" /> <edge from-layer="4032" from-port="0" to-layer="4033" to-port="1" /> <edge from-layer="4033" from-port="2" to-layer="4034" to-port="1" /> <edge from-layer="4034" from-port="2" to-layer="4036" to-port="0" /> <edge from-layer="4035" from-port="0" to-layer="4036" to-port="1" /> <edge from-layer="4036" from-port="2" to-layer="4038" to-port="0" /> <edge from-layer="4037" from-port="0" to-layer="4038" to-port="1" /> <edge from-layer="4038" from-port="2" to-layer="4054" to-port="0" /> <edge from-layer="4038" from-port="2" to-layer="4160" to-port="0" /> <edge from-layer="4038" from-port="2" to-layer="4042" to-port="0" /> <edge from-layer="4039" from-port="0" to-layer="4042" to-port="1" /> <edge from-layer="4040" from-port="0" to-layer="4042" to-port="2" /> <edge from-layer="4041" from-port="0" to-layer="4042" to-port="3" /> <edge from-layer="4042" from-port="4" to-layer="4044" to-port="0" /> <edge from-layer="4043" from-port="0" to-layer="4044" to-port="1" /> <edge from-layer="4044" from-port="2" to-layer="4048" to-port="0" /> <edge from-layer="4044" from-port="2" to-layer="4113" to-port="0" /> <edge from-layer="4045" from-port="0" to-layer="4048" to-port="1" /> <edge from-layer="4046" from-port="0" to-layer="4048" to-port="2" /> <edge from-layer="4047" from-port="0" to-layer="4048" to-port="3" /> <edge from-layer="4048" from-port="4" to-layer="4094" to-port="0" /> <edge from-layer="4048" from-port="4" to-layer="4080" to-port="0" /> <edge from-layer="4048" from-port="4" to-layer="4099" to-port="0" /> <edge from-layer="4048" from-port="4" to-layer="4083" to-port="0" /> <edge from-layer="4049" from-port="0" to-layer="4076" to-port="0" /> <edge from-layer="4050" from-port="0" to-layer="4076" to-port="1" /> <edge from-layer="4051" from-port="0" to-layer="4054" to-port="1" /> <edge from-layer="4052" from-port="0" to-layer="4054" to-port="2" /> <edge from-layer="4053" from-port="0" to-layer="4054" to-port="3" /> <edge from-layer="4054" from-port="4" to-layer="4056" to-port="0" /> <edge from-layer="4055" from-port="0" to-layer="4056" to-port="1" /> <edge from-layer="4056" from-port="2" to-layer="4057" to-port="0" /> <edge from-layer="4056" from-port="2" to-layer="4145" to-port="0" /> <edge from-layer="4056" from-port="2" to-layer="4118" to-port="0" /> <edge from-layer="4057" from-port="1" to-layer="4059" to-port="0" /> <edge from-layer="4058" from-port="0" to-layer="4059" to-port="2" /> <edge from-layer="4059" from-port="3" to-layer="4072" to-port="0" /> <edge from-layer="4060" from-port="0" to-layer="4065" to-port="0" /> <edge from-layer="4061" from-port="0" to-layer="4064" to-port="1" /> <edge from-layer="4062" from-port="0" to-layer="4064" to-port="2" /> <edge from-layer="4063" from-port="0" to-layer="4064" to-port="3" /> <edge from-layer="4064" from-port="4" to-layer="4065" to-port="1" /> <edge from-layer="4065" from-port="2" to-layer="4066" to-port="0" /> <edge from-layer="4066" from-port="1" to-layer="4068" to-port="0" /> <edge from-layer="4067" from-port="0" to-layer="4068" to-port="2" /> <edge from-layer="4068" from-port="3" to-layer="4069" to-port="0" /> <edge from-layer="4068" from-port="3" to-layer="4147" to-port="0" /> <edge from-layer="4069" from-port="1" to-layer="4071" to-port="0" /> <edge from-layer="4070" from-port="0" to-layer="4071" to-port="2" /> <edge from-layer="4071" from-port="3" to-layer="4072" to-port="1" /> <edge from-layer="4072" from-port="2" to-layer="4074" to-port="0" /> <edge from-layer="4073" from-port="0" to-layer="4074" to-port="1" /> <edge from-layer="4074" from-port="2" to-layer="4076" to-port="2" /> <edge from-layer="4074" from-port="2" to-layer="4104" to-port="2" /> <edge from-layer="4075" from-port="0" to-layer="4076" to-port="3" /> <edge from-layer="4076" from-port="4" to-layer="4078" to-port="0" /> <edge from-layer="4077" from-port="0" to-layer="4078" to-port="2" /> <edge from-layer="4078" from-port="3" to-layer="4079" to-port="0" /> <edge from-layer="4079" from-port="2" to-layer="4119" to-port="1" /> <edge from-layer="4079" from-port="2" to-layer="4080" to-port="1" /> <edge from-layer="4080" from-port="2" to-layer="4109" to-port="0" /> <edge from-layer="4081" from-port="0" to-layer="4091" to-port="0" /> <edge from-layer="4082" from-port="0" to-layer="4091" to-port="1" /> <edge from-layer="4083" from-port="1" to-layer="4085" to-port="0" /> <edge from-layer="4084" from-port="0" to-layer="4085" to-port="2" /> <edge from-layer="4085" from-port="3" to-layer="4086" to-port="0" /> <edge from-layer="4086" from-port="2" to-layer="4087" to-port="0" /> <edge from-layer="4087" from-port="1" to-layer="4089" to-port="0" /> <edge from-layer="4088" from-port="0" to-layer="4089" to-port="1" /> <edge from-layer="4089" from-port="2" to-layer="4091" to-port="2" /> <edge from-layer="4090" from-port="0" to-layer="4091" to-port="3" /> <edge from-layer="4091" from-port="4" to-layer="4094" to-port="1" /> <edge from-layer="4091" from-port="4" to-layer="4099" to-port="2" /> <edge from-layer="4092" from-port="0" to-layer="4094" to-port="2" /> <edge from-layer="4093" from-port="0" to-layer="4094" to-port="3" /> <edge from-layer="4094" from-port="4" to-layer="4096" to-port="0" /> <edge from-layer="4095" from-port="0" to-layer="4096" to-port="1" /> <edge from-layer="4096" from-port="2" to-layer="4100" to-port="0" /> <edge from-layer="4097" from-port="0" to-layer="4099" to-port="1" /> <edge from-layer="4098" from-port="0" to-layer="4099" to-port="3" /> <edge from-layer="4099" from-port="4" to-layer="4100" to-port="1" /> <edge from-layer="4100" from-port="2" to-layer="4108" to-port="0" /> <edge from-layer="4101" from-port="0" to-layer="4104" to-port="0" /> <edge from-layer="4102" from-port="0" to-layer="4104" to-port="1" /> <edge from-layer="4103" from-port="0" to-layer="4104" to-port="3" /> <edge from-layer="4104" from-port="4" to-layer="4106" to-port="0" /> <edge from-layer="4105" from-port="0" to-layer="4106" to-port="2" /> <edge from-layer="4106" from-port="3" to-layer="4107" to-port="0" /> <edge from-layer="4107" from-port="2" to-layer="4140" to-port="1" /> <edge from-layer="4107" from-port="2" to-layer="4108" to-port="1" /> <edge from-layer="4108" from-port="2" to-layer="4109" to-port="1" /> <edge from-layer="4109" from-port="2" to-layer="4114" to-port="0" /> <edge from-layer="4110" from-port="0" to-layer="4113" to-port="1" /> <edge from-layer="4111" from-port="0" to-layer="4113" to-port="2" /> <edge from-layer="4112" from-port="0" to-layer="4113" to-port="3" /> <edge from-layer="4113" from-port="4" to-layer="4114" to-port="1" /> <edge from-layer="4114" from-port="2" to-layer="4170" to-port="0" /> <edge from-layer="4114" from-port="2" to-layer="4208" to-port="0" /> <edge from-layer="4115" from-port="0" to-layer="4118" to-port="1" /> <edge from-layer="4116" from-port="0" to-layer="4118" to-port="2" /> <edge from-layer="4117" from-port="0" to-layer="4118" to-port="3" /> <edge from-layer="4118" from-port="4" to-layer="4133" to-port="0" /> <edge from-layer="4118" from-port="4" to-layer="4122" to-port="0" /> <edge from-layer="4118" from-port="4" to-layer="4138" to-port="0" /> <edge from-layer="4118" from-port="4" to-layer="4119" to-port="0" /> <edge from-layer="4119" from-port="2" to-layer="4141" to-port="0" /> <edge from-layer="4120" from-port="0" to-layer="4130" to-port="0" /> <edge from-layer="4121" from-port="0" to-layer="4130" to-port="1" /> <edge from-layer="4122" from-port="1" to-layer="4124" to-port="0" /> <edge from-layer="4123" from-port="0" to-layer="4124" to-port="2" /> <edge from-layer="4124" from-port="3" to-layer="4125" to-port="0" /> <edge from-layer="4125" from-port="2" to-layer="4126" to-port="0" /> <edge from-layer="4126" from-port="1" to-layer="4128" to-port="0" /> <edge from-layer="4127" from-port="0" to-layer="4128" to-port="1" /> <edge from-layer="4128" from-port="2" to-layer="4130" to-port="2" /> <edge from-layer="4129" from-port="0" to-layer="4130" to-port="3" /> <edge from-layer="4130" from-port="4" to-layer="4138" to-port="2" /> <edge from-layer="4130" from-port="4" to-layer="4133" to-port="1" /> <edge from-layer="4131" from-port="0" to-layer="4133" to-port="2" /> <edge from-layer="4132" from-port="0" to-layer="4133" to-port="3" /> <edge from-layer="4133" from-port="4" to-layer="4135" to-port="0" /> <edge from-layer="4134" from-port="0" to-layer="4135" to-port="1" /> <edge from-layer="4135" from-port="2" to-layer="4139" to-port="0" /> <edge from-layer="4136" from-port="0" to-layer="4138" to-port="1" /> <edge from-layer="4137" from-port="0" to-layer="4138" to-port="3" /> <edge from-layer="4138" from-port="4" to-layer="4139" to-port="1" /> <edge from-layer="4139" from-port="2" to-layer="4140" to-port="0" /> <edge from-layer="4140" from-port="2" to-layer="4141" to-port="1" /> <edge from-layer="4141" from-port="2" to-layer="4146" to-port="0" /> <edge from-layer="4142" from-port="0" to-layer="4145" to-port="1" /> <edge from-layer="4143" from-port="0" to-layer="4145" to-port="2" /> <edge from-layer="4144" from-port="0" to-layer="4145" to-port="3" /> <edge from-layer="4145" from-port="4" to-layer="4146" to-port="1" /> <edge from-layer="4146" from-port="2" to-layer="4147" to-port="1" /> <edge from-layer="4147" from-port="2" to-layer="4167" to-port="0" /> <edge from-layer="4147" from-port="2" to-layer="10155" to-port="0" /> <edge from-layer="4147" from-port="2" to-layer="4208" to-port="1" /> <edge from-layer="4148" from-port="0" to-layer="4153" to-port="0" /> <edge from-layer="4149" from-port="0" to-layer="4152" to-port="1" /> <edge from-layer="4150" from-port="0" to-layer="4152" to-port="2" /> <edge from-layer="4151" from-port="0" to-layer="4152" to-port="3" /> <edge from-layer="4152" from-port="4" to-layer="4153" to-port="1" /> <edge from-layer="4153" from-port="2" to-layer="4154" to-port="0" /> <edge from-layer="4154" from-port="1" to-layer="4156" to-port="0" /> <edge from-layer="4155" from-port="0" to-layer="4156" to-port="2" /> <edge from-layer="4156" from-port="3" to-layer="4163" to-port="0" /> <edge from-layer="4157" from-port="0" to-layer="4160" to-port="1" /> <edge from-layer="4158" from-port="0" to-layer="4160" to-port="2" /> <edge from-layer="4159" from-port="0" to-layer="4160" to-port="3" /> <edge from-layer="4160" from-port="4" to-layer="4162" to-port="0" /> <edge from-layer="4161" from-port="0" to-layer="4162" to-port="1" /> <edge from-layer="4162" from-port="2" to-layer="4163" to-port="1" /> <edge from-layer="4163" from-port="2" to-layer="4208" to-port="2" /> <edge from-layer="4163" from-port="2" to-layer="10154" to-port="0" /> <edge from-layer="4164" from-port="0" to-layer="4182" to-port="0" /> <edge from-layer="4165" from-port="0" to-layer="4177" to-port="0" /> <edge from-layer="4166" from-port="0" to-layer="4177" to-port="1" /> <edge from-layer="4166" from-port="0" to-layer="4180" to-port="1" /> <edge from-layer="4167" from-port="1" to-layer="4169" to-port="0" /> <edge from-layer="4168" from-port="0" to-layer="4169" to-port="2" /> <edge from-layer="4169" from-port="3" to-layer="4173" to-port="0" /> <edge from-layer="4169" from-port="3" to-layer="4179" to-port="0" /> <edge from-layer="4170" from-port="1" to-layer="4172" to-port="0" /> <edge from-layer="4170" from-port="1" to-layer="4197" to-port="0" /> <edge from-layer="4171" from-port="0" to-layer="4172" to-port="2" /> <edge from-layer="4172" from-port="3" to-layer="4173" to-port="1" /> <edge from-layer="4173" from-port="2" to-layer="4175" to-port="0" /> <edge from-layer="4174" from-port="0" to-layer="4175" to-port="1" /> <edge from-layer="4174" from-port="0" to-layer="4179" to-port="1" /> <edge from-layer="4175" from-port="2" to-layer="4177" to-port="2" /> <edge from-layer="4176" from-port="0" to-layer="4177" to-port="3" /> <edge from-layer="4176" from-port="0" to-layer="4180" to-port="3" /> <edge from-layer="4177" from-port="4" to-layer="4182" to-port="1" /> <edge from-layer="4178" from-port="0" to-layer="4180" to-port="0" /> <edge from-layer="4179" from-port="2" to-layer="4187" to-port="2" /> <edge from-layer="4179" from-port="2" to-layer="4180" to-port="2" /> <edge from-layer="4180" from-port="4" to-layer="4182" to-port="2" /> <edge from-layer="4181" from-port="0" to-layer="4182" to-port="3" /> <edge from-layer="4182" from-port="4" to-layer="4189" to-port="0" /> <edge from-layer="4183" from-port="0" to-layer="4189" to-port="1" /> <edge from-layer="4184" from-port="0" to-layer="4187" to-port="0" /> <edge from-layer="4185" from-port="0" to-layer="4187" to-port="1" /> <edge from-layer="4186" from-port="0" to-layer="4187" to-port="3" /> <edge from-layer="4187" from-port="4" to-layer="4189" to-port="2" /> <edge from-layer="4188" from-port="0" to-layer="4189" to-port="3" /> <edge from-layer="4189" from-port="4" to-layer="4192" to-port="0" /> <edge from-layer="4190" from-port="0" to-layer="4192" to-port="1" /> <edge from-layer="4191" from-port="0" to-layer="4192" to-port="2" /> <edge from-layer="4192" from-port="3" to-layer="4193" to-port="0" /> <edge from-layer="4193" from-port="2" to-layer="4194" to-port="0" /> <edge from-layer="4193" from-port="2" to-layer="4207" to-port="0" /> <edge from-layer="4194" from-port="1" to-layer="4206" to-port="0" /> <edge from-layer="4195" from-port="0" to-layer="4197" to-port="1" /> <edge from-layer="4196" from-port="0" to-layer="4197" to-port="2" /> <edge from-layer="4197" from-port="3" to-layer="4201" to-port="0" /> <edge from-layer="4198" from-port="0" to-layer="4201" to-port="1" /> <edge from-layer="4199" from-port="0" to-layer="4201" to-port="2" /> <edge from-layer="4200" from-port="0" to-layer="4201" to-port="3" /> <edge from-layer="4201" from-port="4" to-layer="4203" to-port="0" /> <edge from-layer="4201" from-port="4" to-layer="4205" to-port="2" /> <edge from-layer="4202" from-port="0" to-layer="4203" to-port="1" /> <edge from-layer="4203" from-port="2" to-layer="4205" to-port="0" /> <edge from-layer="4204" from-port="0" to-layer="4205" to-port="1" /> <edge from-layer="4205" from-port="3" to-layer="4206" to-port="1" /> <edge from-layer="4206" from-port="2" to-layer="4207" to-port="1" /> <edge from-layer="4207" from-port="2" to-layer="4208" to-port="3" /> <edge from-layer="4208" from-port="4" to-layer="4210" to-port="0" /> <edge from-layer="4209" from-port="0" to-layer="4210" to-port="1" /> <edge from-layer="4210" from-port="2" to-layer="4212" to-port="0" /> <edge from-layer="4211" from-port="0" to-layer="4212" to-port="1" /> <edge from-layer="4212" from-port="2" to-layer="4220" to-port="0" /> <edge from-layer="4213" from-port="0" to-layer="4214" to-port="0" /> <edge from-layer="4214" from-port="1" to-layer="4217" to-port="0" /> <edge from-layer="4215" from-port="0" to-layer="4216" to-port="0" /> <edge from-layer="4216" from-port="1" to-layer="4217" to-port="1" /> <edge from-layer="4217" from-port="2" to-layer="4219" to-port="0" /> <edge from-layer="4218" from-port="0" to-layer="4219" to-port="1" /> <edge from-layer="4219" from-port="2" to-layer="4220" to-port="1" /> <edge from-layer="4220" from-port="2" to-layer="4222" to-port="0" /> <edge from-layer="4221" from-port="0" to-layer="4222" to-port="1" /> <edge from-layer="4222" from-port="2" to-layer="4223" to-port="1" /> <edge from-layer="4223" from-port="2" to-layer="4224" to-port="0" /> <edge from-layer="4224" from-port="2" to-layer="4455" to-port="1" /> <edge from-layer="4224" from-port="2" to-layer="4253" to-port="0" /> <edge from-layer="4224" from-port="2" to-layer="4226" to-port="0" /> <edge from-layer="4225" from-port="0" to-layer="4226" to-port="1" /> <edge from-layer="4226" from-port="2" to-layer="4228" to-port="0" /> <edge from-layer="4227" from-port="0" to-layer="4228" to-port="1" /> <edge from-layer="4228" from-port="2" to-layer="4230" to-port="0" /> <edge from-layer="4229" from-port="0" to-layer="4230" to-port="1" /> <edge from-layer="4230" from-port="2" to-layer="4238" to-port="0" /> <edge from-layer="4231" from-port="0" to-layer="4232" to-port="0" /> <edge from-layer="4232" from-port="1" to-layer="4235" to-port="0" /> <edge from-layer="4233" from-port="0" to-layer="4234" to-port="0" /> <edge from-layer="4234" from-port="1" to-layer="4235" to-port="1" /> <edge from-layer="4235" from-port="2" to-layer="4237" to-port="0" /> <edge from-layer="4236" from-port="0" to-layer="4237" to-port="1" /> <edge from-layer="4237" from-port="2" to-layer="4238" to-port="1" /> <edge from-layer="4238" from-port="2" to-layer="4240" to-port="0" /> <edge from-layer="4239" from-port="0" to-layer="4240" to-port="1" /> <edge from-layer="4240" from-port="2" to-layer="4241" to-port="0" /> <edge from-layer="4241" from-port="1" to-layer="4249" to-port="0" /> <edge from-layer="4242" from-port="0" to-layer="4243" to-port="0" /> <edge from-layer="4243" from-port="1" to-layer="4246" to-port="0" /> <edge from-layer="4244" from-port="0" to-layer="4245" to-port="0" /> <edge from-layer="4245" from-port="1" to-layer="4246" to-port="1" /> <edge from-layer="4246" from-port="2" to-layer="4248" to-port="0" /> <edge from-layer="4247" from-port="0" to-layer="4248" to-port="1" /> <edge from-layer="4248" from-port="2" to-layer="4249" to-port="1" /> <edge from-layer="4249" from-port="2" to-layer="4251" to-port="0" /> <edge from-layer="4250" from-port="0" to-layer="4251" to-port="1" /> <edge from-layer="4251" from-port="2" to-layer="4454" to-port="0" /> <edge from-layer="4252" from-port="0" to-layer="4253" to-port="1" /> <edge from-layer="4253" from-port="2" to-layer="4255" to-port="0" /> <edge from-layer="4254" from-port="0" to-layer="4255" to-port="1" /> <edge from-layer="4255" from-port="2" to-layer="4257" to-port="0" /> <edge from-layer="4256" from-port="0" to-layer="4257" to-port="1" /> <edge from-layer="4257" from-port="2" to-layer="4265" to-port="0" /> <edge from-layer="4258" from-port="0" to-layer="4259" to-port="0" /> <edge from-layer="4259" from-port="1" to-layer="4262" to-port="0" /> <edge from-layer="4260" from-port="0" to-layer="4261" to-port="0" /> <edge from-layer="4261" from-port="1" to-layer="4262" to-port="1" /> <edge from-layer="4262" from-port="2" to-layer="4264" to-port="0" /> <edge from-layer="4263" from-port="0" to-layer="4264" to-port="1" /> <edge from-layer="4264" from-port="2" to-layer="4265" to-port="1" /> <edge from-layer="4265" from-port="2" to-layer="4267" to-port="0" /> <edge from-layer="4266" from-port="0" to-layer="4267" to-port="1" /> <edge from-layer="4267" from-port="2" to-layer="4269" to-port="0" /> <edge from-layer="4268" from-port="0" to-layer="4269" to-port="1" /> <edge from-layer="4269" from-port="2" to-layer="4391" to-port="0" /> <edge from-layer="4269" from-port="2" to-layer="4273" to-port="0" /> <edge from-layer="4269" from-port="2" to-layer="4285" to-port="0" /> <edge from-layer="4270" from-port="0" to-layer="4273" to-port="1" /> <edge from-layer="4271" from-port="0" to-layer="4273" to-port="2" /> <edge from-layer="4272" from-port="0" to-layer="4273" to-port="3" /> <edge from-layer="4273" from-port="4" to-layer="4275" to-port="0" /> <edge from-layer="4274" from-port="0" to-layer="4275" to-port="1" /> <edge from-layer="4275" from-port="2" to-layer="4279" to-port="0" /> <edge from-layer="4275" from-port="2" to-layer="4344" to-port="0" /> <edge from-layer="4276" from-port="0" to-layer="4279" to-port="1" /> <edge from-layer="4277" from-port="0" to-layer="4279" to-port="2" /> <edge from-layer="4278" from-port="0" to-layer="4279" to-port="3" /> <edge from-layer="4279" from-port="4" to-layer="4325" to-port="0" /> <edge from-layer="4279" from-port="4" to-layer="4330" to-port="0" /> <edge from-layer="4279" from-port="4" to-layer="4314" to-port="0" /> <edge from-layer="4279" from-port="4" to-layer="4311" to-port="0" /> <edge from-layer="4280" from-port="0" to-layer="4307" to-port="0" /> <edge from-layer="4281" from-port="0" to-layer="4307" to-port="1" /> <edge from-layer="4282" from-port="0" to-layer="4285" to-port="1" /> <edge from-layer="4283" from-port="0" to-layer="4285" to-port="2" /> <edge from-layer="4284" from-port="0" to-layer="4285" to-port="3" /> <edge from-layer="4285" from-port="4" to-layer="4287" to-port="0" /> <edge from-layer="4286" from-port="0" to-layer="4287" to-port="1" /> <edge from-layer="4287" from-port="2" to-layer="4349" to-port="0" /> <edge from-layer="4287" from-port="2" to-layer="4288" to-port="0" /> <edge from-layer="4287" from-port="2" to-layer="4376" to-port="0" /> <edge from-layer="4288" from-port="1" to-layer="4290" to-port="0" /> <edge from-layer="4289" from-port="0" to-layer="4290" to-port="2" /> <edge from-layer="4290" from-port="3" to-layer="4303" to-port="0" /> <edge from-layer="4291" from-port="0" to-layer="4296" to-port="0" /> <edge from-layer="4292" from-port="0" to-layer="4295" to-port="1" /> <edge from-layer="4293" from-port="0" to-layer="4295" to-port="2" /> <edge from-layer="4294" from-port="0" to-layer="4295" to-port="3" /> <edge from-layer="4295" from-port="4" to-layer="4296" to-port="1" /> <edge from-layer="4296" from-port="2" to-layer="4297" to-port="0" /> <edge from-layer="4297" from-port="1" to-layer="4299" to-port="0" /> <edge from-layer="4298" from-port="0" to-layer="4299" to-port="2" /> <edge from-layer="4299" from-port="3" to-layer="4300" to-port="0" /> <edge from-layer="4299" from-port="3" to-layer="4378" to-port="0" /> <edge from-layer="4300" from-port="1" to-layer="4302" to-port="0" /> <edge from-layer="4301" from-port="0" to-layer="4302" to-port="2" /> <edge from-layer="4302" from-port="3" to-layer="4303" to-port="1" /> <edge from-layer="4303" from-port="2" to-layer="4305" to-port="0" /> <edge from-layer="4304" from-port="0" to-layer="4305" to-port="1" /> <edge from-layer="4305" from-port="2" to-layer="4335" to-port="2" /> <edge from-layer="4305" from-port="2" to-layer="4307" to-port="2" /> <edge from-layer="4306" from-port="0" to-layer="4307" to-port="3" /> <edge from-layer="4307" from-port="4" to-layer="4309" to-port="0" /> <edge from-layer="4308" from-port="0" to-layer="4309" to-port="2" /> <edge from-layer="4309" from-port="3" to-layer="4310" to-port="0" /> <edge from-layer="4310" from-port="2" to-layer="4350" to-port="1" /> <edge from-layer="4310" from-port="2" to-layer="4311" to-port="1" /> <edge from-layer="4311" from-port="2" to-layer="4340" to-port="0" /> <edge from-layer="4312" from-port="0" to-layer="4322" to-port="0" /> <edge from-layer="4313" from-port="0" to-layer="4322" to-port="1" /> <edge from-layer="4314" from-port="1" to-layer="4316" to-port="0" /> <edge from-layer="4315" from-port="0" to-layer="4316" to-port="2" /> <edge from-layer="4316" from-port="3" to-layer="4317" to-port="0" /> <edge from-layer="4317" from-port="2" to-layer="4318" to-port="0" /> <edge from-layer="4318" from-port="1" to-layer="4320" to-port="0" /> <edge from-layer="4319" from-port="0" to-layer="4320" to-port="1" /> <edge from-layer="4320" from-port="2" to-layer="4322" to-port="2" /> <edge from-layer="4321" from-port="0" to-layer="4322" to-port="3" /> <edge from-layer="4322" from-port="4" to-layer="4325" to-port="1" /> <edge from-layer="4322" from-port="4" to-layer="4330" to-port="2" /> <edge from-layer="4323" from-port="0" to-layer="4325" to-port="2" /> <edge from-layer="4324" from-port="0" to-layer="4325" to-port="3" /> <edge from-layer="4325" from-port="4" to-layer="4327" to-port="0" /> <edge from-layer="4326" from-port="0" to-layer="4327" to-port="1" /> <edge from-layer="4327" from-port="2" to-layer="4331" to-port="0" /> <edge from-layer="4328" from-port="0" to-layer="4330" to-port="1" /> <edge from-layer="4329" from-port="0" to-layer="4330" to-port="3" /> <edge from-layer="4330" from-port="4" to-layer="4331" to-port="1" /> <edge from-layer="4331" from-port="2" to-layer="4339" to-port="0" /> <edge from-layer="4332" from-port="0" to-layer="4335" to-port="0" /> <edge from-layer="4333" from-port="0" to-layer="4335" to-port="1" /> <edge from-layer="4334" from-port="0" to-layer="4335" to-port="3" /> <edge from-layer="4335" from-port="4" to-layer="4337" to-port="0" /> <edge from-layer="4336" from-port="0" to-layer="4337" to-port="2" /> <edge from-layer="4337" from-port="3" to-layer="4338" to-port="0" /> <edge from-layer="4338" from-port="2" to-layer="4371" to-port="1" /> <edge from-layer="4338" from-port="2" to-layer="4339" to-port="1" /> <edge from-layer="4339" from-port="2" to-layer="4340" to-port="1" /> <edge from-layer="4340" from-port="2" to-layer="4345" to-port="0" /> <edge from-layer="4341" from-port="0" to-layer="4344" to-port="1" /> <edge from-layer="4342" from-port="0" to-layer="4344" to-port="2" /> <edge from-layer="4343" from-port="0" to-layer="4344" to-port="3" /> <edge from-layer="4344" from-port="4" to-layer="4345" to-port="1" /> <edge from-layer="4345" from-port="2" to-layer="4439" to-port="0" /> <edge from-layer="4345" from-port="2" to-layer="4401" to-port="0" /> <edge from-layer="4346" from-port="0" to-layer="4349" to-port="1" /> <edge from-layer="4347" from-port="0" to-layer="4349" to-port="2" /> <edge from-layer="4348" from-port="0" to-layer="4349" to-port="3" /> <edge from-layer="4349" from-port="4" to-layer="4353" to-port="0" /> <edge from-layer="4349" from-port="4" to-layer="4364" to-port="0" /> <edge from-layer="4349" from-port="4" to-layer="4369" to-port="0" /> <edge from-layer="4349" from-port="4" to-layer="4350" to-port="0" /> <edge from-layer="4350" from-port="2" to-layer="4372" to-port="0" /> <edge from-layer="4351" from-port="0" to-layer="4361" to-port="0" /> <edge from-layer="4352" from-port="0" to-layer="4361" to-port="1" /> <edge from-layer="4353" from-port="1" to-layer="4355" to-port="0" /> <edge from-layer="4354" from-port="0" to-layer="4355" to-port="2" /> <edge from-layer="4355" from-port="3" to-layer="4356" to-port="0" /> <edge from-layer="4356" from-port="2" to-layer="4357" to-port="0" /> <edge from-layer="4357" from-port="1" to-layer="4359" to-port="0" /> <edge from-layer="4358" from-port="0" to-layer="4359" to-port="1" /> <edge from-layer="4359" from-port="2" to-layer="4361" to-port="2" /> <edge from-layer="4360" from-port="0" to-layer="4361" to-port="3" /> <edge from-layer="4361" from-port="4" to-layer="4369" to-port="2" /> <edge from-layer="4361" from-port="4" to-layer="4364" to-port="1" /> <edge from-layer="4362" from-port="0" to-layer="4364" to-port="2" /> <edge from-layer="4363" from-port="0" to-layer="4364" to-port="3" /> <edge from-layer="4364" from-port="4" to-layer="4366" to-port="0" /> <edge from-layer="4365" from-port="0" to-layer="4366" to-port="1" /> <edge from-layer="4366" from-port="2" to-layer="4370" to-port="0" /> <edge from-layer="4367" from-port="0" to-layer="4369" to-port="1" /> <edge from-layer="4368" from-port="0" to-layer="4369" to-port="3" /> <edge from-layer="4369" from-port="4" to-layer="4370" to-port="1" /> <edge from-layer="4370" from-port="2" to-layer="4371" to-port="0" /> <edge from-layer="4371" from-port="2" to-layer="4372" to-port="1" /> <edge from-layer="4372" from-port="2" to-layer="4377" to-port="0" /> <edge from-layer="4373" from-port="0" to-layer="4376" to-port="1" /> <edge from-layer="4374" from-port="0" to-layer="4376" to-port="2" /> <edge from-layer="4375" from-port="0" to-layer="4376" to-port="3" /> <edge from-layer="4376" from-port="4" to-layer="4377" to-port="1" /> <edge from-layer="4377" from-port="2" to-layer="4378" to-port="1" /> <edge from-layer="4378" from-port="2" to-layer="10153" to-port="0" /> <edge from-layer="4378" from-port="2" to-layer="4439" to-port="1" /> <edge from-layer="4378" from-port="2" to-layer="4398" to-port="0" /> <edge from-layer="4379" from-port="0" to-layer="4384" to-port="0" /> <edge from-layer="4380" from-port="0" to-layer="4383" to-port="1" /> <edge from-layer="4381" from-port="0" to-layer="4383" to-port="2" /> <edge from-layer="4382" from-port="0" to-layer="4383" to-port="3" /> <edge from-layer="4383" from-port="4" to-layer="4384" to-port="1" /> <edge from-layer="4384" from-port="2" to-layer="4385" to-port="0" /> <edge from-layer="4385" from-port="1" to-layer="4387" to-port="0" /> <edge from-layer="4386" from-port="0" to-layer="4387" to-port="2" /> <edge from-layer="4387" from-port="3" to-layer="4394" to-port="0" /> <edge from-layer="4388" from-port="0" to-layer="4391" to-port="1" /> <edge from-layer="4389" from-port="0" to-layer="4391" to-port="2" /> <edge from-layer="4390" from-port="0" to-layer="4391" to-port="3" /> <edge from-layer="4391" from-port="4" to-layer="4393" to-port="0" /> <edge from-layer="4392" from-port="0" to-layer="4393" to-port="1" /> <edge from-layer="4393" from-port="2" to-layer="4394" to-port="1" /> <edge from-layer="4394" from-port="2" to-layer="4439" to-port="2" /> <edge from-layer="4394" from-port="2" to-layer="10152" to-port="0" /> <edge from-layer="4395" from-port="0" to-layer="4413" to-port="0" /> <edge from-layer="4396" from-port="0" to-layer="4408" to-port="0" /> <edge from-layer="4397" from-port="0" to-layer="4408" to-port="1" /> <edge from-layer="4397" from-port="0" to-layer="4411" to-port="1" /> <edge from-layer="4398" from-port="1" to-layer="4400" to-port="0" /> <edge from-layer="4399" from-port="0" to-layer="4400" to-port="2" /> <edge from-layer="4400" from-port="3" to-layer="4404" to-port="0" /> <edge from-layer="4400" from-port="3" to-layer="4410" to-port="0" /> <edge from-layer="4401" from-port="1" to-layer="4428" to-port="0" /> <edge from-layer="4401" from-port="1" to-layer="4403" to-port="0" /> <edge from-layer="4402" from-port="0" to-layer="4403" to-port="2" /> <edge from-layer="4403" from-port="3" to-layer="4404" to-port="1" /> <edge from-layer="4404" from-port="2" to-layer="4406" to-port="0" /> <edge from-layer="4405" from-port="0" to-layer="4406" to-port="1" /> <edge from-layer="4405" from-port="0" to-layer="4410" to-port="1" /> <edge from-layer="4406" from-port="2" to-layer="4408" to-port="2" /> <edge from-layer="4407" from-port="0" to-layer="4408" to-port="3" /> <edge from-layer="4407" from-port="0" to-layer="4411" to-port="3" /> <edge from-layer="4408" from-port="4" to-layer="4413" to-port="1" /> <edge from-layer="4409" from-port="0" to-layer="4411" to-port="0" /> <edge from-layer="4410" from-port="2" to-layer="4411" to-port="2" /> <edge from-layer="4410" from-port="2" to-layer="4418" to-port="2" /> <edge from-layer="4411" from-port="4" to-layer="4413" to-port="2" /> <edge from-layer="4412" from-port="0" to-layer="4413" to-port="3" /> <edge from-layer="4413" from-port="4" to-layer="4420" to-port="0" /> <edge from-layer="4414" from-port="0" to-layer="4420" to-port="1" /> <edge from-layer="4415" from-port="0" to-layer="4418" to-port="0" /> <edge from-layer="4416" from-port="0" to-layer="4418" to-port="1" /> <edge from-layer="4417" from-port="0" to-layer="4418" to-port="3" /> <edge from-layer="4418" from-port="4" to-layer="4420" to-port="2" /> <edge from-layer="4419" from-port="0" to-layer="4420" to-port="3" /> <edge from-layer="4420" from-port="4" to-layer="4423" to-port="0" /> <edge from-layer="4421" from-port="0" to-layer="4423" to-port="1" /> <edge from-layer="4422" from-port="0" to-layer="4423" to-port="2" /> <edge from-layer="4423" from-port="3" to-layer="4424" to-port="0" /> <edge from-layer="4424" from-port="2" to-layer="4425" to-port="0" /> <edge from-layer="4424" from-port="2" to-layer="4438" to-port="0" /> <edge from-layer="4425" from-port="1" to-layer="4437" to-port="0" /> <edge from-layer="4426" from-port="0" to-layer="4428" to-port="1" /> <edge from-layer="4427" from-port="0" to-layer="4428" to-port="2" /> <edge from-layer="4428" from-port="3" to-layer="4432" to-port="0" /> <edge from-layer="4429" from-port="0" to-layer="4432" to-port="1" /> <edge from-layer="4430" from-port="0" to-layer="4432" to-port="2" /> <edge from-layer="4431" from-port="0" to-layer="4432" to-port="3" /> <edge from-layer="4432" from-port="4" to-layer="4434" to-port="0" /> <edge from-layer="4432" from-port="4" to-layer="4436" to-port="2" /> <edge from-layer="4433" from-port="0" to-layer="4434" to-port="1" /> <edge from-layer="4434" from-port="2" to-layer="4436" to-port="0" /> <edge from-layer="4435" from-port="0" to-layer="4436" to-port="1" /> <edge from-layer="4436" from-port="3" to-layer="4437" to-port="1" /> <edge from-layer="4437" from-port="2" to-layer="4438" to-port="1" /> <edge from-layer="4438" from-port="2" to-layer="4439" to-port="3" /> <edge from-layer="4439" from-port="4" to-layer="4441" to-port="0" /> <edge from-layer="4440" from-port="0" to-layer="4441" to-port="1" /> <edge from-layer="4441" from-port="2" to-layer="4443" to-port="0" /> <edge from-layer="4442" from-port="0" to-layer="4443" to-port="1" /> <edge from-layer="4443" from-port="2" to-layer="4451" to-port="0" /> <edge from-layer="4444" from-port="0" to-layer="4445" to-port="0" /> <edge from-layer="4445" from-port="1" to-layer="4448" to-port="0" /> <edge from-layer="4446" from-port="0" to-layer="4447" to-port="0" /> <edge from-layer="4447" from-port="1" to-layer="4448" to-port="1" /> <edge from-layer="4448" from-port="2" to-layer="4450" to-port="0" /> <edge from-layer="4449" from-port="0" to-layer="4450" to-port="1" /> <edge from-layer="4450" from-port="2" to-layer="4451" to-port="1" /> <edge from-layer="4451" from-port="2" to-layer="4453" to-port="0" /> <edge from-layer="4452" from-port="0" to-layer="4453" to-port="1" /> <edge from-layer="4453" from-port="2" to-layer="4454" to-port="1" /> <edge from-layer="4454" from-port="2" to-layer="4455" to-port="0" /> <edge from-layer="4455" from-port="2" to-layer="4686" to-port="1" /> <edge from-layer="4455" from-port="2" to-layer="4457" to-port="0" /> <edge from-layer="4455" from-port="2" to-layer="4484" to-port="0" /> <edge from-layer="4456" from-port="0" to-layer="4457" to-port="1" /> <edge from-layer="4457" from-port="2" to-layer="4459" to-port="0" /> <edge from-layer="4458" from-port="0" to-layer="4459" to-port="1" /> <edge from-layer="4459" from-port="2" to-layer="4461" to-port="0" /> <edge from-layer="4460" from-port="0" to-layer="4461" to-port="1" /> <edge from-layer="4461" from-port="2" to-layer="4469" to-port="0" /> <edge from-layer="4462" from-port="0" to-layer="4463" to-port="0" /> <edge from-layer="4463" from-port="1" to-layer="4466" to-port="0" /> <edge from-layer="4464" from-port="0" to-layer="4465" to-port="0" /> <edge from-layer="4465" from-port="1" to-layer="4466" to-port="1" /> <edge from-layer="4466" from-port="2" to-layer="4468" to-port="0" /> <edge from-layer="4467" from-port="0" to-layer="4468" to-port="1" /> <edge from-layer="4468" from-port="2" to-layer="4469" to-port="1" /> <edge from-layer="4469" from-port="2" to-layer="4471" to-port="0" /> <edge from-layer="4470" from-port="0" to-layer="4471" to-port="1" /> <edge from-layer="4471" from-port="2" to-layer="4472" to-port="0" /> <edge from-layer="4472" from-port="1" to-layer="4480" to-port="0" /> <edge from-layer="4473" from-port="0" to-layer="4474" to-port="0" /> <edge from-layer="4474" from-port="1" to-layer="4477" to-port="0" /> <edge from-layer="4475" from-port="0" to-layer="4476" to-port="0" /> <edge from-layer="4476" from-port="1" to-layer="4477" to-port="1" /> <edge from-layer="4477" from-port="2" to-layer="4479" to-port="0" /> <edge from-layer="4478" from-port="0" to-layer="4479" to-port="1" /> <edge from-layer="4479" from-port="2" to-layer="4480" to-port="1" /> <edge from-layer="4480" from-port="2" to-layer="4482" to-port="0" /> <edge from-layer="4481" from-port="0" to-layer="4482" to-port="1" /> <edge from-layer="4482" from-port="2" to-layer="4685" to-port="0" /> <edge from-layer="4483" from-port="0" to-layer="4484" to-port="1" /> <edge from-layer="4484" from-port="2" to-layer="4486" to-port="0" /> <edge from-layer="4485" from-port="0" to-layer="4486" to-port="1" /> <edge from-layer="4486" from-port="2" to-layer="4488" to-port="0" /> <edge from-layer="4487" from-port="0" to-layer="4488" to-port="1" /> <edge from-layer="4488" from-port="2" to-layer="4496" to-port="0" /> <edge from-layer="4489" from-port="0" to-layer="4490" to-port="0" /> <edge from-layer="4490" from-port="1" to-layer="4493" to-port="0" /> <edge from-layer="4491" from-port="0" to-layer="4492" to-port="0" /> <edge from-layer="4492" from-port="1" to-layer="4493" to-port="1" /> <edge from-layer="4493" from-port="2" to-layer="4495" to-port="0" /> <edge from-layer="4494" from-port="0" to-layer="4495" to-port="1" /> <edge from-layer="4495" from-port="2" to-layer="4496" to-port="1" /> <edge from-layer="4496" from-port="2" to-layer="4498" to-port="0" /> <edge from-layer="4497" from-port="0" to-layer="4498" to-port="1" /> <edge from-layer="4498" from-port="2" to-layer="4500" to-port="0" /> <edge from-layer="4499" from-port="0" to-layer="4500" to-port="1" /> <edge from-layer="4500" from-port="2" to-layer="4504" to-port="0" /> <edge from-layer="4500" from-port="2" to-layer="4622" to-port="0" /> <edge from-layer="4500" from-port="2" to-layer="4516" to-port="0" /> <edge from-layer="4501" from-port="0" to-layer="4504" to-port="1" /> <edge from-layer="4502" from-port="0" to-layer="4504" to-port="2" /> <edge from-layer="4503" from-port="0" to-layer="4504" to-port="3" /> <edge from-layer="4504" from-port="4" to-layer="4506" to-port="0" /> <edge from-layer="4505" from-port="0" to-layer="4506" to-port="1" /> <edge from-layer="4506" from-port="2" to-layer="4575" to-port="0" /> <edge from-layer="4506" from-port="2" to-layer="4510" to-port="0" /> <edge from-layer="4507" from-port="0" to-layer="4510" to-port="1" /> <edge from-layer="4508" from-port="0" to-layer="4510" to-port="2" /> <edge from-layer="4509" from-port="0" to-layer="4510" to-port="3" /> <edge from-layer="4510" from-port="4" to-layer="4561" to-port="0" /> <edge from-layer="4510" from-port="4" to-layer="4556" to-port="0" /> <edge from-layer="4510" from-port="4" to-layer="4545" to-port="0" /> <edge from-layer="4510" from-port="4" to-layer="4542" to-port="0" /> <edge from-layer="4511" from-port="0" to-layer="4538" to-port="0" /> <edge from-layer="4512" from-port="0" to-layer="4538" to-port="1" /> <edge from-layer="4513" from-port="0" to-layer="4516" to-port="1" /> <edge from-layer="4514" from-port="0" to-layer="4516" to-port="2" /> <edge from-layer="4515" from-port="0" to-layer="4516" to-port="3" /> <edge from-layer="4516" from-port="4" to-layer="4518" to-port="0" /> <edge from-layer="4517" from-port="0" to-layer="4518" to-port="1" /> <edge from-layer="4518" from-port="2" to-layer="4580" to-port="0" /> <edge from-layer="4518" from-port="2" to-layer="4607" to-port="0" /> <edge from-layer="4518" from-port="2" to-layer="4519" to-port="0" /> <edge from-layer="4519" from-port="1" to-layer="4521" to-port="0" /> <edge from-layer="4520" from-port="0" to-layer="4521" to-port="2" /> <edge from-layer="4521" from-port="3" to-layer="4534" to-port="0" /> <edge from-layer="4522" from-port="0" to-layer="4527" to-port="0" /> <edge from-layer="4523" from-port="0" to-layer="4526" to-port="1" /> <edge from-layer="4524" from-port="0" to-layer="4526" to-port="2" /> <edge from-layer="4525" from-port="0" to-layer="4526" to-port="3" /> <edge from-layer="4526" from-port="4" to-layer="4527" to-port="1" /> <edge from-layer="4527" from-port="2" to-layer="4528" to-port="0" /> <edge from-layer="4528" from-port="1" to-layer="4530" to-port="0" /> <edge from-layer="4529" from-port="0" to-layer="4530" to-port="2" /> <edge from-layer="4530" from-port="3" to-layer="4609" to-port="0" /> <edge from-layer="4530" from-port="3" to-layer="4531" to-port="0" /> <edge from-layer="4531" from-port="1" to-layer="4533" to-port="0" /> <edge from-layer="4532" from-port="0" to-layer="4533" to-port="2" /> <edge from-layer="4533" from-port="3" to-layer="4534" to-port="1" /> <edge from-layer="4534" from-port="2" to-layer="4536" to-port="0" /> <edge from-layer="4535" from-port="0" to-layer="4536" to-port="1" /> <edge from-layer="4536" from-port="2" to-layer="4538" to-port="2" /> <edge from-layer="4536" from-port="2" to-layer="4566" to-port="2" /> <edge from-layer="4537" from-port="0" to-layer="4538" to-port="3" /> <edge from-layer="4538" from-port="4" to-layer="4540" to-port="0" /> <edge from-layer="4539" from-port="0" to-layer="4540" to-port="2" /> <edge from-layer="4540" from-port="3" to-layer="4541" to-port="0" /> <edge from-layer="4541" from-port="2" to-layer="4581" to-port="1" /> <edge from-layer="4541" from-port="2" to-layer="4542" to-port="1" /> <edge from-layer="4542" from-port="2" to-layer="4571" to-port="0" /> <edge from-layer="4543" from-port="0" to-layer="4553" to-port="0" /> <edge from-layer="4544" from-port="0" to-layer="4553" to-port="1" /> <edge from-layer="4545" from-port="1" to-layer="4547" to-port="0" /> <edge from-layer="4546" from-port="0" to-layer="4547" to-port="2" /> <edge from-layer="4547" from-port="3" to-layer="4548" to-port="0" /> <edge from-layer="4548" from-port="2" to-layer="4549" to-port="0" /> <edge from-layer="4549" from-port="1" to-layer="4551" to-port="0" /> <edge from-layer="4550" from-port="0" to-layer="4551" to-port="1" /> <edge from-layer="4551" from-port="2" to-layer="4553" to-port="2" /> <edge from-layer="4552" from-port="0" to-layer="4553" to-port="3" /> <edge from-layer="4553" from-port="4" to-layer="4561" to-port="2" /> <edge from-layer="4553" from-port="4" to-layer="4556" to-port="1" /> <edge from-layer="4554" from-port="0" to-layer="4556" to-port="2" /> <edge from-layer="4555" from-port="0" to-layer="4556" to-port="3" /> <edge from-layer="4556" from-port="4" to-layer="4558" to-port="0" /> <edge from-layer="4557" from-port="0" to-layer="4558" to-port="1" /> <edge from-layer="4558" from-port="2" to-layer="4562" to-port="0" /> <edge from-layer="4559" from-port="0" to-layer="4561" to-port="1" /> <edge from-layer="4560" from-port="0" to-layer="4561" to-port="3" /> <edge from-layer="4561" from-port="4" to-layer="4562" to-port="1" /> <edge from-layer="4562" from-port="2" to-layer="4570" to-port="0" /> <edge from-layer="4563" from-port="0" to-layer="4566" to-port="0" /> <edge from-layer="4564" from-port="0" to-layer="4566" to-port="1" /> <edge from-layer="4565" from-port="0" to-layer="4566" to-port="3" /> <edge from-layer="4566" from-port="4" to-layer="4568" to-port="0" /> <edge from-layer="4567" from-port="0" to-layer="4568" to-port="2" /> <edge from-layer="4568" from-port="3" to-layer="4569" to-port="0" /> <edge from-layer="4569" from-port="2" to-layer="4602" to-port="1" /> <edge from-layer="4569" from-port="2" to-layer="4570" to-port="1" /> <edge from-layer="4570" from-port="2" to-layer="4571" to-port="1" /> <edge from-layer="4571" from-port="2" to-layer="4576" to-port="0" /> <edge from-layer="4572" from-port="0" to-layer="4575" to-port="1" /> <edge from-layer="4573" from-port="0" to-layer="4575" to-port="2" /> <edge from-layer="4574" from-port="0" to-layer="4575" to-port="3" /> <edge from-layer="4575" from-port="4" to-layer="4576" to-port="1" /> <edge from-layer="4576" from-port="2" to-layer="4632" to-port="0" /> <edge from-layer="4576" from-port="2" to-layer="4670" to-port="0" /> <edge from-layer="4577" from-port="0" to-layer="4580" to-port="1" /> <edge from-layer="4578" from-port="0" to-layer="4580" to-port="2" /> <edge from-layer="4579" from-port="0" to-layer="4580" to-port="3" /> <edge from-layer="4580" from-port="4" to-layer="4600" to-port="0" /> <edge from-layer="4580" from-port="4" to-layer="4581" to-port="0" /> <edge from-layer="4580" from-port="4" to-layer="4595" to-port="0" /> <edge from-layer="4580" from-port="4" to-layer="4584" to-port="0" /> <edge from-layer="4581" from-port="2" to-layer="4603" to-port="0" /> <edge from-layer="4582" from-port="0" to-layer="4592" to-port="0" /> <edge from-layer="4583" from-port="0" to-layer="4592" to-port="1" /> <edge from-layer="4584" from-port="1" to-layer="4586" to-port="0" /> <edge from-layer="4585" from-port="0" to-layer="4586" to-port="2" /> <edge from-layer="4586" from-port="3" to-layer="4587" to-port="0" /> <edge from-layer="4587" from-port="2" to-layer="4588" to-port="0" /> <edge from-layer="4588" from-port="1" to-layer="4590" to-port="0" /> <edge from-layer="4589" from-port="0" to-layer="4590" to-port="1" /> <edge from-layer="4590" from-port="2" to-layer="4592" to-port="2" /> <edge from-layer="4591" from-port="0" to-layer="4592" to-port="3" /> <edge from-layer="4592" from-port="4" to-layer="4595" to-port="1" /> <edge from-layer="4592" from-port="4" to-layer="4600" to-port="2" /> <edge from-layer="4593" from-port="0" to-layer="4595" to-port="2" /> <edge from-layer="4594" from-port="0" to-layer="4595" to-port="3" /> <edge from-layer="4595" from-port="4" to-layer="4597" to-port="0" /> <edge from-layer="4596" from-port="0" to-layer="4597" to-port="1" /> <edge from-layer="4597" from-port="2" to-layer="4601" to-port="0" /> <edge from-layer="4598" from-port="0" to-layer="4600" to-port="1" /> <edge from-layer="4599" from-port="0" to-layer="4600" to-port="3" /> <edge from-layer="4600" from-port="4" to-layer="4601" to-port="1" /> <edge from-layer="4601" from-port="2" to-layer="4602" to-port="0" /> <edge from-layer="4602" from-port="2" to-layer="4603" to-port="1" /> <edge from-layer="4603" from-port="2" to-layer="4608" to-port="0" /> <edge from-layer="4604" from-port="0" to-layer="4607" to-port="1" /> <edge from-layer="4605" from-port="0" to-layer="4607" to-port="2" /> <edge from-layer="4606" from-port="0" to-layer="4607" to-port="3" /> <edge from-layer="4607" from-port="4" to-layer="4608" to-port="1" /> <edge from-layer="4608" from-port="2" to-layer="4609" to-port="1" /> <edge from-layer="4609" from-port="2" to-layer="4629" to-port="0" /> <edge from-layer="4609" from-port="2" to-layer="4670" to-port="1" /> <edge from-layer="4609" from-port="2" to-layer="10151" to-port="0" /> <edge from-layer="4610" from-port="0" to-layer="4615" to-port="0" /> <edge from-layer="4611" from-port="0" to-layer="4614" to-port="1" /> <edge from-layer="4612" from-port="0" to-layer="4614" to-port="2" /> <edge from-layer="4613" from-port="0" to-layer="4614" to-port="3" /> <edge from-layer="4614" from-port="4" to-layer="4615" to-port="1" /> <edge from-layer="4615" from-port="2" to-layer="4616" to-port="0" /> <edge from-layer="4616" from-port="1" to-layer="4618" to-port="0" /> <edge from-layer="4617" from-port="0" to-layer="4618" to-port="2" /> <edge from-layer="4618" from-port="3" to-layer="4625" to-port="0" /> <edge from-layer="4619" from-port="0" to-layer="4622" to-port="1" /> <edge from-layer="4620" from-port="0" to-layer="4622" to-port="2" /> <edge from-layer="4621" from-port="0" to-layer="4622" to-port="3" /> <edge from-layer="4622" from-port="4" to-layer="4624" to-port="0" /> <edge from-layer="4623" from-port="0" to-layer="4624" to-port="1" /> <edge from-layer="4624" from-port="2" to-layer="4625" to-port="1" /> <edge from-layer="4625" from-port="2" to-layer="10150" to-port="0" /> <edge from-layer="4625" from-port="2" to-layer="4670" to-port="2" /> <edge from-layer="4626" from-port="0" to-layer="4644" to-port="0" /> <edge from-layer="4627" from-port="0" to-layer="4639" to-port="0" /> <edge from-layer="4628" from-port="0" to-layer="4642" to-port="1" /> <edge from-layer="4628" from-port="0" to-layer="4639" to-port="1" /> <edge from-layer="4629" from-port="1" to-layer="4631" to-port="0" /> <edge from-layer="4630" from-port="0" to-layer="4631" to-port="2" /> <edge from-layer="4631" from-port="3" to-layer="4635" to-port="0" /> <edge from-layer="4631" from-port="3" to-layer="4641" to-port="0" /> <edge from-layer="4632" from-port="1" to-layer="4659" to-port="0" /> <edge from-layer="4632" from-port="1" to-layer="4634" to-port="0" /> <edge from-layer="4633" from-port="0" to-layer="4634" to-port="2" /> <edge from-layer="4634" from-port="3" to-layer="4635" to-port="1" /> <edge from-layer="4635" from-port="2" to-layer="4637" to-port="0" /> <edge from-layer="4636" from-port="0" to-layer="4641" to-port="1" /> <edge from-layer="4636" from-port="0" to-layer="4637" to-port="1" /> <edge from-layer="4637" from-port="2" to-layer="4639" to-port="2" /> <edge from-layer="4638" from-port="0" to-layer="4639" to-port="3" /> <edge from-layer="4638" from-port="0" to-layer="4642" to-port="3" /> <edge from-layer="4639" from-port="4" to-layer="4644" to-port="1" /> <edge from-layer="4640" from-port="0" to-layer="4642" to-port="0" /> <edge from-layer="4641" from-port="2" to-layer="4642" to-port="2" /> <edge from-layer="4641" from-port="2" to-layer="4649" to-port="2" /> <edge from-layer="4642" from-port="4" to-layer="4644" to-port="2" /> <edge from-layer="4643" from-port="0" to-layer="4644" to-port="3" /> <edge from-layer="4644" from-port="4" to-layer="4651" to-port="0" /> <edge from-layer="4645" from-port="0" to-layer="4651" to-port="1" /> <edge from-layer="4646" from-port="0" to-layer="4649" to-port="0" /> <edge from-layer="4647" from-port="0" to-layer="4649" to-port="1" /> <edge from-layer="4648" from-port="0" to-layer="4649" to-port="3" /> <edge from-layer="4649" from-port="4" to-layer="4651" to-port="2" /> <edge from-layer="4650" from-port="0" to-layer="4651" to-port="3" /> <edge from-layer="4651" from-port="4" to-layer="4654" to-port="0" /> <edge from-layer="4652" from-port="0" to-layer="4654" to-port="1" /> <edge from-layer="4653" from-port="0" to-layer="4654" to-port="2" /> <edge from-layer="4654" from-port="3" to-layer="4655" to-port="0" /> <edge from-layer="4655" from-port="2" to-layer="4669" to-port="0" /> <edge from-layer="4655" from-port="2" to-layer="4656" to-port="0" /> <edge from-layer="4656" from-port="1" to-layer="4668" to-port="0" /> <edge from-layer="4657" from-port="0" to-layer="4659" to-port="1" /> <edge from-layer="4658" from-port="0" to-layer="4659" to-port="2" /> <edge from-layer="4659" from-port="3" to-layer="4663" to-port="0" /> <edge from-layer="4660" from-port="0" to-layer="4663" to-port="1" /> <edge from-layer="4661" from-port="0" to-layer="4663" to-port="2" /> <edge from-layer="4662" from-port="0" to-layer="4663" to-port="3" /> <edge from-layer="4663" from-port="4" to-layer="4665" to-port="0" /> <edge from-layer="4663" from-port="4" to-layer="4667" to-port="2" /> <edge from-layer="4664" from-port="0" to-layer="4665" to-port="1" /> <edge from-layer="4665" from-port="2" to-layer="4667" to-port="0" /> <edge from-layer="4666" from-port="0" to-layer="4667" to-port="1" /> <edge from-layer="4667" from-port="3" to-layer="4668" to-port="1" /> <edge from-layer="4668" from-port="2" to-layer="4669" to-port="1" /> <edge from-layer="4669" from-port="2" to-layer="4670" to-port="3" /> <edge from-layer="4670" from-port="4" to-layer="4672" to-port="0" /> <edge from-layer="4671" from-port="0" to-layer="4672" to-port="1" /> <edge from-layer="4672" from-port="2" to-layer="4674" to-port="0" /> <edge from-layer="4673" from-port="0" to-layer="4674" to-port="1" /> <edge from-layer="4674" from-port="2" to-layer="4682" to-port="0" /> <edge from-layer="4675" from-port="0" to-layer="4676" to-port="0" /> <edge from-layer="4676" from-port="1" to-layer="4679" to-port="0" /> <edge from-layer="4677" from-port="0" to-layer="4678" to-port="0" /> <edge from-layer="4678" from-port="1" to-layer="4679" to-port="1" /> <edge from-layer="4679" from-port="2" to-layer="4681" to-port="0" /> <edge from-layer="4680" from-port="0" to-layer="4681" to-port="1" /> <edge from-layer="4681" from-port="2" to-layer="4682" to-port="1" /> <edge from-layer="4682" from-port="2" to-layer="4684" to-port="0" /> <edge from-layer="4683" from-port="0" to-layer="4684" to-port="1" /> <edge from-layer="4684" from-port="2" to-layer="4685" to-port="1" /> <edge from-layer="4685" from-port="2" to-layer="4686" to-port="0" /> <edge from-layer="4686" from-port="2" to-layer="4917" to-port="1" /> <edge from-layer="4686" from-port="2" to-layer="4688" to-port="0" /> <edge from-layer="4686" from-port="2" to-layer="4715" to-port="0" /> <edge from-layer="4687" from-port="0" to-layer="4688" to-port="1" /> <edge from-layer="4688" from-port="2" to-layer="4690" to-port="0" /> <edge from-layer="4689" from-port="0" to-layer="4690" to-port="1" /> <edge from-layer="4690" from-port="2" to-layer="4692" to-port="0" /> <edge from-layer="4691" from-port="0" to-layer="4692" to-port="1" /> <edge from-layer="4692" from-port="2" to-layer="4700" to-port="0" /> <edge from-layer="4693" from-port="0" to-layer="4694" to-port="0" /> <edge from-layer="4694" from-port="1" to-layer="4697" to-port="0" /> <edge from-layer="4695" from-port="0" to-layer="4696" to-port="0" /> <edge from-layer="4696" from-port="1" to-layer="4697" to-port="1" /> <edge from-layer="4697" from-port="2" to-layer="4699" to-port="0" /> <edge from-layer="4698" from-port="0" to-layer="4699" to-port="1" /> <edge from-layer="4699" from-port="2" to-layer="4700" to-port="1" /> <edge from-layer="4700" from-port="2" to-layer="4702" to-port="0" /> <edge from-layer="4701" from-port="0" to-layer="4702" to-port="1" /> <edge from-layer="4702" from-port="2" to-layer="4703" to-port="0" /> <edge from-layer="4703" from-port="1" to-layer="4711" to-port="0" /> <edge from-layer="4704" from-port="0" to-layer="4705" to-port="0" /> <edge from-layer="4705" from-port="1" to-layer="4708" to-port="0" /> <edge from-layer="4706" from-port="0" to-layer="4707" to-port="0" /> <edge from-layer="4707" from-port="1" to-layer="4708" to-port="1" /> <edge from-layer="4708" from-port="2" to-layer="4710" to-port="0" /> <edge from-layer="4709" from-port="0" to-layer="4710" to-port="1" /> <edge from-layer="4710" from-port="2" to-layer="4711" to-port="1" /> <edge from-layer="4711" from-port="2" to-layer="4713" to-port="0" /> <edge from-layer="4712" from-port="0" to-layer="4713" to-port="1" /> <edge from-layer="4713" from-port="2" to-layer="4916" to-port="0" /> <edge from-layer="4714" from-port="0" to-layer="4715" to-port="1" /> <edge from-layer="4715" from-port="2" to-layer="4717" to-port="0" /> <edge from-layer="4716" from-port="0" to-layer="4717" to-port="1" /> <edge from-layer="4717" from-port="2" to-layer="4719" to-port="0" /> <edge from-layer="4718" from-port="0" to-layer="4719" to-port="1" /> <edge from-layer="4719" from-port="2" to-layer="4727" to-port="0" /> <edge from-layer="4720" from-port="0" to-layer="4721" to-port="0" /> <edge from-layer="4721" from-port="1" to-layer="4724" to-port="0" /> <edge from-layer="4722" from-port="0" to-layer="4723" to-port="0" /> <edge from-layer="4723" from-port="1" to-layer="4724" to-port="1" /> <edge from-layer="4724" from-port="2" to-layer="4726" to-port="0" /> <edge from-layer="4725" from-port="0" to-layer="4726" to-port="1" /> <edge from-layer="4726" from-port="2" to-layer="4727" to-port="1" /> <edge from-layer="4727" from-port="2" to-layer="4729" to-port="0" /> <edge from-layer="4728" from-port="0" to-layer="4729" to-port="1" /> <edge from-layer="4729" from-port="2" to-layer="4731" to-port="0" /> <edge from-layer="4730" from-port="0" to-layer="4731" to-port="1" /> <edge from-layer="4731" from-port="2" to-layer="4735" to-port="0" /> <edge from-layer="4731" from-port="2" to-layer="4747" to-port="0" /> <edge from-layer="4731" from-port="2" to-layer="4853" to-port="0" /> <edge from-layer="4732" from-port="0" to-layer="4735" to-port="1" /> <edge from-layer="4733" from-port="0" to-layer="4735" to-port="2" /> <edge from-layer="4734" from-port="0" to-layer="4735" to-port="3" /> <edge from-layer="4735" from-port="4" to-layer="4737" to-port="0" /> <edge from-layer="4736" from-port="0" to-layer="4737" to-port="1" /> <edge from-layer="4737" from-port="2" to-layer="4741" to-port="0" /> <edge from-layer="4737" from-port="2" to-layer="4806" to-port="0" /> <edge from-layer="4738" from-port="0" to-layer="4741" to-port="1" /> <edge from-layer="4739" from-port="0" to-layer="4741" to-port="2" /> <edge from-layer="4740" from-port="0" to-layer="4741" to-port="3" /> <edge from-layer="4741" from-port="4" to-layer="4776" to-port="0" /> <edge from-layer="4741" from-port="4" to-layer="4792" to-port="0" /> <edge from-layer="4741" from-port="4" to-layer="4773" to-port="0" /> <edge from-layer="4741" from-port="4" to-layer="4787" to-port="0" /> <edge from-layer="4742" from-port="0" to-layer="4769" to-port="0" /> <edge from-layer="4743" from-port="0" to-layer="4769" to-port="1" /> <edge from-layer="4744" from-port="0" to-layer="4747" to-port="1" /> <edge from-layer="4745" from-port="0" to-layer="4747" to-port="2" /> <edge from-layer="4746" from-port="0" to-layer="4747" to-port="3" /> <edge from-layer="4747" from-port="4" to-layer="4749" to-port="0" /> <edge from-layer="4748" from-port="0" to-layer="4749" to-port="1" /> <edge from-layer="4749" from-port="2" to-layer="4750" to-port="0" /> <edge from-layer="4749" from-port="2" to-layer="4838" to-port="0" /> <edge from-layer="4749" from-port="2" to-layer="4811" to-port="0" /> <edge from-layer="4750" from-port="1" to-layer="4752" to-port="0" /> <edge from-layer="4751" from-port="0" to-layer="4752" to-port="2" /> <edge from-layer="4752" from-port="3" to-layer="4765" to-port="0" /> <edge from-layer="4753" from-port="0" to-layer="4758" to-port="0" /> <edge from-layer="4754" from-port="0" to-layer="4757" to-port="1" /> <edge from-layer="4755" from-port="0" to-layer="4757" to-port="2" /> <edge from-layer="4756" from-port="0" to-layer="4757" to-port="3" /> <edge from-layer="4757" from-port="4" to-layer="4758" to-port="1" /> <edge from-layer="4758" from-port="2" to-layer="4759" to-port="0" /> <edge from-layer="4759" from-port="1" to-layer="4761" to-port="0" /> <edge from-layer="4760" from-port="0" to-layer="4761" to-port="2" /> <edge from-layer="4761" from-port="3" to-layer="4762" to-port="0" /> <edge from-layer="4761" from-port="3" to-layer="4840" to-port="0" /> <edge from-layer="4762" from-port="1" to-layer="4764" to-port="0" /> <edge from-layer="4763" from-port="0" to-layer="4764" to-port="2" /> <edge from-layer="4764" from-port="3" to-layer="4765" to-port="1" /> <edge from-layer="4765" from-port="2" to-layer="4767" to-port="0" /> <edge from-layer="4766" from-port="0" to-layer="4767" to-port="1" /> <edge from-layer="4767" from-port="2" to-layer="4797" to-port="2" /> <edge from-layer="4767" from-port="2" to-layer="4769" to-port="2" /> <edge from-layer="4768" from-port="0" to-layer="4769" to-port="3" /> <edge from-layer="4769" from-port="4" to-layer="4771" to-port="0" /> <edge from-layer="4770" from-port="0" to-layer="4771" to-port="2" /> <edge from-layer="4771" from-port="3" to-layer="4772" to-port="0" /> <edge from-layer="4772" from-port="2" to-layer="4773" to-port="1" /> <edge from-layer="4772" from-port="2" to-layer="4812" to-port="1" /> <edge from-layer="4773" from-port="2" to-layer="4802" to-port="0" /> <edge from-layer="4774" from-port="0" to-layer="4784" to-port="0" /> <edge from-layer="4775" from-port="0" to-layer="4784" to-port="1" /> <edge from-layer="4776" from-port="1" to-layer="4778" to-port="0" /> <edge from-layer="4777" from-port="0" to-layer="4778" to-port="2" /> <edge from-layer="4778" from-port="3" to-layer="4779" to-port="0" /> <edge from-layer="4779" from-port="2" to-layer="4780" to-port="0" /> <edge from-layer="4780" from-port="1" to-layer="4782" to-port="0" /> <edge from-layer="4781" from-port="0" to-layer="4782" to-port="1" /> <edge from-layer="4782" from-port="2" to-layer="4784" to-port="2" /> <edge from-layer="4783" from-port="0" to-layer="4784" to-port="3" /> <edge from-layer="4784" from-port="4" to-layer="4787" to-port="1" /> <edge from-layer="4784" from-port="4" to-layer="4792" to-port="2" /> <edge from-layer="4785" from-port="0" to-layer="4787" to-port="2" /> <edge from-layer="4786" from-port="0" to-layer="4787" to-port="3" /> <edge from-layer="4787" from-port="4" to-layer="4789" to-port="0" /> <edge from-layer="4788" from-port="0" to-layer="4789" to-port="1" /> <edge from-layer="4789" from-port="2" to-layer="4793" to-port="0" /> <edge from-layer="4790" from-port="0" to-layer="4792" to-port="1" /> <edge from-layer="4791" from-port="0" to-layer="4792" to-port="3" /> <edge from-layer="4792" from-port="4" to-layer="4793" to-port="1" /> <edge from-layer="4793" from-port="2" to-layer="4801" to-port="0" /> <edge from-layer="4794" from-port="0" to-layer="4797" to-port="0" /> <edge from-layer="4795" from-port="0" to-layer="4797" to-port="1" /> <edge from-layer="4796" from-port="0" to-layer="4797" to-port="3" /> <edge from-layer="4797" from-port="4" to-layer="4799" to-port="0" /> <edge from-layer="4798" from-port="0" to-layer="4799" to-port="2" /> <edge from-layer="4799" from-port="3" to-layer="4800" to-port="0" /> <edge from-layer="4800" from-port="2" to-layer="4801" to-port="1" /> <edge from-layer="4800" from-port="2" to-layer="4833" to-port="1" /> <edge from-layer="4801" from-port="2" to-layer="4802" to-port="1" /> <edge from-layer="4802" from-port="2" to-layer="4807" to-port="0" /> <edge from-layer="4803" from-port="0" to-layer="4806" to-port="1" /> <edge from-layer="4804" from-port="0" to-layer="4806" to-port="2" /> <edge from-layer="4805" from-port="0" to-layer="4806" to-port="3" /> <edge from-layer="4806" from-port="4" to-layer="4807" to-port="1" /> <edge from-layer="4807" from-port="2" to-layer="4901" to-port="0" /> <edge from-layer="4807" from-port="2" to-layer="4863" to-port="0" /> <edge from-layer="4808" from-port="0" to-layer="4811" to-port="1" /> <edge from-layer="4809" from-port="0" to-layer="4811" to-port="2" /> <edge from-layer="4810" from-port="0" to-layer="4811" to-port="3" /> <edge from-layer="4811" from-port="4" to-layer="4812" to-port="0" /> <edge from-layer="4811" from-port="4" to-layer="4815" to-port="0" /> <edge from-layer="4811" from-port="4" to-layer="4831" to-port="0" /> <edge from-layer="4811" from-port="4" to-layer="4826" to-port="0" /> <edge from-layer="4812" from-port="2" to-layer="4834" to-port="0" /> <edge from-layer="4813" from-port="0" to-layer="4823" to-port="0" /> <edge from-layer="4814" from-port="0" to-layer="4823" to-port="1" /> <edge from-layer="4815" from-port="1" to-layer="4817" to-port="0" /> <edge from-layer="4816" from-port="0" to-layer="4817" to-port="2" /> <edge from-layer="4817" from-port="3" to-layer="4818" to-port="0" /> <edge from-layer="4818" from-port="2" to-layer="4819" to-port="0" /> <edge from-layer="4819" from-port="1" to-layer="4821" to-port="0" /> <edge from-layer="4820" from-port="0" to-layer="4821" to-port="1" /> <edge from-layer="4821" from-port="2" to-layer="4823" to-port="2" /> <edge from-layer="4822" from-port="0" to-layer="4823" to-port="3" /> <edge from-layer="4823" from-port="4" to-layer="4826" to-port="1" /> <edge from-layer="4823" from-port="4" to-layer="4831" to-port="2" /> <edge from-layer="4824" from-port="0" to-layer="4826" to-port="2" /> <edge from-layer="4825" from-port="0" to-layer="4826" to-port="3" /> <edge from-layer="4826" from-port="4" to-layer="4828" to-port="0" /> <edge from-layer="4827" from-port="0" to-layer="4828" to-port="1" /> <edge from-layer="4828" from-port="2" to-layer="4832" to-port="0" /> <edge from-layer="4829" from-port="0" to-layer="4831" to-port="1" /> <edge from-layer="4830" from-port="0" to-layer="4831" to-port="3" /> <edge from-layer="4831" from-port="4" to-layer="4832" to-port="1" /> <edge from-layer="4832" from-port="2" to-layer="4833" to-port="0" /> <edge from-layer="4833" from-port="2" to-layer="4834" to-port="1" /> <edge from-layer="4834" from-port="2" to-layer="4839" to-port="0" /> <edge from-layer="4835" from-port="0" to-layer="4838" to-port="1" /> <edge from-layer="4836" from-port="0" to-layer="4838" to-port="2" /> <edge from-layer="4837" from-port="0" to-layer="4838" to-port="3" /> <edge from-layer="4838" from-port="4" to-layer="4839" to-port="1" /> <edge from-layer="4839" from-port="2" to-layer="4840" to-port="1" /> <edge from-layer="4840" from-port="2" to-layer="10147" to-port="0" /> <edge from-layer="4840" from-port="2" to-layer="4901" to-port="1" /> <edge from-layer="4840" from-port="2" to-layer="4860" to-port="0" /> <edge from-layer="4841" from-port="0" to-layer="4846" to-port="0" /> <edge from-layer="4842" from-port="0" to-layer="4845" to-port="1" /> <edge from-layer="4843" from-port="0" to-layer="4845" to-port="2" /> <edge from-layer="4844" from-port="0" to-layer="4845" to-port="3" /> <edge from-layer="4845" from-port="4" to-layer="4846" to-port="1" /> <edge from-layer="4846" from-port="2" to-layer="4847" to-port="0" /> <edge from-layer="4847" from-port="1" to-layer="4849" to-port="0" /> <edge from-layer="4848" from-port="0" to-layer="4849" to-port="2" /> <edge from-layer="4849" from-port="3" to-layer="4856" to-port="0" /> <edge from-layer="4850" from-port="0" to-layer="4853" to-port="1" /> <edge from-layer="4851" from-port="0" to-layer="4853" to-port="2" /> <edge from-layer="4852" from-port="0" to-layer="4853" to-port="3" /> <edge from-layer="4853" from-port="4" to-layer="4855" to-port="0" /> <edge from-layer="4854" from-port="0" to-layer="4855" to-port="1" /> <edge from-layer="4855" from-port="2" to-layer="4856" to-port="1" /> <edge from-layer="4856" from-port="2" to-layer="10146" to-port="0" /> <edge from-layer="4856" from-port="2" to-layer="4901" to-port="2" /> <edge from-layer="4857" from-port="0" to-layer="4875" to-port="0" /> <edge from-layer="4858" from-port="0" to-layer="4870" to-port="0" /> <edge from-layer="4859" from-port="0" to-layer="4873" to-port="1" /> <edge from-layer="4859" from-port="0" to-layer="4870" to-port="1" /> <edge from-layer="4860" from-port="1" to-layer="4862" to-port="0" /> <edge from-layer="4861" from-port="0" to-layer="4862" to-port="2" /> <edge from-layer="4862" from-port="3" to-layer="4872" to-port="0" /> <edge from-layer="4862" from-port="3" to-layer="4866" to-port="0" /> <edge from-layer="4863" from-port="1" to-layer="4890" to-port="0" /> <edge from-layer="4863" from-port="1" to-layer="4865" to-port="0" /> <edge from-layer="4864" from-port="0" to-layer="4865" to-port="2" /> <edge from-layer="4865" from-port="3" to-layer="4866" to-port="1" /> <edge from-layer="4866" from-port="2" to-layer="4868" to-port="0" /> <edge from-layer="4867" from-port="0" to-layer="4872" to-port="1" /> <edge from-layer="4867" from-port="0" to-layer="4868" to-port="1" /> <edge from-layer="4868" from-port="2" to-layer="4870" to-port="2" /> <edge from-layer="4869" from-port="0" to-layer="4873" to-port="3" /> <edge from-layer="4869" from-port="0" to-layer="4870" to-port="3" /> <edge from-layer="4870" from-port="4" to-layer="4875" to-port="1" /> <edge from-layer="4871" from-port="0" to-layer="4873" to-port="0" /> <edge from-layer="4872" from-port="2" to-layer="4873" to-port="2" /> <edge from-layer="4872" from-port="2" to-layer="4880" to-port="2" /> <edge from-layer="4873" from-port="4" to-layer="4875" to-port="2" /> <edge from-layer="4874" from-port="0" to-layer="4875" to-port="3" /> <edge from-layer="4875" from-port="4" to-layer="4882" to-port="0" /> <edge from-layer="4876" from-port="0" to-layer="4882" to-port="1" /> <edge from-layer="4877" from-port="0" to-layer="4880" to-port="0" /> <edge from-layer="4878" from-port="0" to-layer="4880" to-port="1" /> <edge from-layer="4879" from-port="0" to-layer="4880" to-port="3" /> <edge from-layer="4880" from-port="4" to-layer="4882" to-port="2" /> <edge from-layer="4881" from-port="0" to-layer="4882" to-port="3" /> <edge from-layer="4882" from-port="4" to-layer="4885" to-port="0" /> <edge from-layer="4883" from-port="0" to-layer="4885" to-port="1" /> <edge from-layer="4884" from-port="0" to-layer="4885" to-port="2" /> <edge from-layer="4885" from-port="3" to-layer="4886" to-port="0" /> <edge from-layer="4886" from-port="2" to-layer="4900" to-port="0" /> <edge from-layer="4886" from-port="2" to-layer="4887" to-port="0" /> <edge from-layer="4887" from-port="1" to-layer="4899" to-port="0" /> <edge from-layer="4888" from-port="0" to-layer="4890" to-port="1" /> <edge from-layer="4889" from-port="0" to-layer="4890" to-port="2" /> <edge from-layer="4890" from-port="3" to-layer="4894" to-port="0" /> <edge from-layer="4891" from-port="0" to-layer="4894" to-port="1" /> <edge from-layer="4892" from-port="0" to-layer="4894" to-port="2" /> <edge from-layer="4893" from-port="0" to-layer="4894" to-port="3" /> <edge from-layer="4894" from-port="4" to-layer="4896" to-port="0" /> <edge from-layer="4894" from-port="4" to-layer="4898" to-port="2" /> <edge from-layer="4895" from-port="0" to-layer="4896" to-port="1" /> <edge from-layer="4896" from-port="2" to-layer="4898" to-port="0" /> <edge from-layer="4897" from-port="0" to-layer="4898" to-port="1" /> <edge from-layer="4898" from-port="3" to-layer="4899" to-port="1" /> <edge from-layer="4899" from-port="2" to-layer="4900" to-port="1" /> <edge from-layer="4900" from-port="2" to-layer="4901" to-port="3" /> <edge from-layer="4901" from-port="4" to-layer="4903" to-port="0" /> <edge from-layer="4902" from-port="0" to-layer="4903" to-port="1" /> <edge from-layer="4903" from-port="2" to-layer="4905" to-port="0" /> <edge from-layer="4904" from-port="0" to-layer="4905" to-port="1" /> <edge from-layer="4905" from-port="2" to-layer="4913" to-port="0" /> <edge from-layer="4906" from-port="0" to-layer="4907" to-port="0" /> <edge from-layer="4907" from-port="1" to-layer="4910" to-port="0" /> <edge from-layer="4908" from-port="0" to-layer="4909" to-port="0" /> <edge from-layer="4909" from-port="1" to-layer="4910" to-port="1" /> <edge from-layer="4910" from-port="2" to-layer="4912" to-port="0" /> <edge from-layer="4911" from-port="0" to-layer="4912" to-port="1" /> <edge from-layer="4912" from-port="2" to-layer="4913" to-port="1" /> <edge from-layer="4913" from-port="2" to-layer="4915" to-port="0" /> <edge from-layer="4914" from-port="0" to-layer="4915" to-port="1" /> <edge from-layer="4915" from-port="2" to-layer="4916" to-port="1" /> <edge from-layer="4916" from-port="2" to-layer="4917" to-port="0" /> <edge from-layer="4917" from-port="2" to-layer="4946" to-port="0" /> <edge from-layer="4917" from-port="2" to-layer="4919" to-port="0" /> <edge from-layer="4917" from-port="2" to-layer="5148" to-port="1" /> <edge from-layer="4918" from-port="0" to-layer="4919" to-port="1" /> <edge from-layer="4919" from-port="2" to-layer="4921" to-port="0" /> <edge from-layer="4920" from-port="0" to-layer="4921" to-port="1" /> <edge from-layer="4921" from-port="2" to-layer="4923" to-port="0" /> <edge from-layer="4922" from-port="0" to-layer="4923" to-port="1" /> <edge from-layer="4923" from-port="2" to-layer="4931" to-port="0" /> <edge from-layer="4924" from-port="0" to-layer="4925" to-port="0" /> <edge from-layer="4925" from-port="1" to-layer="4928" to-port="0" /> <edge from-layer="4926" from-port="0" to-layer="4927" to-port="0" /> <edge from-layer="4927" from-port="1" to-layer="4928" to-port="1" /> <edge from-layer="4928" from-port="2" to-layer="4930" to-port="0" /> <edge from-layer="4929" from-port="0" to-layer="4930" to-port="1" /> <edge from-layer="4930" from-port="2" to-layer="4931" to-port="1" /> <edge from-layer="4931" from-port="2" to-layer="4933" to-port="0" /> <edge from-layer="4932" from-port="0" to-layer="4933" to-port="1" /> <edge from-layer="4933" from-port="2" to-layer="4934" to-port="0" /> <edge from-layer="4934" from-port="1" to-layer="4942" to-port="0" /> <edge from-layer="4935" from-port="0" to-layer="4936" to-port="0" /> <edge from-layer="4936" from-port="1" to-layer="4939" to-port="0" /> <edge from-layer="4937" from-port="0" to-layer="4938" to-port="0" /> <edge from-layer="4938" from-port="1" to-layer="4939" to-port="1" /> <edge from-layer="4939" from-port="2" to-layer="4941" to-port="0" /> <edge from-layer="4940" from-port="0" to-layer="4941" to-port="1" /> <edge from-layer="4941" from-port="2" to-layer="4942" to-port="1" /> <edge from-layer="4942" from-port="2" to-layer="4944" to-port="0" /> <edge from-layer="4943" from-port="0" to-layer="4944" to-port="1" /> <edge from-layer="4944" from-port="2" to-layer="5147" to-port="0" /> <edge from-layer="4945" from-port="0" to-layer="4946" to-port="1" /> <edge from-layer="4946" from-port="2" to-layer="4948" to-port="0" /> <edge from-layer="4947" from-port="0" to-layer="4948" to-port="1" /> <edge from-layer="4948" from-port="2" to-layer="4950" to-port="0" /> <edge from-layer="4949" from-port="0" to-layer="4950" to-port="1" /> <edge from-layer="4950" from-port="2" to-layer="4958" to-port="0" /> <edge from-layer="4951" from-port="0" to-layer="4952" to-port="0" /> <edge from-layer="4952" from-port="1" to-layer="4955" to-port="0" /> <edge from-layer="4953" from-port="0" to-layer="4954" to-port="0" /> <edge from-layer="4954" from-port="1" to-layer="4955" to-port="1" /> <edge from-layer="4955" from-port="2" to-layer="4957" to-port="0" /> <edge from-layer="4956" from-port="0" to-layer="4957" to-port="1" /> <edge from-layer="4957" from-port="2" to-layer="4958" to-port="1" /> <edge from-layer="4958" from-port="2" to-layer="4960" to-port="0" /> <edge from-layer="4959" from-port="0" to-layer="4960" to-port="1" /> <edge from-layer="4960" from-port="2" to-layer="4962" to-port="0" /> <edge from-layer="4961" from-port="0" to-layer="4962" to-port="1" /> <edge from-layer="4962" from-port="2" to-layer="4978" to-port="0" /> <edge from-layer="4962" from-port="2" to-layer="4966" to-port="0" /> <edge from-layer="4962" from-port="2" to-layer="5084" to-port="0" /> <edge from-layer="4963" from-port="0" to-layer="4966" to-port="1" /> <edge from-layer="4964" from-port="0" to-layer="4966" to-port="2" /> <edge from-layer="4965" from-port="0" to-layer="4966" to-port="3" /> <edge from-layer="4966" from-port="4" to-layer="4968" to-port="0" /> <edge from-layer="4967" from-port="0" to-layer="4968" to-port="1" /> <edge from-layer="4968" from-port="2" to-layer="4972" to-port="0" /> <edge from-layer="4968" from-port="2" to-layer="5037" to-port="0" /> <edge from-layer="4969" from-port="0" to-layer="4972" to-port="1" /> <edge from-layer="4970" from-port="0" to-layer="4972" to-port="2" /> <edge from-layer="4971" from-port="0" to-layer="4972" to-port="3" /> <edge from-layer="4972" from-port="4" to-layer="5004" to-port="0" /> <edge from-layer="4972" from-port="4" to-layer="5007" to-port="0" /> <edge from-layer="4972" from-port="4" to-layer="5018" to-port="0" /> <edge from-layer="4972" from-port="4" to-layer="5023" to-port="0" /> <edge from-layer="4973" from-port="0" to-layer="5000" to-port="0" /> <edge from-layer="4974" from-port="0" to-layer="5000" to-port="1" /> <edge from-layer="4975" from-port="0" to-layer="4978" to-port="1" /> <edge from-layer="4976" from-port="0" to-layer="4978" to-port="2" /> <edge from-layer="4977" from-port="0" to-layer="4978" to-port="3" /> <edge from-layer="4978" from-port="4" to-layer="4980" to-port="0" /> <edge from-layer="4979" from-port="0" to-layer="4980" to-port="1" /> <edge from-layer="4980" from-port="2" to-layer="5069" to-port="0" /> <edge from-layer="4980" from-port="2" to-layer="4981" to-port="0" /> <edge from-layer="4980" from-port="2" to-layer="5042" to-port="0" /> <edge from-layer="4981" from-port="1" to-layer="4983" to-port="0" /> <edge from-layer="4982" from-port="0" to-layer="4983" to-port="2" /> <edge from-layer="4983" from-port="3" to-layer="4996" to-port="0" /> <edge from-layer="4984" from-port="0" to-layer="4989" to-port="0" /> <edge from-layer="4985" from-port="0" to-layer="4988" to-port="1" /> <edge from-layer="4986" from-port="0" to-layer="4988" to-port="2" /> <edge from-layer="4987" from-port="0" to-layer="4988" to-port="3" /> <edge from-layer="4988" from-port="4" to-layer="4989" to-port="1" /> <edge from-layer="4989" from-port="2" to-layer="4990" to-port="0" /> <edge from-layer="4990" from-port="1" to-layer="4992" to-port="0" /> <edge from-layer="4991" from-port="0" to-layer="4992" to-port="2" /> <edge from-layer="4992" from-port="3" to-layer="4993" to-port="0" /> <edge from-layer="4992" from-port="3" to-layer="5071" to-port="0" /> <edge from-layer="4993" from-port="1" to-layer="4995" to-port="0" /> <edge from-layer="4994" from-port="0" to-layer="4995" to-port="2" /> <edge from-layer="4995" from-port="3" to-layer="4996" to-port="1" /> <edge from-layer="4996" from-port="2" to-layer="4998" to-port="0" /> <edge from-layer="4997" from-port="0" to-layer="4998" to-port="1" /> <edge from-layer="4998" from-port="2" to-layer="5000" to-port="2" /> <edge from-layer="4998" from-port="2" to-layer="5028" to-port="2" /> <edge from-layer="4999" from-port="0" to-layer="5000" to-port="3" /> <edge from-layer="5000" from-port="4" to-layer="5002" to-port="0" /> <edge from-layer="5001" from-port="0" to-layer="5002" to-port="2" /> <edge from-layer="5002" from-port="3" to-layer="5003" to-port="0" /> <edge from-layer="5003" from-port="2" to-layer="5004" to-port="1" /> <edge from-layer="5003" from-port="2" to-layer="5043" to-port="1" /> <edge from-layer="5004" from-port="2" to-layer="5033" to-port="0" /> <edge from-layer="5005" from-port="0" to-layer="5015" to-port="0" /> <edge from-layer="5006" from-port="0" to-layer="5015" to-port="1" /> <edge from-layer="5007" from-port="1" to-layer="5009" to-port="0" /> <edge from-layer="5008" from-port="0" to-layer="5009" to-port="2" /> <edge from-layer="5009" from-port="3" to-layer="5010" to-port="0" /> <edge from-layer="5010" from-port="2" to-layer="5011" to-port="0" /> <edge from-layer="5011" from-port="1" to-layer="5013" to-port="0" /> <edge from-layer="5012" from-port="0" to-layer="5013" to-port="1" /> <edge from-layer="5013" from-port="2" to-layer="5015" to-port="2" /> <edge from-layer="5014" from-port="0" to-layer="5015" to-port="3" /> <edge from-layer="5015" from-port="4" to-layer="5018" to-port="1" /> <edge from-layer="5015" from-port="4" to-layer="5023" to-port="2" /> <edge from-layer="5016" from-port="0" to-layer="5018" to-port="2" /> <edge from-layer="5017" from-port="0" to-layer="5018" to-port="3" /> <edge from-layer="5018" from-port="4" to-layer="5020" to-port="0" /> <edge from-layer="5019" from-port="0" to-layer="5020" to-port="1" /> <edge from-layer="5020" from-port="2" to-layer="5024" to-port="0" /> <edge from-layer="5021" from-port="0" to-layer="5023" to-port="1" /> <edge from-layer="5022" from-port="0" to-layer="5023" to-port="3" /> <edge from-layer="5023" from-port="4" to-layer="5024" to-port="1" /> <edge from-layer="5024" from-port="2" to-layer="5032" to-port="0" /> <edge from-layer="5025" from-port="0" to-layer="5028" to-port="0" /> <edge from-layer="5026" from-port="0" to-layer="5028" to-port="1" /> <edge from-layer="5027" from-port="0" to-layer="5028" to-port="3" /> <edge from-layer="5028" from-port="4" to-layer="5030" to-port="0" /> <edge from-layer="5029" from-port="0" to-layer="5030" to-port="2" /> <edge from-layer="5030" from-port="3" to-layer="5031" to-port="0" /> <edge from-layer="5031" from-port="2" to-layer="5032" to-port="1" /> <edge from-layer="5031" from-port="2" to-layer="5064" to-port="1" /> <edge from-layer="5032" from-port="2" to-layer="5033" to-port="1" /> <edge from-layer="5033" from-port="2" to-layer="5038" to-port="0" /> <edge from-layer="5034" from-port="0" to-layer="5037" to-port="1" /> <edge from-layer="5035" from-port="0" to-layer="5037" to-port="2" /> <edge from-layer="5036" from-port="0" to-layer="5037" to-port="3" /> <edge from-layer="5037" from-port="4" to-layer="5038" to-port="1" /> <edge from-layer="5038" from-port="2" to-layer="5132" to-port="0" /> <edge from-layer="5038" from-port="2" to-layer="5094" to-port="0" /> <edge from-layer="5039" from-port="0" to-layer="5042" to-port="1" /> <edge from-layer="5040" from-port="0" to-layer="5042" to-port="2" /> <edge from-layer="5041" from-port="0" to-layer="5042" to-port="3" /> <edge from-layer="5042" from-port="4" to-layer="5057" to-port="0" /> <edge from-layer="5042" from-port="4" to-layer="5046" to-port="0" /> <edge from-layer="5042" from-port="4" to-layer="5062" to-port="0" /> <edge from-layer="5042" from-port="4" to-layer="5043" to-port="0" /> <edge from-layer="5043" from-port="2" to-layer="5065" to-port="0" /> <edge from-layer="5044" from-port="0" to-layer="5054" to-port="0" /> <edge from-layer="5045" from-port="0" to-layer="5054" to-port="1" /> <edge from-layer="5046" from-port="1" to-layer="5048" to-port="0" /> <edge from-layer="5047" from-port="0" to-layer="5048" to-port="2" /> <edge from-layer="5048" from-port="3" to-layer="5049" to-port="0" /> <edge from-layer="5049" from-port="2" to-layer="5050" to-port="0" /> <edge from-layer="5050" from-port="1" to-layer="5052" to-port="0" /> <edge from-layer="5051" from-port="0" to-layer="5052" to-port="1" /> <edge from-layer="5052" from-port="2" to-layer="5054" to-port="2" /> <edge from-layer="5053" from-port="0" to-layer="5054" to-port="3" /> <edge from-layer="5054" from-port="4" to-layer="5062" to-port="2" /> <edge from-layer="5054" from-port="4" to-layer="5057" to-port="1" /> <edge from-layer="5055" from-port="0" to-layer="5057" to-port="2" /> <edge from-layer="5056" from-port="0" to-layer="5057" to-port="3" /> <edge from-layer="5057" from-port="4" to-layer="5059" to-port="0" /> <edge from-layer="5058" from-port="0" to-layer="5059" to-port="1" /> <edge from-layer="5059" from-port="2" to-layer="5063" to-port="0" /> <edge from-layer="5060" from-port="0" to-layer="5062" to-port="1" /> <edge from-layer="5061" from-port="0" to-layer="5062" to-port="3" /> <edge from-layer="5062" from-port="4" to-layer="5063" to-port="1" /> <edge from-layer="5063" from-port="2" to-layer="5064" to-port="0" /> <edge from-layer="5064" from-port="2" to-layer="5065" to-port="1" /> <edge from-layer="5065" from-port="2" to-layer="5070" to-port="0" /> <edge from-layer="5066" from-port="0" to-layer="5069" to-port="1" /> <edge from-layer="5067" from-port="0" to-layer="5069" to-port="2" /> <edge from-layer="5068" from-port="0" to-layer="5069" to-port="3" /> <edge from-layer="5069" from-port="4" to-layer="5070" to-port="1" /> <edge from-layer="5070" from-port="2" to-layer="5071" to-port="1" /> <edge from-layer="5071" from-port="2" to-layer="10145" to-port="0" /> <edge from-layer="5071" from-port="2" to-layer="5132" to-port="1" /> <edge from-layer="5071" from-port="2" to-layer="5091" to-port="0" /> <edge from-layer="5072" from-port="0" to-layer="5077" to-port="0" /> <edge from-layer="5073" from-port="0" to-layer="5076" to-port="1" /> <edge from-layer="5074" from-port="0" to-layer="5076" to-port="2" /> <edge from-layer="5075" from-port="0" to-layer="5076" to-port="3" /> <edge from-layer="5076" from-port="4" to-layer="5077" to-port="1" /> <edge from-layer="5077" from-port="2" to-layer="5078" to-port="0" /> <edge from-layer="5078" from-port="1" to-layer="5080" to-port="0" /> <edge from-layer="5079" from-port="0" to-layer="5080" to-port="2" /> <edge from-layer="5080" from-port="3" to-layer="5087" to-port="0" /> <edge from-layer="5081" from-port="0" to-layer="5084" to-port="1" /> <edge from-layer="5082" from-port="0" to-layer="5084" to-port="2" /> <edge from-layer="5083" from-port="0" to-layer="5084" to-port="3" /> <edge from-layer="5084" from-port="4" to-layer="5086" to-port="0" /> <edge from-layer="5085" from-port="0" to-layer="5086" to-port="1" /> <edge from-layer="5086" from-port="2" to-layer="5087" to-port="1" /> <edge from-layer="5087" from-port="2" to-layer="5132" to-port="2" /> <edge from-layer="5087" from-port="2" to-layer="10144" to-port="0" /> <edge from-layer="5088" from-port="0" to-layer="5106" to-port="0" /> <edge from-layer="5089" from-port="0" to-layer="5101" to-port="0" /> <edge from-layer="5090" from-port="0" to-layer="5104" to-port="1" /> <edge from-layer="5090" from-port="0" to-layer="5101" to-port="1" /> <edge from-layer="5091" from-port="1" to-layer="5093" to-port="0" /> <edge from-layer="5092" from-port="0" to-layer="5093" to-port="2" /> <edge from-layer="5093" from-port="3" to-layer="5103" to-port="0" /> <edge from-layer="5093" from-port="3" to-layer="5097" to-port="0" /> <edge from-layer="5094" from-port="1" to-layer="5096" to-port="0" /> <edge from-layer="5094" from-port="1" to-layer="5121" to-port="0" /> <edge from-layer="5095" from-port="0" to-layer="5096" to-port="2" /> <edge from-layer="5096" from-port="3" to-layer="5097" to-port="1" /> <edge from-layer="5097" from-port="2" to-layer="5099" to-port="0" /> <edge from-layer="5098" from-port="0" to-layer="5103" to-port="1" /> <edge from-layer="5098" from-port="0" to-layer="5099" to-port="1" /> <edge from-layer="5099" from-port="2" to-layer="5101" to-port="2" /> <edge from-layer="5100" from-port="0" to-layer="5104" to-port="3" /> <edge from-layer="5100" from-port="0" to-layer="5101" to-port="3" /> <edge from-layer="5101" from-port="4" to-layer="5106" to-port="1" /> <edge from-layer="5102" from-port="0" to-layer="5104" to-port="0" /> <edge from-layer="5103" from-port="2" to-layer="5104" to-port="2" /> <edge from-layer="5103" from-port="2" to-layer="5111" to-port="2" /> <edge from-layer="5104" from-port="4" to-layer="5106" to-port="2" /> <edge from-layer="5105" from-port="0" to-layer="5106" to-port="3" /> <edge from-layer="5106" from-port="4" to-layer="5113" to-port="0" /> <edge from-layer="5107" from-port="0" to-layer="5113" to-port="1" /> <edge from-layer="5108" from-port="0" to-layer="5111" to-port="0" /> <edge from-layer="5109" from-port="0" to-layer="5111" to-port="1" /> <edge from-layer="5110" from-port="0" to-layer="5111" to-port="3" /> <edge from-layer="5111" from-port="4" to-layer="5113" to-port="2" /> <edge from-layer="5112" from-port="0" to-layer="5113" to-port="3" /> <edge from-layer="5113" from-port="4" to-layer="5116" to-port="0" /> <edge from-layer="5114" from-port="0" to-layer="5116" to-port="1" /> <edge from-layer="5115" from-port="0" to-layer="5116" to-port="2" /> <edge from-layer="5116" from-port="3" to-layer="5117" to-port="0" /> <edge from-layer="5117" from-port="2" to-layer="5118" to-port="0" /> <edge from-layer="5117" from-port="2" to-layer="5131" to-port="0" /> <edge from-layer="5118" from-port="1" to-layer="5130" to-port="0" /> <edge from-layer="5119" from-port="0" to-layer="5121" to-port="1" /> <edge from-layer="5120" from-port="0" to-layer="5121" to-port="2" /> <edge from-layer="5121" from-port="3" to-layer="5125" to-port="0" /> <edge from-layer="5122" from-port="0" to-layer="5125" to-port="1" /> <edge from-layer="5123" from-port="0" to-layer="5125" to-port="2" /> <edge from-layer="5124" from-port="0" to-layer="5125" to-port="3" /> <edge from-layer="5125" from-port="4" to-layer="5127" to-port="0" /> <edge from-layer="5125" from-port="4" to-layer="5129" to-port="2" /> <edge from-layer="5126" from-port="0" to-layer="5127" to-port="1" /> <edge from-layer="5127" from-port="2" to-layer="5129" to-port="0" /> <edge from-layer="5128" from-port="0" to-layer="5129" to-port="1" /> <edge from-layer="5129" from-port="3" to-layer="5130" to-port="1" /> <edge from-layer="5130" from-port="2" to-layer="5131" to-port="1" /> <edge from-layer="5131" from-port="2" to-layer="5132" to-port="3" /> <edge from-layer="5132" from-port="4" to-layer="5134" to-port="0" /> <edge from-layer="5133" from-port="0" to-layer="5134" to-port="1" /> <edge from-layer="5134" from-port="2" to-layer="5136" to-port="0" /> <edge from-layer="5135" from-port="0" to-layer="5136" to-port="1" /> <edge from-layer="5136" from-port="2" to-layer="5144" to-port="0" /> <edge from-layer="5137" from-port="0" to-layer="5138" to-port="0" /> <edge from-layer="5138" from-port="1" to-layer="5141" to-port="0" /> <edge from-layer="5139" from-port="0" to-layer="5140" to-port="0" /> <edge from-layer="5140" from-port="1" to-layer="5141" to-port="1" /> <edge from-layer="5141" from-port="2" to-layer="5143" to-port="0" /> <edge from-layer="5142" from-port="0" to-layer="5143" to-port="1" /> <edge from-layer="5143" from-port="2" to-layer="5144" to-port="1" /> <edge from-layer="5144" from-port="2" to-layer="5146" to-port="0" /> <edge from-layer="5145" from-port="0" to-layer="5146" to-port="1" /> <edge from-layer="5146" from-port="2" to-layer="5147" to-port="1" /> <edge from-layer="5147" from-port="2" to-layer="5148" to-port="0" /> <edge from-layer="5148" from-port="2" to-layer="5150" to-port="0" /> <edge from-layer="5148" from-port="2" to-layer="5379" to-port="1" /> <edge from-layer="5148" from-port="2" to-layer="5177" to-port="0" /> <edge from-layer="5149" from-port="0" to-layer="5150" to-port="1" /> <edge from-layer="5150" from-port="2" to-layer="5152" to-port="0" /> <edge from-layer="5151" from-port="0" to-layer="5152" to-port="1" /> <edge from-layer="5152" from-port="2" to-layer="5154" to-port="0" /> <edge from-layer="5153" from-port="0" to-layer="5154" to-port="1" /> <edge from-layer="5154" from-port="2" to-layer="5162" to-port="0" /> <edge from-layer="5155" from-port="0" to-layer="5156" to-port="0" /> <edge from-layer="5156" from-port="1" to-layer="5159" to-port="0" /> <edge from-layer="5157" from-port="0" to-layer="5158" to-port="0" /> <edge from-layer="5158" from-port="1" to-layer="5159" to-port="1" /> <edge from-layer="5159" from-port="2" to-layer="5161" to-port="0" /> <edge from-layer="5160" from-port="0" to-layer="5161" to-port="1" /> <edge from-layer="5161" from-port="2" to-layer="5162" to-port="1" /> <edge from-layer="5162" from-port="2" to-layer="5164" to-port="0" /> <edge from-layer="5163" from-port="0" to-layer="5164" to-port="1" /> <edge from-layer="5164" from-port="2" to-layer="5165" to-port="0" /> <edge from-layer="5165" from-port="1" to-layer="5173" to-port="0" /> <edge from-layer="5166" from-port="0" to-layer="5167" to-port="0" /> <edge from-layer="5167" from-port="1" to-layer="5170" to-port="0" /> <edge from-layer="5168" from-port="0" to-layer="5169" to-port="0" /> <edge from-layer="5169" from-port="1" to-layer="5170" to-port="1" /> <edge from-layer="5170" from-port="2" to-layer="5172" to-port="0" /> <edge from-layer="5171" from-port="0" to-layer="5172" to-port="1" /> <edge from-layer="5172" from-port="2" to-layer="5173" to-port="1" /> <edge from-layer="5173" from-port="2" to-layer="5175" to-port="0" /> <edge from-layer="5174" from-port="0" to-layer="5175" to-port="1" /> <edge from-layer="5175" from-port="2" to-layer="5378" to-port="0" /> <edge from-layer="5176" from-port="0" to-layer="5177" to-port="1" /> <edge from-layer="5177" from-port="2" to-layer="5179" to-port="0" /> <edge from-layer="5178" from-port="0" to-layer="5179" to-port="1" /> <edge from-layer="5179" from-port="2" to-layer="5181" to-port="0" /> <edge from-layer="5180" from-port="0" to-layer="5181" to-port="1" /> <edge from-layer="5181" from-port="2" to-layer="5189" to-port="0" /> <edge from-layer="5182" from-port="0" to-layer="5183" to-port="0" /> <edge from-layer="5183" from-port="1" to-layer="5186" to-port="0" /> <edge from-layer="5184" from-port="0" to-layer="5185" to-port="0" /> <edge from-layer="5185" from-port="1" to-layer="5186" to-port="1" /> <edge from-layer="5186" from-port="2" to-layer="5188" to-port="0" /> <edge from-layer="5187" from-port="0" to-layer="5188" to-port="1" /> <edge from-layer="5188" from-port="2" to-layer="5189" to-port="1" /> <edge from-layer="5189" from-port="2" to-layer="5191" to-port="0" /> <edge from-layer="5190" from-port="0" to-layer="5191" to-port="1" /> <edge from-layer="5191" from-port="2" to-layer="5193" to-port="0" /> <edge from-layer="5192" from-port="0" to-layer="5193" to-port="1" /> <edge from-layer="5193" from-port="2" to-layer="5209" to-port="0" /> <edge from-layer="5193" from-port="2" to-layer="5197" to-port="0" /> <edge from-layer="5193" from-port="2" to-layer="5315" to-port="0" /> <edge from-layer="5194" from-port="0" to-layer="5197" to-port="1" /> <edge from-layer="5195" from-port="0" to-layer="5197" to-port="2" /> <edge from-layer="5196" from-port="0" to-layer="5197" to-port="3" /> <edge from-layer="5197" from-port="4" to-layer="5199" to-port="0" /> <edge from-layer="5198" from-port="0" to-layer="5199" to-port="1" /> <edge from-layer="5199" from-port="2" to-layer="5203" to-port="0" /> <edge from-layer="5199" from-port="2" to-layer="5268" to-port="0" /> <edge from-layer="5200" from-port="0" to-layer="5203" to-port="1" /> <edge from-layer="5201" from-port="0" to-layer="5203" to-port="2" /> <edge from-layer="5202" from-port="0" to-layer="5203" to-port="3" /> <edge from-layer="5203" from-port="4" to-layer="5249" to-port="0" /> <edge from-layer="5203" from-port="4" to-layer="5235" to-port="0" /> <edge from-layer="5203" from-port="4" to-layer="5254" to-port="0" /> <edge from-layer="5203" from-port="4" to-layer="5238" to-port="0" /> <edge from-layer="5204" from-port="0" to-layer="5231" to-port="0" /> <edge from-layer="5205" from-port="0" to-layer="5231" to-port="1" /> <edge from-layer="5206" from-port="0" to-layer="5209" to-port="1" /> <edge from-layer="5207" from-port="0" to-layer="5209" to-port="2" /> <edge from-layer="5208" from-port="0" to-layer="5209" to-port="3" /> <edge from-layer="5209" from-port="4" to-layer="5211" to-port="0" /> <edge from-layer="5210" from-port="0" to-layer="5211" to-port="1" /> <edge from-layer="5211" from-port="2" to-layer="5212" to-port="0" /> <edge from-layer="5211" from-port="2" to-layer="5300" to-port="0" /> <edge from-layer="5211" from-port="2" to-layer="5273" to-port="0" /> <edge from-layer="5212" from-port="1" to-layer="5214" to-port="0" /> <edge from-layer="5213" from-port="0" to-layer="5214" to-port="2" /> <edge from-layer="5214" from-port="3" to-layer="5227" to-port="0" /> <edge from-layer="5215" from-port="0" to-layer="5220" to-port="0" /> <edge from-layer="5216" from-port="0" to-layer="5219" to-port="1" /> <edge from-layer="5217" from-port="0" to-layer="5219" to-port="2" /> <edge from-layer="5218" from-port="0" to-layer="5219" to-port="3" /> <edge from-layer="5219" from-port="4" to-layer="5220" to-port="1" /> <edge from-layer="5220" from-port="2" to-layer="5221" to-port="0" /> <edge from-layer="5221" from-port="1" to-layer="5223" to-port="0" /> <edge from-layer="5222" from-port="0" to-layer="5223" to-port="2" /> <edge from-layer="5223" from-port="3" to-layer="5302" to-port="0" /> <edge from-layer="5223" from-port="3" to-layer="5224" to-port="0" /> <edge from-layer="5224" from-port="1" to-layer="5226" to-port="0" /> <edge from-layer="5225" from-port="0" to-layer="5226" to-port="2" /> <edge from-layer="5226" from-port="3" to-layer="5227" to-port="1" /> <edge from-layer="5227" from-port="2" to-layer="5229" to-port="0" /> <edge from-layer="5228" from-port="0" to-layer="5229" to-port="1" /> <edge from-layer="5229" from-port="2" to-layer="5259" to-port="2" /> <edge from-layer="5229" from-port="2" to-layer="5231" to-port="2" /> <edge from-layer="5230" from-port="0" to-layer="5231" to-port="3" /> <edge from-layer="5231" from-port="4" to-layer="5233" to-port="0" /> <edge from-layer="5232" from-port="0" to-layer="5233" to-port="2" /> <edge from-layer="5233" from-port="3" to-layer="5234" to-port="0" /> <edge from-layer="5234" from-port="2" to-layer="5235" to-port="1" /> <edge from-layer="5234" from-port="2" to-layer="5274" to-port="1" /> <edge from-layer="5235" from-port="2" to-layer="5264" to-port="0" /> <edge from-layer="5236" from-port="0" to-layer="5246" to-port="0" /> <edge from-layer="5237" from-port="0" to-layer="5246" to-port="1" /> <edge from-layer="5238" from-port="1" to-layer="5240" to-port="0" /> <edge from-layer="5239" from-port="0" to-layer="5240" to-port="2" /> <edge from-layer="5240" from-port="3" to-layer="5241" to-port="0" /> <edge from-layer="5241" from-port="2" to-layer="5242" to-port="0" /> <edge from-layer="5242" from-port="1" to-layer="5244" to-port="0" /> <edge from-layer="5243" from-port="0" to-layer="5244" to-port="1" /> <edge from-layer="5244" from-port="2" to-layer="5246" to-port="2" /> <edge from-layer="5245" from-port="0" to-layer="5246" to-port="3" /> <edge from-layer="5246" from-port="4" to-layer="5249" to-port="1" /> <edge from-layer="5246" from-port="4" to-layer="5254" to-port="2" /> <edge from-layer="5247" from-port="0" to-layer="5249" to-port="2" /> <edge from-layer="5248" from-port="0" to-layer="5249" to-port="3" /> <edge from-layer="5249" from-port="4" to-layer="5251" to-port="0" /> <edge from-layer="5250" from-port="0" to-layer="5251" to-port="1" /> <edge from-layer="5251" from-port="2" to-layer="5255" to-port="0" /> <edge from-layer="5252" from-port="0" to-layer="5254" to-port="1" /> <edge from-layer="5253" from-port="0" to-layer="5254" to-port="3" /> <edge from-layer="5254" from-port="4" to-layer="5255" to-port="1" /> <edge from-layer="5255" from-port="2" to-layer="5263" to-port="0" /> <edge from-layer="5256" from-port="0" to-layer="5259" to-port="0" /> <edge from-layer="5257" from-port="0" to-layer="5259" to-port="1" /> <edge from-layer="5258" from-port="0" to-layer="5259" to-port="3" /> <edge from-layer="5259" from-port="4" to-layer="5261" to-port="0" /> <edge from-layer="5260" from-port="0" to-layer="5261" to-port="2" /> <edge from-layer="5261" from-port="3" to-layer="5262" to-port="0" /> <edge from-layer="5262" from-port="2" to-layer="5263" to-port="1" /> <edge from-layer="5262" from-port="2" to-layer="5295" to-port="1" /> <edge from-layer="5263" from-port="2" to-layer="5264" to-port="1" /> <edge from-layer="5264" from-port="2" to-layer="5269" to-port="0" /> <edge from-layer="5265" from-port="0" to-layer="5268" to-port="1" /> <edge from-layer="5266" from-port="0" to-layer="5268" to-port="2" /> <edge from-layer="5267" from-port="0" to-layer="5268" to-port="3" /> <edge from-layer="5268" from-port="4" to-layer="5269" to-port="1" /> <edge from-layer="5269" from-port="2" to-layer="5363" to-port="0" /> <edge from-layer="5269" from-port="2" to-layer="5325" to-port="0" /> <edge from-layer="5270" from-port="0" to-layer="5273" to-port="1" /> <edge from-layer="5271" from-port="0" to-layer="5273" to-port="2" /> <edge from-layer="5272" from-port="0" to-layer="5273" to-port="3" /> <edge from-layer="5273" from-port="4" to-layer="5293" to-port="0" /> <edge from-layer="5273" from-port="4" to-layer="5288" to-port="0" /> <edge from-layer="5273" from-port="4" to-layer="5274" to-port="0" /> <edge from-layer="5273" from-port="4" to-layer="5277" to-port="0" /> <edge from-layer="5274" from-port="2" to-layer="5296" to-port="0" /> <edge from-layer="5275" from-port="0" to-layer="5285" to-port="0" /> <edge from-layer="5276" from-port="0" to-layer="5285" to-port="1" /> <edge from-layer="5277" from-port="1" to-layer="5279" to-port="0" /> <edge from-layer="5278" from-port="0" to-layer="5279" to-port="2" /> <edge from-layer="5279" from-port="3" to-layer="5280" to-port="0" /> <edge from-layer="5280" from-port="2" to-layer="5281" to-port="0" /> <edge from-layer="5281" from-port="1" to-layer="5283" to-port="0" /> <edge from-layer="5282" from-port="0" to-layer="5283" to-port="1" /> <edge from-layer="5283" from-port="2" to-layer="5285" to-port="2" /> <edge from-layer="5284" from-port="0" to-layer="5285" to-port="3" /> <edge from-layer="5285" from-port="4" to-layer="5288" to-port="1" /> <edge from-layer="5285" from-port="4" to-layer="5293" to-port="2" /> <edge from-layer="5286" from-port="0" to-layer="5288" to-port="2" /> <edge from-layer="5287" from-port="0" to-layer="5288" to-port="3" /> <edge from-layer="5288" from-port="4" to-layer="5290" to-port="0" /> <edge from-layer="5289" from-port="0" to-layer="5290" to-port="1" /> <edge from-layer="5290" from-port="2" to-layer="5294" to-port="0" /> <edge from-layer="5291" from-port="0" to-layer="5293" to-port="1" /> <edge from-layer="5292" from-port="0" to-layer="5293" to-port="3" /> <edge from-layer="5293" from-port="4" to-layer="5294" to-port="1" /> <edge from-layer="5294" from-port="2" to-layer="5295" to-port="0" /> <edge from-layer="5295" from-port="2" to-layer="5296" to-port="1" /> <edge from-layer="5296" from-port="2" to-layer="5301" to-port="0" /> <edge from-layer="5297" from-port="0" to-layer="5300" to-port="1" /> <edge from-layer="5298" from-port="0" to-layer="5300" to-port="2" /> <edge from-layer="5299" from-port="0" to-layer="5300" to-port="3" /> <edge from-layer="5300" from-port="4" to-layer="5301" to-port="1" /> <edge from-layer="5301" from-port="2" to-layer="5302" to-port="1" /> <edge from-layer="5302" from-port="2" to-layer="10143" to-port="0" /> <edge from-layer="5302" from-port="2" to-layer="5363" to-port="1" /> <edge from-layer="5302" from-port="2" to-layer="5322" to-port="0" /> <edge from-layer="5303" from-port="0" to-layer="5308" to-port="0" /> <edge from-layer="5304" from-port="0" to-layer="5307" to-port="1" /> <edge from-layer="5305" from-port="0" to-layer="5307" to-port="2" /> <edge from-layer="5306" from-port="0" to-layer="5307" to-port="3" /> <edge from-layer="5307" from-port="4" to-layer="5308" to-port="1" /> <edge from-layer="5308" from-port="2" to-layer="5309" to-port="0" /> <edge from-layer="5309" from-port="1" to-layer="5311" to-port="0" /> <edge from-layer="5310" from-port="0" to-layer="5311" to-port="2" /> <edge from-layer="5311" from-port="3" to-layer="5318" to-port="0" /> <edge from-layer="5312" from-port="0" to-layer="5315" to-port="1" /> <edge from-layer="5313" from-port="0" to-layer="5315" to-port="2" /> <edge from-layer="5314" from-port="0" to-layer="5315" to-port="3" /> <edge from-layer="5315" from-port="4" to-layer="5317" to-port="0" /> <edge from-layer="5316" from-port="0" to-layer="5317" to-port="1" /> <edge from-layer="5317" from-port="2" to-layer="5318" to-port="1" /> <edge from-layer="5318" from-port="2" to-layer="5363" to-port="2" /> <edge from-layer="5318" from-port="2" to-layer="10142" to-port="0" /> <edge from-layer="5319" from-port="0" to-layer="5337" to-port="0" /> <edge from-layer="5320" from-port="0" to-layer="5332" to-port="0" /> <edge from-layer="5321" from-port="0" to-layer="5335" to-port="1" /> <edge from-layer="5321" from-port="0" to-layer="5332" to-port="1" /> <edge from-layer="5322" from-port="1" to-layer="5324" to-port="0" /> <edge from-layer="5323" from-port="0" to-layer="5324" to-port="2" /> <edge from-layer="5324" from-port="3" to-layer="5334" to-port="0" /> <edge from-layer="5324" from-port="3" to-layer="5328" to-port="0" /> <edge from-layer="5325" from-port="1" to-layer="5352" to-port="0" /> <edge from-layer="5325" from-port="1" to-layer="5327" to-port="0" /> <edge from-layer="5326" from-port="0" to-layer="5327" to-port="2" /> <edge from-layer="5327" from-port="3" to-layer="5328" to-port="1" /> <edge from-layer="5328" from-port="2" to-layer="5330" to-port="0" /> <edge from-layer="5329" from-port="0" to-layer="5334" to-port="1" /> <edge from-layer="5329" from-port="0" to-layer="5330" to-port="1" /> <edge from-layer="5330" from-port="2" to-layer="5332" to-port="2" /> <edge from-layer="5331" from-port="0" to-layer="5332" to-port="3" /> <edge from-layer="5331" from-port="0" to-layer="5335" to-port="3" /> <edge from-layer="5332" from-port="4" to-layer="5337" to-port="1" /> <edge from-layer="5333" from-port="0" to-layer="5335" to-port="0" /> <edge from-layer="5334" from-port="2" to-layer="5342" to-port="2" /> <edge from-layer="5334" from-port="2" to-layer="5335" to-port="2" /> <edge from-layer="5335" from-port="4" to-layer="5337" to-port="2" /> <edge from-layer="5336" from-port="0" to-layer="5337" to-port="3" /> <edge from-layer="5337" from-port="4" to-layer="5344" to-port="0" /> <edge from-layer="5338" from-port="0" to-layer="5344" to-port="1" /> <edge from-layer="5339" from-port="0" to-layer="5342" to-port="0" /> <edge from-layer="5340" from-port="0" to-layer="5342" to-port="1" /> <edge from-layer="5341" from-port="0" to-layer="5342" to-port="3" /> <edge from-layer="5342" from-port="4" to-layer="5344" to-port="2" /> <edge from-layer="5343" from-port="0" to-layer="5344" to-port="3" /> <edge from-layer="5344" from-port="4" to-layer="5347" to-port="0" /> <edge from-layer="5345" from-port="0" to-layer="5347" to-port="1" /> <edge from-layer="5346" from-port="0" to-layer="5347" to-port="2" /> <edge from-layer="5347" from-port="3" to-layer="5348" to-port="0" /> <edge from-layer="5348" from-port="2" to-layer="5349" to-port="0" /> <edge from-layer="5348" from-port="2" to-layer="5362" to-port="0" /> <edge from-layer="5349" from-port="1" to-layer="5361" to-port="0" /> <edge from-layer="5350" from-port="0" to-layer="5352" to-port="1" /> <edge from-layer="5351" from-port="0" to-layer="5352" to-port="2" /> <edge from-layer="5352" from-port="3" to-layer="5356" to-port="0" /> <edge from-layer="5353" from-port="0" to-layer="5356" to-port="1" /> <edge from-layer="5354" from-port="0" to-layer="5356" to-port="2" /> <edge from-layer="5355" from-port="0" to-layer="5356" to-port="3" /> <edge from-layer="5356" from-port="4" to-layer="5360" to-port="2" /> <edge from-layer="5356" from-port="4" to-layer="5358" to-port="0" /> <edge from-layer="5357" from-port="0" to-layer="5358" to-port="1" /> <edge from-layer="5358" from-port="2" to-layer="5360" to-port="0" /> <edge from-layer="5359" from-port="0" to-layer="5360" to-port="1" /> <edge from-layer="5360" from-port="3" to-layer="5361" to-port="1" /> <edge from-layer="5361" from-port="2" to-layer="5362" to-port="1" /> <edge from-layer="5362" from-port="2" to-layer="5363" to-port="3" /> <edge from-layer="5363" from-port="4" to-layer="5365" to-port="0" /> <edge from-layer="5364" from-port="0" to-layer="5365" to-port="1" /> <edge from-layer="5365" from-port="2" to-layer="5367" to-port="0" /> <edge from-layer="5366" from-port="0" to-layer="5367" to-port="1" /> <edge from-layer="5367" from-port="2" to-layer="5375" to-port="0" /> <edge from-layer="5368" from-port="0" to-layer="5369" to-port="0" /> <edge from-layer="5369" from-port="1" to-layer="5372" to-port="0" /> <edge from-layer="5370" from-port="0" to-layer="5371" to-port="0" /> <edge from-layer="5371" from-port="1" to-layer="5372" to-port="1" /> <edge from-layer="5372" from-port="2" to-layer="5374" to-port="0" /> <edge from-layer="5373" from-port="0" to-layer="5374" to-port="1" /> <edge from-layer="5374" from-port="2" to-layer="5375" to-port="1" /> <edge from-layer="5375" from-port="2" to-layer="5377" to-port="0" /> <edge from-layer="5376" from-port="0" to-layer="5377" to-port="1" /> <edge from-layer="5377" from-port="2" to-layer="5378" to-port="1" /> <edge from-layer="5378" from-port="2" to-layer="5379" to-port="0" /> <edge from-layer="5379" from-port="2" to-layer="5381" to-port="0" /> <edge from-layer="5379" from-port="2" to-layer="5610" to-port="1" /> <edge from-layer="5379" from-port="2" to-layer="5408" to-port="0" /> <edge from-layer="5380" from-port="0" to-layer="5381" to-port="1" /> <edge from-layer="5381" from-port="2" to-layer="5383" to-port="0" /> <edge from-layer="5382" from-port="0" to-layer="5383" to-port="1" /> <edge from-layer="5383" from-port="2" to-layer="5385" to-port="0" /> <edge from-layer="5384" from-port="0" to-layer="5385" to-port="1" /> <edge from-layer="5385" from-port="2" to-layer="5393" to-port="0" /> <edge from-layer="5386" from-port="0" to-layer="5387" to-port="0" /> <edge from-layer="5387" from-port="1" to-layer="5390" to-port="0" /> <edge from-layer="5388" from-port="0" to-layer="5389" to-port="0" /> <edge from-layer="5389" from-port="1" to-layer="5390" to-port="1" /> <edge from-layer="5390" from-port="2" to-layer="5392" to-port="0" /> <edge from-layer="5391" from-port="0" to-layer="5392" to-port="1" /> <edge from-layer="5392" from-port="2" to-layer="5393" to-port="1" /> <edge from-layer="5393" from-port="2" to-layer="5395" to-port="0" /> <edge from-layer="5394" from-port="0" to-layer="5395" to-port="1" /> <edge from-layer="5395" from-port="2" to-layer="5396" to-port="0" /> <edge from-layer="5396" from-port="1" to-layer="5404" to-port="0" /> <edge from-layer="5397" from-port="0" to-layer="5398" to-port="0" /> <edge from-layer="5398" from-port="1" to-layer="5401" to-port="0" /> <edge from-layer="5399" from-port="0" to-layer="5400" to-port="0" /> <edge from-layer="5400" from-port="1" to-layer="5401" to-port="1" /> <edge from-layer="5401" from-port="2" to-layer="5403" to-port="0" /> <edge from-layer="5402" from-port="0" to-layer="5403" to-port="1" /> <edge from-layer="5403" from-port="2" to-layer="5404" to-port="1" /> <edge from-layer="5404" from-port="2" to-layer="5406" to-port="0" /> <edge from-layer="5405" from-port="0" to-layer="5406" to-port="1" /> <edge from-layer="5406" from-port="2" to-layer="5609" to-port="0" /> <edge from-layer="5407" from-port="0" to-layer="5408" to-port="1" /> <edge from-layer="5408" from-port="2" to-layer="5410" to-port="0" /> <edge from-layer="5409" from-port="0" to-layer="5410" to-port="1" /> <edge from-layer="5410" from-port="2" to-layer="5412" to-port="0" /> <edge from-layer="5411" from-port="0" to-layer="5412" to-port="1" /> <edge from-layer="5412" from-port="2" to-layer="5420" to-port="0" /> <edge from-layer="5413" from-port="0" to-layer="5414" to-port="0" /> <edge from-layer="5414" from-port="1" to-layer="5417" to-port="0" /> <edge from-layer="5415" from-port="0" to-layer="5416" to-port="0" /> <edge from-layer="5416" from-port="1" to-layer="5417" to-port="1" /> <edge from-layer="5417" from-port="2" to-layer="5419" to-port="0" /> <edge from-layer="5418" from-port="0" to-layer="5419" to-port="1" /> <edge from-layer="5419" from-port="2" to-layer="5420" to-port="1" /> <edge from-layer="5420" from-port="2" to-layer="5422" to-port="0" /> <edge from-layer="5421" from-port="0" to-layer="5422" to-port="1" /> <edge from-layer="5422" from-port="2" to-layer="5424" to-port="0" /> <edge from-layer="5423" from-port="0" to-layer="5424" to-port="1" /> <edge from-layer="5424" from-port="2" to-layer="5440" to-port="0" /> <edge from-layer="5424" from-port="2" to-layer="5428" to-port="0" /> <edge from-layer="5424" from-port="2" to-layer="5546" to-port="0" /> <edge from-layer="5425" from-port="0" to-layer="5428" to-port="1" /> <edge from-layer="5426" from-port="0" to-layer="5428" to-port="2" /> <edge from-layer="5427" from-port="0" to-layer="5428" to-port="3" /> <edge from-layer="5428" from-port="4" to-layer="5430" to-port="0" /> <edge from-layer="5429" from-port="0" to-layer="5430" to-port="1" /> <edge from-layer="5430" from-port="2" to-layer="5499" to-port="0" /> <edge from-layer="5430" from-port="2" to-layer="5434" to-port="0" /> <edge from-layer="5431" from-port="0" to-layer="5434" to-port="1" /> <edge from-layer="5432" from-port="0" to-layer="5434" to-port="2" /> <edge from-layer="5433" from-port="0" to-layer="5434" to-port="3" /> <edge from-layer="5434" from-port="4" to-layer="5480" to-port="0" /> <edge from-layer="5434" from-port="4" to-layer="5466" to-port="0" /> <edge from-layer="5434" from-port="4" to-layer="5485" to-port="0" /> <edge from-layer="5434" from-port="4" to-layer="5469" to-port="0" /> <edge from-layer="5435" from-port="0" to-layer="5462" to-port="0" /> <edge from-layer="5436" from-port="0" to-layer="5462" to-port="1" /> <edge from-layer="5437" from-port="0" to-layer="5440" to-port="1" /> <edge from-layer="5438" from-port="0" to-layer="5440" to-port="2" /> <edge from-layer="5439" from-port="0" to-layer="5440" to-port="3" /> <edge from-layer="5440" from-port="4" to-layer="5442" to-port="0" /> <edge from-layer="5441" from-port="0" to-layer="5442" to-port="1" /> <edge from-layer="5442" from-port="2" to-layer="5443" to-port="0" /> <edge from-layer="5442" from-port="2" to-layer="5504" to-port="0" /> <edge from-layer="5442" from-port="2" to-layer="5531" to-port="0" /> <edge from-layer="5443" from-port="1" to-layer="5445" to-port="0" /> <edge from-layer="5444" from-port="0" to-layer="5445" to-port="2" /> <edge from-layer="5445" from-port="3" to-layer="5458" to-port="0" /> <edge from-layer="5446" from-port="0" to-layer="5451" to-port="0" /> <edge from-layer="5447" from-port="0" to-layer="5450" to-port="1" /> <edge from-layer="5448" from-port="0" to-layer="5450" to-port="2" /> <edge from-layer="5449" from-port="0" to-layer="5450" to-port="3" /> <edge from-layer="5450" from-port="4" to-layer="5451" to-port="1" /> <edge from-layer="5451" from-port="2" to-layer="5452" to-port="0" /> <edge from-layer="5452" from-port="1" to-layer="5454" to-port="0" /> <edge from-layer="5453" from-port="0" to-layer="5454" to-port="2" /> <edge from-layer="5454" from-port="3" to-layer="5533" to-port="0" /> <edge from-layer="5454" from-port="3" to-layer="5455" to-port="0" /> <edge from-layer="5455" from-port="1" to-layer="5457" to-port="0" /> <edge from-layer="5456" from-port="0" to-layer="5457" to-port="2" /> <edge from-layer="5457" from-port="3" to-layer="5458" to-port="1" /> <edge from-layer="5458" from-port="2" to-layer="5460" to-port="0" /> <edge from-layer="5459" from-port="0" to-layer="5460" to-port="1" /> <edge from-layer="5460" from-port="2" to-layer="5462" to-port="2" /> <edge from-layer="5460" from-port="2" to-layer="5490" to-port="2" /> <edge from-layer="5461" from-port="0" to-layer="5462" to-port="3" /> <edge from-layer="5462" from-port="4" to-layer="5464" to-port="0" /> <edge from-layer="5463" from-port="0" to-layer="5464" to-port="2" /> <edge from-layer="5464" from-port="3" to-layer="5465" to-port="0" /> <edge from-layer="5465" from-port="2" to-layer="5466" to-port="1" /> <edge from-layer="5465" from-port="2" to-layer="5505" to-port="1" /> <edge from-layer="5466" from-port="2" to-layer="5495" to-port="0" /> <edge from-layer="5467" from-port="0" to-layer="5477" to-port="0" /> <edge from-layer="5468" from-port="0" to-layer="5477" to-port="1" /> <edge from-layer="5469" from-port="1" to-layer="5471" to-port="0" /> <edge from-layer="5470" from-port="0" to-layer="5471" to-port="2" /> <edge from-layer="5471" from-port="3" to-layer="5472" to-port="0" /> <edge from-layer="5472" from-port="2" to-layer="5473" to-port="0" /> <edge from-layer="5473" from-port="1" to-layer="5475" to-port="0" /> <edge from-layer="5474" from-port="0" to-layer="5475" to-port="1" /> <edge from-layer="5475" from-port="2" to-layer="5477" to-port="2" /> <edge from-layer="5476" from-port="0" to-layer="5477" to-port="3" /> <edge from-layer="5477" from-port="4" to-layer="5480" to-port="1" /> <edge from-layer="5477" from-port="4" to-layer="5485" to-port="2" /> <edge from-layer="5478" from-port="0" to-layer="5480" to-port="2" /> <edge from-layer="5479" from-port="0" to-layer="5480" to-port="3" /> <edge from-layer="5480" from-port="4" to-layer="5482" to-port="0" /> <edge from-layer="5481" from-port="0" to-layer="5482" to-port="1" /> <edge from-layer="5482" from-port="2" to-layer="5486" to-port="0" /> <edge from-layer="5483" from-port="0" to-layer="5485" to-port="1" /> <edge from-layer="5484" from-port="0" to-layer="5485" to-port="3" /> <edge from-layer="5485" from-port="4" to-layer="5486" to-port="1" /> <edge from-layer="5486" from-port="2" to-layer="5494" to-port="0" /> <edge from-layer="5487" from-port="0" to-layer="5490" to-port="0" /> <edge from-layer="5488" from-port="0" to-layer="5490" to-port="1" /> <edge from-layer="5489" from-port="0" to-layer="5490" to-port="3" /> <edge from-layer="5490" from-port="4" to-layer="5492" to-port="0" /> <edge from-layer="5491" from-port="0" to-layer="5492" to-port="2" /> <edge from-layer="5492" from-port="3" to-layer="5493" to-port="0" /> <edge from-layer="5493" from-port="2" to-layer="5526" to-port="1" /> <edge from-layer="5493" from-port="2" to-layer="5494" to-port="1" /> <edge from-layer="5494" from-port="2" to-layer="5495" to-port="1" /> <edge from-layer="5495" from-port="2" to-layer="5500" to-port="0" /> <edge from-layer="5496" from-port="0" to-layer="5499" to-port="1" /> <edge from-layer="5497" from-port="0" to-layer="5499" to-port="2" /> <edge from-layer="5498" from-port="0" to-layer="5499" to-port="3" /> <edge from-layer="5499" from-port="4" to-layer="5500" to-port="1" /> <edge from-layer="5500" from-port="2" to-layer="5594" to-port="0" /> <edge from-layer="5500" from-port="2" to-layer="5556" to-port="0" /> <edge from-layer="5501" from-port="0" to-layer="5504" to-port="1" /> <edge from-layer="5502" from-port="0" to-layer="5504" to-port="2" /> <edge from-layer="5503" from-port="0" to-layer="5504" to-port="3" /> <edge from-layer="5504" from-port="4" to-layer="5524" to-port="0" /> <edge from-layer="5504" from-port="4" to-layer="5519" to-port="0" /> <edge from-layer="5504" from-port="4" to-layer="5505" to-port="0" /> <edge from-layer="5504" from-port="4" to-layer="5508" to-port="0" /> <edge from-layer="5505" from-port="2" to-layer="5527" to-port="0" /> <edge from-layer="5506" from-port="0" to-layer="5516" to-port="0" /> <edge from-layer="5507" from-port="0" to-layer="5516" to-port="1" /> <edge from-layer="5508" from-port="1" to-layer="5510" to-port="0" /> <edge from-layer="5509" from-port="0" to-layer="5510" to-port="2" /> <edge from-layer="5510" from-port="3" to-layer="5511" to-port="0" /> <edge from-layer="5511" from-port="2" to-layer="5512" to-port="0" /> <edge from-layer="5512" from-port="1" to-layer="5514" to-port="0" /> <edge from-layer="5513" from-port="0" to-layer="5514" to-port="1" /> <edge from-layer="5514" from-port="2" to-layer="5516" to-port="2" /> <edge from-layer="5515" from-port="0" to-layer="5516" to-port="3" /> <edge from-layer="5516" from-port="4" to-layer="5519" to-port="1" /> <edge from-layer="5516" from-port="4" to-layer="5524" to-port="2" /> <edge from-layer="5517" from-port="0" to-layer="5519" to-port="2" /> <edge from-layer="5518" from-port="0" to-layer="5519" to-port="3" /> <edge from-layer="5519" from-port="4" to-layer="5521" to-port="0" /> <edge from-layer="5520" from-port="0" to-layer="5521" to-port="1" /> <edge from-layer="5521" from-port="2" to-layer="5525" to-port="0" /> <edge from-layer="5522" from-port="0" to-layer="5524" to-port="1" /> <edge from-layer="5523" from-port="0" to-layer="5524" to-port="3" /> <edge from-layer="5524" from-port="4" to-layer="5525" to-port="1" /> <edge from-layer="5525" from-port="2" to-layer="5526" to-port="0" /> <edge from-layer="5526" from-port="2" to-layer="5527" to-port="1" /> <edge from-layer="5527" from-port="2" to-layer="5532" to-port="0" /> <edge from-layer="5528" from-port="0" to-layer="5531" to-port="1" /> <edge from-layer="5529" from-port="0" to-layer="5531" to-port="2" /> <edge from-layer="5530" from-port="0" to-layer="5531" to-port="3" /> <edge from-layer="5531" from-port="4" to-layer="5532" to-port="1" /> <edge from-layer="5532" from-port="2" to-layer="5533" to-port="1" /> <edge from-layer="5533" from-port="2" to-layer="5594" to-port="1" /> <edge from-layer="5533" from-port="2" to-layer="5553" to-port="0" /> <edge from-layer="5533" from-port="2" to-layer="10141" to-port="0" /> <edge from-layer="5534" from-port="0" to-layer="5539" to-port="0" /> <edge from-layer="5535" from-port="0" to-layer="5538" to-port="1" /> <edge from-layer="5536" from-port="0" to-layer="5538" to-port="2" /> <edge from-layer="5537" from-port="0" to-layer="5538" to-port="3" /> <edge from-layer="5538" from-port="4" to-layer="5539" to-port="1" /> <edge from-layer="5539" from-port="2" to-layer="5540" to-port="0" /> <edge from-layer="5540" from-port="1" to-layer="5542" to-port="0" /> <edge from-layer="5541" from-port="0" to-layer="5542" to-port="2" /> <edge from-layer="5542" from-port="3" to-layer="5549" to-port="0" /> <edge from-layer="5543" from-port="0" to-layer="5546" to-port="1" /> <edge from-layer="5544" from-port="0" to-layer="5546" to-port="2" /> <edge from-layer="5545" from-port="0" to-layer="5546" to-port="3" /> <edge from-layer="5546" from-port="4" to-layer="5548" to-port="0" /> <edge from-layer="5547" from-port="0" to-layer="5548" to-port="1" /> <edge from-layer="5548" from-port="2" to-layer="5549" to-port="1" /> <edge from-layer="5549" from-port="2" to-layer="10140" to-port="0" /> <edge from-layer="5549" from-port="2" to-layer="5594" to-port="2" /> <edge from-layer="5550" from-port="0" to-layer="5568" to-port="0" /> <edge from-layer="5551" from-port="0" to-layer="5563" to-port="0" /> <edge from-layer="5552" from-port="0" to-layer="5563" to-port="1" /> <edge from-layer="5552" from-port="0" to-layer="5566" to-port="1" /> <edge from-layer="5553" from-port="1" to-layer="5555" to-port="0" /> <edge from-layer="5554" from-port="0" to-layer="5555" to-port="2" /> <edge from-layer="5555" from-port="3" to-layer="5559" to-port="0" /> <edge from-layer="5555" from-port="3" to-layer="5565" to-port="0" /> <edge from-layer="5556" from-port="1" to-layer="5583" to-port="0" /> <edge from-layer="5556" from-port="1" to-layer="5558" to-port="0" /> <edge from-layer="5557" from-port="0" to-layer="5558" to-port="2" /> <edge from-layer="5558" from-port="3" to-layer="5559" to-port="1" /> <edge from-layer="5559" from-port="2" to-layer="5561" to-port="0" /> <edge from-layer="5560" from-port="0" to-layer="5561" to-port="1" /> <edge from-layer="5560" from-port="0" to-layer="5565" to-port="1" /> <edge from-layer="5561" from-port="2" to-layer="5563" to-port="2" /> <edge from-layer="5562" from-port="0" to-layer="5566" to-port="3" /> <edge from-layer="5562" from-port="0" to-layer="5563" to-port="3" /> <edge from-layer="5563" from-port="4" to-layer="5568" to-port="1" /> <edge from-layer="5564" from-port="0" to-layer="5566" to-port="0" /> <edge from-layer="5565" from-port="2" to-layer="5573" to-port="2" /> <edge from-layer="5565" from-port="2" to-layer="5566" to-port="2" /> <edge from-layer="5566" from-port="4" to-layer="5568" to-port="2" /> <edge from-layer="5567" from-port="0" to-layer="5568" to-port="3" /> <edge from-layer="5568" from-port="4" to-layer="5575" to-port="0" /> <edge from-layer="5569" from-port="0" to-layer="5575" to-port="1" /> <edge from-layer="5570" from-port="0" to-layer="5573" to-port="0" /> <edge from-layer="5571" from-port="0" to-layer="5573" to-port="1" /> <edge from-layer="5572" from-port="0" to-layer="5573" to-port="3" /> <edge from-layer="5573" from-port="4" to-layer="5575" to-port="2" /> <edge from-layer="5574" from-port="0" to-layer="5575" to-port="3" /> <edge from-layer="5575" from-port="4" to-layer="5578" to-port="0" /> <edge from-layer="5576" from-port="0" to-layer="5578" to-port="1" /> <edge from-layer="5577" from-port="0" to-layer="5578" to-port="2" /> <edge from-layer="5578" from-port="3" to-layer="5579" to-port="0" /> <edge from-layer="5579" from-port="2" to-layer="5593" to-port="0" /> <edge from-layer="5579" from-port="2" to-layer="5580" to-port="0" /> <edge from-layer="5580" from-port="1" to-layer="5592" to-port="0" /> <edge from-layer="5581" from-port="0" to-layer="5583" to-port="1" /> <edge from-layer="5582" from-port="0" to-layer="5583" to-port="2" /> <edge from-layer="5583" from-port="3" to-layer="5587" to-port="0" /> <edge from-layer="5584" from-port="0" to-layer="5587" to-port="1" /> <edge from-layer="5585" from-port="0" to-layer="5587" to-port="2" /> <edge from-layer="5586" from-port="0" to-layer="5587" to-port="3" /> <edge from-layer="5587" from-port="4" to-layer="5591" to-port="2" /> <edge from-layer="5587" from-port="4" to-layer="5589" to-port="0" /> <edge from-layer="5588" from-port="0" to-layer="5589" to-port="1" /> <edge from-layer="5589" from-port="2" to-layer="5591" to-port="0" /> <edge from-layer="5590" from-port="0" to-layer="5591" to-port="1" /> <edge from-layer="5591" from-port="3" to-layer="5592" to-port="1" /> <edge from-layer="5592" from-port="2" to-layer="5593" to-port="1" /> <edge from-layer="5593" from-port="2" to-layer="5594" to-port="3" /> <edge from-layer="5594" from-port="4" to-layer="5596" to-port="0" /> <edge from-layer="5595" from-port="0" to-layer="5596" to-port="1" /> <edge from-layer="5596" from-port="2" to-layer="5598" to-port="0" /> <edge from-layer="5597" from-port="0" to-layer="5598" to-port="1" /> <edge from-layer="5598" from-port="2" to-layer="5606" to-port="0" /> <edge from-layer="5599" from-port="0" to-layer="5600" to-port="0" /> <edge from-layer="5600" from-port="1" to-layer="5603" to-port="0" /> <edge from-layer="5601" from-port="0" to-layer="5602" to-port="0" /> <edge from-layer="5602" from-port="1" to-layer="5603" to-port="1" /> <edge from-layer="5603" from-port="2" to-layer="5605" to-port="0" /> <edge from-layer="5604" from-port="0" to-layer="5605" to-port="1" /> <edge from-layer="5605" from-port="2" to-layer="5606" to-port="1" /> <edge from-layer="5606" from-port="2" to-layer="5608" to-port="0" /> <edge from-layer="5607" from-port="0" to-layer="5608" to-port="1" /> <edge from-layer="5608" from-port="2" to-layer="5609" to-port="1" /> <edge from-layer="5609" from-port="2" to-layer="5610" to-port="0" /> <edge from-layer="5610" from-port="2" to-layer="5612" to-port="0" /> <edge from-layer="5610" from-port="2" to-layer="5639" to-port="0" /> <edge from-layer="5610" from-port="2" to-layer="5841" to-port="1" /> <edge from-layer="5611" from-port="0" to-layer="5612" to-port="1" /> <edge from-layer="5612" from-port="2" to-layer="5614" to-port="0" /> <edge from-layer="5613" from-port="0" to-layer="5614" to-port="1" /> <edge from-layer="5614" from-port="2" to-layer="5616" to-port="0" /> <edge from-layer="5615" from-port="0" to-layer="5616" to-port="1" /> <edge from-layer="5616" from-port="2" to-layer="5624" to-port="0" /> <edge from-layer="5617" from-port="0" to-layer="5618" to-port="0" /> <edge from-layer="5618" from-port="1" to-layer="5621" to-port="0" /> <edge from-layer="5619" from-port="0" to-layer="5620" to-port="0" /> <edge from-layer="5620" from-port="1" to-layer="5621" to-port="1" /> <edge from-layer="5621" from-port="2" to-layer="5623" to-port="0" /> <edge from-layer="5622" from-port="0" to-layer="5623" to-port="1" /> <edge from-layer="5623" from-port="2" to-layer="5624" to-port="1" /> <edge from-layer="5624" from-port="2" to-layer="5626" to-port="0" /> <edge from-layer="5625" from-port="0" to-layer="5626" to-port="1" /> <edge from-layer="5626" from-port="2" to-layer="5627" to-port="0" /> <edge from-layer="5627" from-port="1" to-layer="5635" to-port="0" /> <edge from-layer="5628" from-port="0" to-layer="5629" to-port="0" /> <edge from-layer="5629" from-port="1" to-layer="5632" to-port="0" /> <edge from-layer="5630" from-port="0" to-layer="5631" to-port="0" /> <edge from-layer="5631" from-port="1" to-layer="5632" to-port="1" /> <edge from-layer="5632" from-port="2" to-layer="5634" to-port="0" /> <edge from-layer="5633" from-port="0" to-layer="5634" to-port="1" /> <edge from-layer="5634" from-port="2" to-layer="5635" to-port="1" /> <edge from-layer="5635" from-port="2" to-layer="5637" to-port="0" /> <edge from-layer="5636" from-port="0" to-layer="5637" to-port="1" /> <edge from-layer="5637" from-port="2" to-layer="5840" to-port="0" /> <edge from-layer="5638" from-port="0" to-layer="5639" to-port="1" /> <edge from-layer="5639" from-port="2" to-layer="5641" to-port="0" /> <edge from-layer="5640" from-port="0" to-layer="5641" to-port="1" /> <edge from-layer="5641" from-port="2" to-layer="5643" to-port="0" /> <edge from-layer="5642" from-port="0" to-layer="5643" to-port="1" /> <edge from-layer="5643" from-port="2" to-layer="5651" to-port="0" /> <edge from-layer="5644" from-port="0" to-layer="5645" to-port="0" /> <edge from-layer="5645" from-port="1" to-layer="5648" to-port="0" /> <edge from-layer="5646" from-port="0" to-layer="5647" to-port="0" /> <edge from-layer="5647" from-port="1" to-layer="5648" to-port="1" /> <edge from-layer="5648" from-port="2" to-layer="5650" to-port="0" /> <edge from-layer="5649" from-port="0" to-layer="5650" to-port="1" /> <edge from-layer="5650" from-port="2" to-layer="5651" to-port="1" /> <edge from-layer="5651" from-port="2" to-layer="5653" to-port="0" /> <edge from-layer="5652" from-port="0" to-layer="5653" to-port="1" /> <edge from-layer="5653" from-port="2" to-layer="5655" to-port="0" /> <edge from-layer="5654" from-port="0" to-layer="5655" to-port="1" /> <edge from-layer="5655" from-port="2" to-layer="5671" to-port="0" /> <edge from-layer="5655" from-port="2" to-layer="5659" to-port="0" /> <edge from-layer="5655" from-port="2" to-layer="5777" to-port="0" /> <edge from-layer="5656" from-port="0" to-layer="5659" to-port="1" /> <edge from-layer="5657" from-port="0" to-layer="5659" to-port="2" /> <edge from-layer="5658" from-port="0" to-layer="5659" to-port="3" /> <edge from-layer="5659" from-port="4" to-layer="5661" to-port="0" /> <edge from-layer="5660" from-port="0" to-layer="5661" to-port="1" /> <edge from-layer="5661" from-port="2" to-layer="5665" to-port="0" /> <edge from-layer="5661" from-port="2" to-layer="5730" to-port="0" /> <edge from-layer="5662" from-port="0" to-layer="5665" to-port="1" /> <edge from-layer="5663" from-port="0" to-layer="5665" to-port="2" /> <edge from-layer="5664" from-port="0" to-layer="5665" to-port="3" /> <edge from-layer="5665" from-port="4" to-layer="5716" to-port="0" /> <edge from-layer="5665" from-port="4" to-layer="5700" to-port="0" /> <edge from-layer="5665" from-port="4" to-layer="5697" to-port="0" /> <edge from-layer="5665" from-port="4" to-layer="5711" to-port="0" /> <edge from-layer="5666" from-port="0" to-layer="5693" to-port="0" /> <edge from-layer="5667" from-port="0" to-layer="5693" to-port="1" /> <edge from-layer="5668" from-port="0" to-layer="5671" to-port="1" /> <edge from-layer="5669" from-port="0" to-layer="5671" to-port="2" /> <edge from-layer="5670" from-port="0" to-layer="5671" to-port="3" /> <edge from-layer="5671" from-port="4" to-layer="5673" to-port="0" /> <edge from-layer="5672" from-port="0" to-layer="5673" to-port="1" /> <edge from-layer="5673" from-port="2" to-layer="5735" to-port="0" /> <edge from-layer="5673" from-port="2" to-layer="5674" to-port="0" /> <edge from-layer="5673" from-port="2" to-layer="5762" to-port="0" /> <edge from-layer="5674" from-port="1" to-layer="5676" to-port="0" /> <edge from-layer="5675" from-port="0" to-layer="5676" to-port="2" /> <edge from-layer="5676" from-port="3" to-layer="5689" to-port="0" /> <edge from-layer="5677" from-port="0" to-layer="5682" to-port="0" /> <edge from-layer="5678" from-port="0" to-layer="5681" to-port="1" /> <edge from-layer="5679" from-port="0" to-layer="5681" to-port="2" /> <edge from-layer="5680" from-port="0" to-layer="5681" to-port="3" /> <edge from-layer="5681" from-port="4" to-layer="5682" to-port="1" /> <edge from-layer="5682" from-port="2" to-layer="5683" to-port="0" /> <edge from-layer="5683" from-port="1" to-layer="5685" to-port="0" /> <edge from-layer="5684" from-port="0" to-layer="5685" to-port="2" /> <edge from-layer="5685" from-port="3" to-layer="5764" to-port="0" /> <edge from-layer="5685" from-port="3" to-layer="5686" to-port="0" /> <edge from-layer="5686" from-port="1" to-layer="5688" to-port="0" /> <edge from-layer="5687" from-port="0" to-layer="5688" to-port="2" /> <edge from-layer="5688" from-port="3" to-layer="5689" to-port="1" /> <edge from-layer="5689" from-port="2" to-layer="5691" to-port="0" /> <edge from-layer="5690" from-port="0" to-layer="5691" to-port="1" /> <edge from-layer="5691" from-port="2" to-layer="5693" to-port="2" /> <edge from-layer="5691" from-port="2" to-layer="5721" to-port="2" /> <edge from-layer="5692" from-port="0" to-layer="5693" to-port="3" /> <edge from-layer="5693" from-port="4" to-layer="5695" to-port="0" /> <edge from-layer="5694" from-port="0" to-layer="5695" to-port="2" /> <edge from-layer="5695" from-port="3" to-layer="5696" to-port="0" /> <edge from-layer="5696" from-port="2" to-layer="5697" to-port="1" /> <edge from-layer="5696" from-port="2" to-layer="5736" to-port="1" /> <edge from-layer="5697" from-port="2" to-layer="5726" to-port="0" /> <edge from-layer="5698" from-port="0" to-layer="5708" to-port="0" /> <edge from-layer="5699" from-port="0" to-layer="5708" to-port="1" /> <edge from-layer="5700" from-port="1" to-layer="5702" to-port="0" /> <edge from-layer="5701" from-port="0" to-layer="5702" to-port="2" /> <edge from-layer="5702" from-port="3" to-layer="5703" to-port="0" /> <edge from-layer="5703" from-port="2" to-layer="5704" to-port="0" /> <edge from-layer="5704" from-port="1" to-layer="5706" to-port="0" /> <edge from-layer="5705" from-port="0" to-layer="5706" to-port="1" /> <edge from-layer="5706" from-port="2" to-layer="5708" to-port="2" /> <edge from-layer="5707" from-port="0" to-layer="5708" to-port="3" /> <edge from-layer="5708" from-port="4" to-layer="5711" to-port="1" /> <edge from-layer="5708" from-port="4" to-layer="5716" to-port="2" /> <edge from-layer="5709" from-port="0" to-layer="5711" to-port="2" /> <edge from-layer="5710" from-port="0" to-layer="5711" to-port="3" /> <edge from-layer="5711" from-port="4" to-layer="5713" to-port="0" /> <edge from-layer="5712" from-port="0" to-layer="5713" to-port="1" /> <edge from-layer="5713" from-port="2" to-layer="5717" to-port="0" /> <edge from-layer="5714" from-port="0" to-layer="5716" to-port="1" /> <edge from-layer="5715" from-port="0" to-layer="5716" to-port="3" /> <edge from-layer="5716" from-port="4" to-layer="5717" to-port="1" /> <edge from-layer="5717" from-port="2" to-layer="5725" to-port="0" /> <edge from-layer="5718" from-port="0" to-layer="5721" to-port="0" /> <edge from-layer="5719" from-port="0" to-layer="5721" to-port="1" /> <edge from-layer="5720" from-port="0" to-layer="5721" to-port="3" /> <edge from-layer="5721" from-port="4" to-layer="5723" to-port="0" /> <edge from-layer="5722" from-port="0" to-layer="5723" to-port="2" /> <edge from-layer="5723" from-port="3" to-layer="5724" to-port="0" /> <edge from-layer="5724" from-port="2" to-layer="5725" to-port="1" /> <edge from-layer="5724" from-port="2" to-layer="5757" to-port="1" /> <edge from-layer="5725" from-port="2" to-layer="5726" to-port="1" /> <edge from-layer="5726" from-port="2" to-layer="5731" to-port="0" /> <edge from-layer="5727" from-port="0" to-layer="5730" to-port="1" /> <edge from-layer="5728" from-port="0" to-layer="5730" to-port="2" /> <edge from-layer="5729" from-port="0" to-layer="5730" to-port="3" /> <edge from-layer="5730" from-port="4" to-layer="5731" to-port="1" /> <edge from-layer="5731" from-port="2" to-layer="5787" to-port="0" /> <edge from-layer="5731" from-port="2" to-layer="5825" to-port="0" /> <edge from-layer="5732" from-port="0" to-layer="5735" to-port="1" /> <edge from-layer="5733" from-port="0" to-layer="5735" to-port="2" /> <edge from-layer="5734" from-port="0" to-layer="5735" to-port="3" /> <edge from-layer="5735" from-port="4" to-layer="5750" to-port="0" /> <edge from-layer="5735" from-port="4" to-layer="5736" to-port="0" /> <edge from-layer="5735" from-port="4" to-layer="5755" to-port="0" /> <edge from-layer="5735" from-port="4" to-layer="5739" to-port="0" /> <edge from-layer="5736" from-port="2" to-layer="5758" to-port="0" /> <edge from-layer="5737" from-port="0" to-layer="5747" to-port="0" /> <edge from-layer="5738" from-port="0" to-layer="5747" to-port="1" /> <edge from-layer="5739" from-port="1" to-layer="5741" to-port="0" /> <edge from-layer="5740" from-port="0" to-layer="5741" to-port="2" /> <edge from-layer="5741" from-port="3" to-layer="5742" to-port="0" /> <edge from-layer="5742" from-port="2" to-layer="5743" to-port="0" /> <edge from-layer="5743" from-port="1" to-layer="5745" to-port="0" /> <edge from-layer="5744" from-port="0" to-layer="5745" to-port="1" /> <edge from-layer="5745" from-port="2" to-layer="5747" to-port="2" /> <edge from-layer="5746" from-port="0" to-layer="5747" to-port="3" /> <edge from-layer="5747" from-port="4" to-layer="5755" to-port="2" /> <edge from-layer="5747" from-port="4" to-layer="5750" to-port="1" /> <edge from-layer="5748" from-port="0" to-layer="5750" to-port="2" /> <edge from-layer="5749" from-port="0" to-layer="5750" to-port="3" /> <edge from-layer="5750" from-port="4" to-layer="5752" to-port="0" /> <edge from-layer="5751" from-port="0" to-layer="5752" to-port="1" /> <edge from-layer="5752" from-port="2" to-layer="5756" to-port="0" /> <edge from-layer="5753" from-port="0" to-layer="5755" to-port="1" /> <edge from-layer="5754" from-port="0" to-layer="5755" to-port="3" /> <edge from-layer="5755" from-port="4" to-layer="5756" to-port="1" /> <edge from-layer="5756" from-port="2" to-layer="5757" to-port="0" /> <edge from-layer="5757" from-port="2" to-layer="5758" to-port="1" /> <edge from-layer="5758" from-port="2" to-layer="5763" to-port="0" /> <edge from-layer="5759" from-port="0" to-layer="5762" to-port="1" /> <edge from-layer="5760" from-port="0" to-layer="5762" to-port="2" /> <edge from-layer="5761" from-port="0" to-layer="5762" to-port="3" /> <edge from-layer="5762" from-port="4" to-layer="5763" to-port="1" /> <edge from-layer="5763" from-port="2" to-layer="5764" to-port="1" /> <edge from-layer="5764" from-port="2" to-layer="5825" to-port="1" /> <edge from-layer="5764" from-port="2" to-layer="10139" to-port="0" /> <edge from-layer="5764" from-port="2" to-layer="5784" to-port="0" /> <edge from-layer="5765" from-port="0" to-layer="5770" to-port="0" /> <edge from-layer="5766" from-port="0" to-layer="5769" to-port="1" /> <edge from-layer="5767" from-port="0" to-layer="5769" to-port="2" /> <edge from-layer="5768" from-port="0" to-layer="5769" to-port="3" /> <edge from-layer="5769" from-port="4" to-layer="5770" to-port="1" /> <edge from-layer="5770" from-port="2" to-layer="5771" to-port="0" /> <edge from-layer="5771" from-port="1" to-layer="5773" to-port="0" /> <edge from-layer="5772" from-port="0" to-layer="5773" to-port="2" /> <edge from-layer="5773" from-port="3" to-layer="5780" to-port="0" /> <edge from-layer="5774" from-port="0" to-layer="5777" to-port="1" /> <edge from-layer="5775" from-port="0" to-layer="5777" to-port="2" /> <edge from-layer="5776" from-port="0" to-layer="5777" to-port="3" /> <edge from-layer="5777" from-port="4" to-layer="5779" to-port="0" /> <edge from-layer="5778" from-port="0" to-layer="5779" to-port="1" /> <edge from-layer="5779" from-port="2" to-layer="5780" to-port="1" /> <edge from-layer="5780" from-port="2" to-layer="10138" to-port="0" /> <edge from-layer="5780" from-port="2" to-layer="5825" to-port="2" /> <edge from-layer="5781" from-port="0" to-layer="5799" to-port="0" /> <edge from-layer="5782" from-port="0" to-layer="5794" to-port="0" /> <edge from-layer="5783" from-port="0" to-layer="5797" to-port="1" /> <edge from-layer="5783" from-port="0" to-layer="5794" to-port="1" /> <edge from-layer="5784" from-port="1" to-layer="5786" to-port="0" /> <edge from-layer="5785" from-port="0" to-layer="5786" to-port="2" /> <edge from-layer="5786" from-port="3" to-layer="5796" to-port="0" /> <edge from-layer="5786" from-port="3" to-layer="5790" to-port="0" /> <edge from-layer="5787" from-port="1" to-layer="5789" to-port="0" /> <edge from-layer="5787" from-port="1" to-layer="5814" to-port="0" /> <edge from-layer="5788" from-port="0" to-layer="5789" to-port="2" /> <edge from-layer="5789" from-port="3" to-layer="5790" to-port="1" /> <edge from-layer="5790" from-port="2" to-layer="5792" to-port="0" /> <edge from-layer="5791" from-port="0" to-layer="5796" to-port="1" /> <edge from-layer="5791" from-port="0" to-layer="5792" to-port="1" /> <edge from-layer="5792" from-port="2" to-layer="5794" to-port="2" /> <edge from-layer="5793" from-port="0" to-layer="5797" to-port="3" /> <edge from-layer="5793" from-port="0" to-layer="5794" to-port="3" /> <edge from-layer="5794" from-port="4" to-layer="5799" to-port="1" /> <edge from-layer="5795" from-port="0" to-layer="5797" to-port="0" /> <edge from-layer="5796" from-port="2" to-layer="5797" to-port="2" /> <edge from-layer="5796" from-port="2" to-layer="5804" to-port="2" /> <edge from-layer="5797" from-port="4" to-layer="5799" to-port="2" /> <edge from-layer="5798" from-port="0" to-layer="5799" to-port="3" /> <edge from-layer="5799" from-port="4" to-layer="5806" to-port="0" /> <edge from-layer="5800" from-port="0" to-layer="5806" to-port="1" /> <edge from-layer="5801" from-port="0" to-layer="5804" to-port="0" /> <edge from-layer="5802" from-port="0" to-layer="5804" to-port="1" /> <edge from-layer="5803" from-port="0" to-layer="5804" to-port="3" /> <edge from-layer="5804" from-port="4" to-layer="5806" to-port="2" /> <edge from-layer="5805" from-port="0" to-layer="5806" to-port="3" /> <edge from-layer="5806" from-port="4" to-layer="5809" to-port="0" /> <edge from-layer="5807" from-port="0" to-layer="5809" to-port="1" /> <edge from-layer="5808" from-port="0" to-layer="5809" to-port="2" /> <edge from-layer="5809" from-port="3" to-layer="5810" to-port="0" /> <edge from-layer="5810" from-port="2" to-layer="5811" to-port="0" /> <edge from-layer="5810" from-port="2" to-layer="5824" to-port="0" /> <edge from-layer="5811" from-port="1" to-layer="5823" to-port="0" /> <edge from-layer="5812" from-port="0" to-layer="5814" to-port="1" /> <edge from-layer="5813" from-port="0" to-layer="5814" to-port="2" /> <edge from-layer="5814" from-port="3" to-layer="5818" to-port="0" /> <edge from-layer="5815" from-port="0" to-layer="5818" to-port="1" /> <edge from-layer="5816" from-port="0" to-layer="5818" to-port="2" /> <edge from-layer="5817" from-port="0" to-layer="5818" to-port="3" /> <edge from-layer="5818" from-port="4" to-layer="5820" to-port="0" /> <edge from-layer="5818" from-port="4" to-layer="5822" to-port="2" /> <edge from-layer="5819" from-port="0" to-layer="5820" to-port="1" /> <edge from-layer="5820" from-port="2" to-layer="5822" to-port="0" /> <edge from-layer="5821" from-port="0" to-layer="5822" to-port="1" /> <edge from-layer="5822" from-port="3" to-layer="5823" to-port="1" /> <edge from-layer="5823" from-port="2" to-layer="5824" to-port="1" /> <edge from-layer="5824" from-port="2" to-layer="5825" to-port="3" /> <edge from-layer="5825" from-port="4" to-layer="5827" to-port="0" /> <edge from-layer="5826" from-port="0" to-layer="5827" to-port="1" /> <edge from-layer="5827" from-port="2" to-layer="5829" to-port="0" /> <edge from-layer="5828" from-port="0" to-layer="5829" to-port="1" /> <edge from-layer="5829" from-port="2" to-layer="5837" to-port="0" /> <edge from-layer="5830" from-port="0" to-layer="5831" to-port="0" /> <edge from-layer="5831" from-port="1" to-layer="5834" to-port="0" /> <edge from-layer="5832" from-port="0" to-layer="5833" to-port="0" /> <edge from-layer="5833" from-port="1" to-layer="5834" to-port="1" /> <edge from-layer="5834" from-port="2" to-layer="5836" to-port="0" /> <edge from-layer="5835" from-port="0" to-layer="5836" to-port="1" /> <edge from-layer="5836" from-port="2" to-layer="5837" to-port="1" /> <edge from-layer="5837" from-port="2" to-layer="5839" to-port="0" /> <edge from-layer="5838" from-port="0" to-layer="5839" to-port="1" /> <edge from-layer="5839" from-port="2" to-layer="5840" to-port="1" /> <edge from-layer="5840" from-port="2" to-layer="5841" to-port="0" /> <edge from-layer="5841" from-port="2" to-layer="6072" to-port="1" /> <edge from-layer="5841" from-port="2" to-layer="5843" to-port="0" /> <edge from-layer="5841" from-port="2" to-layer="5870" to-port="0" /> <edge from-layer="5842" from-port="0" to-layer="5843" to-port="1" /> <edge from-layer="5843" from-port="2" to-layer="5845" to-port="0" /> <edge from-layer="5844" from-port="0" to-layer="5845" to-port="1" /> <edge from-layer="5845" from-port="2" to-layer="5847" to-port="0" /> <edge from-layer="5846" from-port="0" to-layer="5847" to-port="1" /> <edge from-layer="5847" from-port="2" to-layer="5855" to-port="0" /> <edge from-layer="5848" from-port="0" to-layer="5849" to-port="0" /> <edge from-layer="5849" from-port="1" to-layer="5852" to-port="0" /> <edge from-layer="5850" from-port="0" to-layer="5851" to-port="0" /> <edge from-layer="5851" from-port="1" to-layer="5852" to-port="1" /> <edge from-layer="5852" from-port="2" to-layer="5854" to-port="0" /> <edge from-layer="5853" from-port="0" to-layer="5854" to-port="1" /> <edge from-layer="5854" from-port="2" to-layer="5855" to-port="1" /> <edge from-layer="5855" from-port="2" to-layer="5857" to-port="0" /> <edge from-layer="5856" from-port="0" to-layer="5857" to-port="1" /> <edge from-layer="5857" from-port="2" to-layer="5858" to-port="0" /> <edge from-layer="5858" from-port="1" to-layer="5866" to-port="0" /> <edge from-layer="5859" from-port="0" to-layer="5860" to-port="0" /> <edge from-layer="5860" from-port="1" to-layer="5863" to-port="0" /> <edge from-layer="5861" from-port="0" to-layer="5862" to-port="0" /> <edge from-layer="5862" from-port="1" to-layer="5863" to-port="1" /> <edge from-layer="5863" from-port="2" to-layer="5865" to-port="0" /> <edge from-layer="5864" from-port="0" to-layer="5865" to-port="1" /> <edge from-layer="5865" from-port="2" to-layer="5866" to-port="1" /> <edge from-layer="5866" from-port="2" to-layer="5868" to-port="0" /> <edge from-layer="5867" from-port="0" to-layer="5868" to-port="1" /> <edge from-layer="5868" from-port="2" to-layer="6071" to-port="0" /> <edge from-layer="5869" from-port="0" to-layer="5870" to-port="1" /> <edge from-layer="5870" from-port="2" to-layer="5872" to-port="0" /> <edge from-layer="5871" from-port="0" to-layer="5872" to-port="1" /> <edge from-layer="5872" from-port="2" to-layer="5874" to-port="0" /> <edge from-layer="5873" from-port="0" to-layer="5874" to-port="1" /> <edge from-layer="5874" from-port="2" to-layer="5882" to-port="0" /> <edge from-layer="5875" from-port="0" to-layer="5876" to-port="0" /> <edge from-layer="5876" from-port="1" to-layer="5879" to-port="0" /> <edge from-layer="5877" from-port="0" to-layer="5878" to-port="0" /> <edge from-layer="5878" from-port="1" to-layer="5879" to-port="1" /> <edge from-layer="5879" from-port="2" to-layer="5881" to-port="0" /> <edge from-layer="5880" from-port="0" to-layer="5881" to-port="1" /> <edge from-layer="5881" from-port="2" to-layer="5882" to-port="1" /> <edge from-layer="5882" from-port="2" to-layer="5884" to-port="0" /> <edge from-layer="5883" from-port="0" to-layer="5884" to-port="1" /> <edge from-layer="5884" from-port="2" to-layer="5886" to-port="0" /> <edge from-layer="5885" from-port="0" to-layer="5886" to-port="1" /> <edge from-layer="5886" from-port="2" to-layer="5902" to-port="0" /> <edge from-layer="5886" from-port="2" to-layer="5890" to-port="0" /> <edge from-layer="5886" from-port="2" to-layer="6008" to-port="0" /> <edge from-layer="5887" from-port="0" to-layer="5890" to-port="1" /> <edge from-layer="5888" from-port="0" to-layer="5890" to-port="2" /> <edge from-layer="5889" from-port="0" to-layer="5890" to-port="3" /> <edge from-layer="5890" from-port="4" to-layer="5892" to-port="0" /> <edge from-layer="5891" from-port="0" to-layer="5892" to-port="1" /> <edge from-layer="5892" from-port="2" to-layer="5961" to-port="0" /> <edge from-layer="5892" from-port="2" to-layer="5896" to-port="0" /> <edge from-layer="5893" from-port="0" to-layer="5896" to-port="1" /> <edge from-layer="5894" from-port="0" to-layer="5896" to-port="2" /> <edge from-layer="5895" from-port="0" to-layer="5896" to-port="3" /> <edge from-layer="5896" from-port="4" to-layer="5942" to-port="0" /> <edge from-layer="5896" from-port="4" to-layer="5928" to-port="0" /> <edge from-layer="5896" from-port="4" to-layer="5947" to-port="0" /> <edge from-layer="5896" from-port="4" to-layer="5931" to-port="0" /> <edge from-layer="5897" from-port="0" to-layer="5924" to-port="0" /> <edge from-layer="5898" from-port="0" to-layer="5924" to-port="1" /> <edge from-layer="5899" from-port="0" to-layer="5902" to-port="1" /> <edge from-layer="5900" from-port="0" to-layer="5902" to-port="2" /> <edge from-layer="5901" from-port="0" to-layer="5902" to-port="3" /> <edge from-layer="5902" from-port="4" to-layer="5904" to-port="0" /> <edge from-layer="5903" from-port="0" to-layer="5904" to-port="1" /> <edge from-layer="5904" from-port="2" to-layer="5966" to-port="0" /> <edge from-layer="5904" from-port="2" to-layer="5993" to-port="0" /> <edge from-layer="5904" from-port="2" to-layer="5905" to-port="0" /> <edge from-layer="5905" from-port="1" to-layer="5907" to-port="0" /> <edge from-layer="5906" from-port="0" to-layer="5907" to-port="2" /> <edge from-layer="5907" from-port="3" to-layer="5920" to-port="0" /> <edge from-layer="5908" from-port="0" to-layer="5913" to-port="0" /> <edge from-layer="5909" from-port="0" to-layer="5912" to-port="1" /> <edge from-layer="5910" from-port="0" to-layer="5912" to-port="2" /> <edge from-layer="5911" from-port="0" to-layer="5912" to-port="3" /> <edge from-layer="5912" from-port="4" to-layer="5913" to-port="1" /> <edge from-layer="5913" from-port="2" to-layer="5914" to-port="0" /> <edge from-layer="5914" from-port="1" to-layer="5916" to-port="0" /> <edge from-layer="5915" from-port="0" to-layer="5916" to-port="2" /> <edge from-layer="5916" from-port="3" to-layer="5995" to-port="0" /> <edge from-layer="5916" from-port="3" to-layer="5917" to-port="0" /> <edge from-layer="5917" from-port="1" to-layer="5919" to-port="0" /> <edge from-layer="5918" from-port="0" to-layer="5919" to-port="2" /> <edge from-layer="5919" from-port="3" to-layer="5920" to-port="1" /> <edge from-layer="5920" from-port="2" to-layer="5922" to-port="0" /> <edge from-layer="5921" from-port="0" to-layer="5922" to-port="1" /> <edge from-layer="5922" from-port="2" to-layer="5952" to-port="2" /> <edge from-layer="5922" from-port="2" to-layer="5924" to-port="2" /> <edge from-layer="5923" from-port="0" to-layer="5924" to-port="3" /> <edge from-layer="5924" from-port="4" to-layer="5926" to-port="0" /> <edge from-layer="5925" from-port="0" to-layer="5926" to-port="2" /> <edge from-layer="5926" from-port="3" to-layer="5927" to-port="0" /> <edge from-layer="5927" from-port="2" to-layer="5928" to-port="1" /> <edge from-layer="5927" from-port="2" to-layer="5967" to-port="1" /> <edge from-layer="5928" from-port="2" to-layer="5957" to-port="0" /> <edge from-layer="5929" from-port="0" to-layer="5939" to-port="0" /> <edge from-layer="5930" from-port="0" to-layer="5939" to-port="1" /> <edge from-layer="5931" from-port="1" to-layer="5933" to-port="0" /> <edge from-layer="5932" from-port="0" to-layer="5933" to-port="2" /> <edge from-layer="5933" from-port="3" to-layer="5934" to-port="0" /> <edge from-layer="5934" from-port="2" to-layer="5935" to-port="0" /> <edge from-layer="5935" from-port="1" to-layer="5937" to-port="0" /> <edge from-layer="5936" from-port="0" to-layer="5937" to-port="1" /> <edge from-layer="5937" from-port="2" to-layer="5939" to-port="2" /> <edge from-layer="5938" from-port="0" to-layer="5939" to-port="3" /> <edge from-layer="5939" from-port="4" to-layer="5947" to-port="2" /> <edge from-layer="5939" from-port="4" to-layer="5942" to-port="1" /> <edge from-layer="5940" from-port="0" to-layer="5942" to-port="2" /> <edge from-layer="5941" from-port="0" to-layer="5942" to-port="3" /> <edge from-layer="5942" from-port="4" to-layer="5944" to-port="0" /> <edge from-layer="5943" from-port="0" to-layer="5944" to-port="1" /> <edge from-layer="5944" from-port="2" to-layer="5948" to-port="0" /> <edge from-layer="5945" from-port="0" to-layer="5947" to-port="1" /> <edge from-layer="5946" from-port="0" to-layer="5947" to-port="3" /> <edge from-layer="5947" from-port="4" to-layer="5948" to-port="1" /> <edge from-layer="5948" from-port="2" to-layer="5956" to-port="0" /> <edge from-layer="5949" from-port="0" to-layer="5952" to-port="0" /> <edge from-layer="5950" from-port="0" to-layer="5952" to-port="1" /> <edge from-layer="5951" from-port="0" to-layer="5952" to-port="3" /> <edge from-layer="5952" from-port="4" to-layer="5954" to-port="0" /> <edge from-layer="5953" from-port="0" to-layer="5954" to-port="2" /> <edge from-layer="5954" from-port="3" to-layer="5955" to-port="0" /> <edge from-layer="5955" from-port="2" to-layer="5956" to-port="1" /> <edge from-layer="5955" from-port="2" to-layer="5988" to-port="1" /> <edge from-layer="5956" from-port="2" to-layer="5957" to-port="1" /> <edge from-layer="5957" from-port="2" to-layer="5962" to-port="0" /> <edge from-layer="5958" from-port="0" to-layer="5961" to-port="1" /> <edge from-layer="5959" from-port="0" to-layer="5961" to-port="2" /> <edge from-layer="5960" from-port="0" to-layer="5961" to-port="3" /> <edge from-layer="5961" from-port="4" to-layer="5962" to-port="1" /> <edge from-layer="5962" from-port="2" to-layer="6056" to-port="0" /> <edge from-layer="5962" from-port="2" to-layer="6018" to-port="0" /> <edge from-layer="5963" from-port="0" to-layer="5966" to-port="1" /> <edge from-layer="5964" from-port="0" to-layer="5966" to-port="2" /> <edge from-layer="5965" from-port="0" to-layer="5966" to-port="3" /> <edge from-layer="5966" from-port="4" to-layer="5967" to-port="0" /> <edge from-layer="5966" from-port="4" to-layer="5981" to-port="0" /> <edge from-layer="5966" from-port="4" to-layer="5970" to-port="0" /> <edge from-layer="5966" from-port="4" to-layer="5986" to-port="0" /> <edge from-layer="5967" from-port="2" to-layer="5989" to-port="0" /> <edge from-layer="5968" from-port="0" to-layer="5978" to-port="0" /> <edge from-layer="5969" from-port="0" to-layer="5978" to-port="1" /> <edge from-layer="5970" from-port="1" to-layer="5972" to-port="0" /> <edge from-layer="5971" from-port="0" to-layer="5972" to-port="2" /> <edge from-layer="5972" from-port="3" to-layer="5973" to-port="0" /> <edge from-layer="5973" from-port="2" to-layer="5974" to-port="0" /> <edge from-layer="5974" from-port="1" to-layer="5976" to-port="0" /> <edge from-layer="5975" from-port="0" to-layer="5976" to-port="1" /> <edge from-layer="5976" from-port="2" to-layer="5978" to-port="2" /> <edge from-layer="5977" from-port="0" to-layer="5978" to-port="3" /> <edge from-layer="5978" from-port="4" to-layer="5981" to-port="1" /> <edge from-layer="5978" from-port="4" to-layer="5986" to-port="2" /> <edge from-layer="5979" from-port="0" to-layer="5981" to-port="2" /> <edge from-layer="5980" from-port="0" to-layer="5981" to-port="3" /> <edge from-layer="5981" from-port="4" to-layer="5983" to-port="0" /> <edge from-layer="5982" from-port="0" to-layer="5983" to-port="1" /> <edge from-layer="5983" from-port="2" to-layer="5987" to-port="0" /> <edge from-layer="5984" from-port="0" to-layer="5986" to-port="1" /> <edge from-layer="5985" from-port="0" to-layer="5986" to-port="3" /> <edge from-layer="5986" from-port="4" to-layer="5987" to-port="1" /> <edge from-layer="5987" from-port="2" to-layer="5988" to-port="0" /> <edge from-layer="5988" from-port="2" to-layer="5989" to-port="1" /> <edge from-layer="5989" from-port="2" to-layer="5994" to-port="0" /> <edge from-layer="5990" from-port="0" to-layer="5993" to-port="1" /> <edge from-layer="5991" from-port="0" to-layer="5993" to-port="2" /> <edge from-layer="5992" from-port="0" to-layer="5993" to-port="3" /> <edge from-layer="5993" from-port="4" to-layer="5994" to-port="1" /> <edge from-layer="5994" from-port="2" to-layer="5995" to-port="1" /> <edge from-layer="5995" from-port="2" to-layer="10137" to-port="0" /> <edge from-layer="5995" from-port="2" to-layer="6056" to-port="1" /> <edge from-layer="5995" from-port="2" to-layer="6015" to-port="0" /> <edge from-layer="5996" from-port="0" to-layer="6001" to-port="0" /> <edge from-layer="5997" from-port="0" to-layer="6000" to-port="1" /> <edge from-layer="5998" from-port="0" to-layer="6000" to-port="2" /> <edge from-layer="5999" from-port="0" to-layer="6000" to-port="3" /> <edge from-layer="6000" from-port="4" to-layer="6001" to-port="1" /> <edge from-layer="6001" from-port="2" to-layer="6002" to-port="0" /> <edge from-layer="6002" from-port="1" to-layer="6004" to-port="0" /> <edge from-layer="6003" from-port="0" to-layer="6004" to-port="2" /> <edge from-layer="6004" from-port="3" to-layer="6011" to-port="0" /> <edge from-layer="6005" from-port="0" to-layer="6008" to-port="1" /> <edge from-layer="6006" from-port="0" to-layer="6008" to-port="2" /> <edge from-layer="6007" from-port="0" to-layer="6008" to-port="3" /> <edge from-layer="6008" from-port="4" to-layer="6010" to-port="0" /> <edge from-layer="6009" from-port="0" to-layer="6010" to-port="1" /> <edge from-layer="6010" from-port="2" to-layer="6011" to-port="1" /> <edge from-layer="6011" from-port="2" to-layer="6056" to-port="2" /> <edge from-layer="6011" from-port="2" to-layer="10136" to-port="0" /> <edge from-layer="6012" from-port="0" to-layer="6030" to-port="0" /> <edge from-layer="6013" from-port="0" to-layer="6025" to-port="0" /> <edge from-layer="6014" from-port="0" to-layer="6025" to-port="1" /> <edge from-layer="6014" from-port="0" to-layer="6028" to-port="1" /> <edge from-layer="6015" from-port="1" to-layer="6017" to-port="0" /> <edge from-layer="6016" from-port="0" to-layer="6017" to-port="2" /> <edge from-layer="6017" from-port="3" to-layer="6027" to-port="0" /> <edge from-layer="6017" from-port="3" to-layer="6021" to-port="0" /> <edge from-layer="6018" from-port="1" to-layer="6045" to-port="0" /> <edge from-layer="6018" from-port="1" to-layer="6020" to-port="0" /> <edge from-layer="6019" from-port="0" to-layer="6020" to-port="2" /> <edge from-layer="6020" from-port="3" to-layer="6021" to-port="1" /> <edge from-layer="6021" from-port="2" to-layer="6023" to-port="0" /> <edge from-layer="6022" from-port="0" to-layer="6027" to-port="1" /> <edge from-layer="6022" from-port="0" to-layer="6023" to-port="1" /> <edge from-layer="6023" from-port="2" to-layer="6025" to-port="2" /> <edge from-layer="6024" from-port="0" to-layer="6025" to-port="3" /> <edge from-layer="6024" from-port="0" to-layer="6028" to-port="3" /> <edge from-layer="6025" from-port="4" to-layer="6030" to-port="1" /> <edge from-layer="6026" from-port="0" to-layer="6028" to-port="0" /> <edge from-layer="6027" from-port="2" to-layer="6028" to-port="2" /> <edge from-layer="6027" from-port="2" to-layer="6035" to-port="2" /> <edge from-layer="6028" from-port="4" to-layer="6030" to-port="2" /> <edge from-layer="6029" from-port="0" to-layer="6030" to-port="3" /> <edge from-layer="6030" from-port="4" to-layer="6037" to-port="0" /> <edge from-layer="6031" from-port="0" to-layer="6037" to-port="1" /> <edge from-layer="6032" from-port="0" to-layer="6035" to-port="0" /> <edge from-layer="6033" from-port="0" to-layer="6035" to-port="1" /> <edge from-layer="6034" from-port="0" to-layer="6035" to-port="3" /> <edge from-layer="6035" from-port="4" to-layer="6037" to-port="2" /> <edge from-layer="6036" from-port="0" to-layer="6037" to-port="3" /> <edge from-layer="6037" from-port="4" to-layer="6040" to-port="0" /> <edge from-layer="6038" from-port="0" to-layer="6040" to-port="1" /> <edge from-layer="6039" from-port="0" to-layer="6040" to-port="2" /> <edge from-layer="6040" from-port="3" to-layer="6041" to-port="0" /> <edge from-layer="6041" from-port="2" to-layer="6055" to-port="0" /> <edge from-layer="6041" from-port="2" to-layer="6042" to-port="0" /> <edge from-layer="6042" from-port="1" to-layer="6054" to-port="0" /> <edge from-layer="6043" from-port="0" to-layer="6045" to-port="1" /> <edge from-layer="6044" from-port="0" to-layer="6045" to-port="2" /> <edge from-layer="6045" from-port="3" to-layer="6049" to-port="0" /> <edge from-layer="6046" from-port="0" to-layer="6049" to-port="1" /> <edge from-layer="6047" from-port="0" to-layer="6049" to-port="2" /> <edge from-layer="6048" from-port="0" to-layer="6049" to-port="3" /> <edge from-layer="6049" from-port="4" to-layer="6053" to-port="2" /> <edge from-layer="6049" from-port="4" to-layer="6051" to-port="0" /> <edge from-layer="6050" from-port="0" to-layer="6051" to-port="1" /> <edge from-layer="6051" from-port="2" to-layer="6053" to-port="0" /> <edge from-layer="6052" from-port="0" to-layer="6053" to-port="1" /> <edge from-layer="6053" from-port="3" to-layer="6054" to-port="1" /> <edge from-layer="6054" from-port="2" to-layer="6055" to-port="1" /> <edge from-layer="6055" from-port="2" to-layer="6056" to-port="3" /> <edge from-layer="6056" from-port="4" to-layer="6058" to-port="0" /> <edge from-layer="6057" from-port="0" to-layer="6058" to-port="1" /> <edge from-layer="6058" from-port="2" to-layer="6060" to-port="0" /> <edge from-layer="6059" from-port="0" to-layer="6060" to-port="1" /> <edge from-layer="6060" from-port="2" to-layer="6068" to-port="0" /> <edge from-layer="6061" from-port="0" to-layer="6062" to-port="0" /> <edge from-layer="6062" from-port="1" to-layer="6065" to-port="0" /> <edge from-layer="6063" from-port="0" to-layer="6064" to-port="0" /> <edge from-layer="6064" from-port="1" to-layer="6065" to-port="1" /> <edge from-layer="6065" from-port="2" to-layer="6067" to-port="0" /> <edge from-layer="6066" from-port="0" to-layer="6067" to-port="1" /> <edge from-layer="6067" from-port="2" to-layer="6068" to-port="1" /> <edge from-layer="6068" from-port="2" to-layer="6070" to-port="0" /> <edge from-layer="6069" from-port="0" to-layer="6070" to-port="1" /> <edge from-layer="6070" from-port="2" to-layer="6071" to-port="1" /> <edge from-layer="6071" from-port="2" to-layer="6072" to-port="0" /> <edge from-layer="6072" from-port="2" to-layer="6303" to-port="1" /> <edge from-layer="6072" from-port="2" to-layer="6074" to-port="0" /> <edge from-layer="6072" from-port="2" to-layer="6101" to-port="0" /> <edge from-layer="6073" from-port="0" to-layer="6074" to-port="1" /> <edge from-layer="6074" from-port="2" to-layer="6076" to-port="0" /> <edge from-layer="6075" from-port="0" to-layer="6076" to-port="1" /> <edge from-layer="6076" from-port="2" to-layer="6078" to-port="0" /> <edge from-layer="6077" from-port="0" to-layer="6078" to-port="1" /> <edge from-layer="6078" from-port="2" to-layer="6086" to-port="0" /> <edge from-layer="6079" from-port="0" to-layer="6080" to-port="0" /> <edge from-layer="6080" from-port="1" to-layer="6083" to-port="0" /> <edge from-layer="6081" from-port="0" to-layer="6082" to-port="0" /> <edge from-layer="6082" from-port="1" to-layer="6083" to-port="1" /> <edge from-layer="6083" from-port="2" to-layer="6085" to-port="0" /> <edge from-layer="6084" from-port="0" to-layer="6085" to-port="1" /> <edge from-layer="6085" from-port="2" to-layer="6086" to-port="1" /> <edge from-layer="6086" from-port="2" to-layer="6088" to-port="0" /> <edge from-layer="6087" from-port="0" to-layer="6088" to-port="1" /> <edge from-layer="6088" from-port="2" to-layer="6089" to-port="0" /> <edge from-layer="6089" from-port="1" to-layer="6097" to-port="0" /> <edge from-layer="6090" from-port="0" to-layer="6091" to-port="0" /> <edge from-layer="6091" from-port="1" to-layer="6094" to-port="0" /> <edge from-layer="6092" from-port="0" to-layer="6093" to-port="0" /> <edge from-layer="6093" from-port="1" to-layer="6094" to-port="1" /> <edge from-layer="6094" from-port="2" to-layer="6096" to-port="0" /> <edge from-layer="6095" from-port="0" to-layer="6096" to-port="1" /> <edge from-layer="6096" from-port="2" to-layer="6097" to-port="1" /> <edge from-layer="6097" from-port="2" to-layer="6099" to-port="0" /> <edge from-layer="6098" from-port="0" to-layer="6099" to-port="1" /> <edge from-layer="6099" from-port="2" to-layer="6302" to-port="0" /> <edge from-layer="6100" from-port="0" to-layer="6101" to-port="1" /> <edge from-layer="6101" from-port="2" to-layer="6103" to-port="0" /> <edge from-layer="6102" from-port="0" to-layer="6103" to-port="1" /> <edge from-layer="6103" from-port="2" to-layer="6105" to-port="0" /> <edge from-layer="6104" from-port="0" to-layer="6105" to-port="1" /> <edge from-layer="6105" from-port="2" to-layer="6113" to-port="0" /> <edge from-layer="6106" from-port="0" to-layer="6107" to-port="0" /> <edge from-layer="6107" from-port="1" to-layer="6110" to-port="0" /> <edge from-layer="6108" from-port="0" to-layer="6109" to-port="0" /> <edge from-layer="6109" from-port="1" to-layer="6110" to-port="1" /> <edge from-layer="6110" from-port="2" to-layer="6112" to-port="0" /> <edge from-layer="6111" from-port="0" to-layer="6112" to-port="1" /> <edge from-layer="6112" from-port="2" to-layer="6113" to-port="1" /> <edge from-layer="6113" from-port="2" to-layer="6115" to-port="0" /> <edge from-layer="6114" from-port="0" to-layer="6115" to-port="1" /> <edge from-layer="6115" from-port="2" to-layer="6117" to-port="0" /> <edge from-layer="6116" from-port="0" to-layer="6117" to-port="1" /> <edge from-layer="6117" from-port="2" to-layer="6133" to-port="0" /> <edge from-layer="6117" from-port="2" to-layer="6239" to-port="0" /> <edge from-layer="6117" from-port="2" to-layer="6121" to-port="0" /> <edge from-layer="6118" from-port="0" to-layer="6121" to-port="1" /> <edge from-layer="6119" from-port="0" to-layer="6121" to-port="2" /> <edge from-layer="6120" from-port="0" to-layer="6121" to-port="3" /> <edge from-layer="6121" from-port="4" to-layer="6123" to-port="0" /> <edge from-layer="6122" from-port="0" to-layer="6123" to-port="1" /> <edge from-layer="6123" from-port="2" to-layer="6127" to-port="0" /> <edge from-layer="6123" from-port="2" to-layer="6192" to-port="0" /> <edge from-layer="6124" from-port="0" to-layer="6127" to-port="1" /> <edge from-layer="6125" from-port="0" to-layer="6127" to-port="2" /> <edge from-layer="6126" from-port="0" to-layer="6127" to-port="3" /> <edge from-layer="6127" from-port="4" to-layer="6173" to-port="0" /> <edge from-layer="6127" from-port="4" to-layer="6178" to-port="0" /> <edge from-layer="6127" from-port="4" to-layer="6162" to-port="0" /> <edge from-layer="6127" from-port="4" to-layer="6159" to-port="0" /> <edge from-layer="6128" from-port="0" to-layer="6155" to-port="0" /> <edge from-layer="6129" from-port="0" to-layer="6155" to-port="1" /> <edge from-layer="6130" from-port="0" to-layer="6133" to-port="1" /> <edge from-layer="6131" from-port="0" to-layer="6133" to-port="2" /> <edge from-layer="6132" from-port="0" to-layer="6133" to-port="3" /> <edge from-layer="6133" from-port="4" to-layer="6135" to-port="0" /> <edge from-layer="6134" from-port="0" to-layer="6135" to-port="1" /> <edge from-layer="6135" from-port="2" to-layer="6136" to-port="0" /> <edge from-layer="6135" from-port="2" to-layer="6197" to-port="0" /> <edge from-layer="6135" from-port="2" to-layer="6224" to-port="0" /> <edge from-layer="6136" from-port="1" to-layer="6138" to-port="0" /> <edge from-layer="6137" from-port="0" to-layer="6138" to-port="2" /> <edge from-layer="6138" from-port="3" to-layer="6151" to-port="0" /> <edge from-layer="6139" from-port="0" to-layer="6144" to-port="0" /> <edge from-layer="6140" from-port="0" to-layer="6143" to-port="1" /> <edge from-layer="6141" from-port="0" to-layer="6143" to-port="2" /> <edge from-layer="6142" from-port="0" to-layer="6143" to-port="3" /> <edge from-layer="6143" from-port="4" to-layer="6144" to-port="1" /> <edge from-layer="6144" from-port="2" to-layer="6145" to-port="0" /> <edge from-layer="6145" from-port="1" to-layer="6147" to-port="0" /> <edge from-layer="6146" from-port="0" to-layer="6147" to-port="2" /> <edge from-layer="6147" from-port="3" to-layer="6148" to-port="0" /> <edge from-layer="6147" from-port="3" to-layer="6226" to-port="0" /> <edge from-layer="6148" from-port="1" to-layer="6150" to-port="0" /> <edge from-layer="6149" from-port="0" to-layer="6150" to-port="2" /> <edge from-layer="6150" from-port="3" to-layer="6151" to-port="1" /> <edge from-layer="6151" from-port="2" to-layer="6153" to-port="0" /> <edge from-layer="6152" from-port="0" to-layer="6153" to-port="1" /> <edge from-layer="6153" from-port="2" to-layer="6155" to-port="2" /> <edge from-layer="6153" from-port="2" to-layer="6183" to-port="2" /> <edge from-layer="6154" from-port="0" to-layer="6155" to-port="3" /> <edge from-layer="6155" from-port="4" to-layer="6157" to-port="0" /> <edge from-layer="6156" from-port="0" to-layer="6157" to-port="2" /> <edge from-layer="6157" from-port="3" to-layer="6158" to-port="0" /> <edge from-layer="6158" from-port="2" to-layer="6159" to-port="1" /> <edge from-layer="6158" from-port="2" to-layer="6198" to-port="1" /> <edge from-layer="6159" from-port="2" to-layer="6188" to-port="0" /> <edge from-layer="6160" from-port="0" to-layer="6170" to-port="0" /> <edge from-layer="6161" from-port="0" to-layer="6170" to-port="1" /> <edge from-layer="6162" from-port="1" to-layer="6164" to-port="0" /> <edge from-layer="6163" from-port="0" to-layer="6164" to-port="2" /> <edge from-layer="6164" from-port="3" to-layer="6165" to-port="0" /> <edge from-layer="6165" from-port="2" to-layer="6166" to-port="0" /> <edge from-layer="6166" from-port="1" to-layer="6168" to-port="0" /> <edge from-layer="6167" from-port="0" to-layer="6168" to-port="1" /> <edge from-layer="6168" from-port="2" to-layer="6170" to-port="2" /> <edge from-layer="6169" from-port="0" to-layer="6170" to-port="3" /> <edge from-layer="6170" from-port="4" to-layer="6173" to-port="1" /> <edge from-layer="6170" from-port="4" to-layer="6178" to-port="2" /> <edge from-layer="6171" from-port="0" to-layer="6173" to-port="2" /> <edge from-layer="6172" from-port="0" to-layer="6173" to-port="3" /> <edge from-layer="6173" from-port="4" to-layer="6175" to-port="0" /> <edge from-layer="6174" from-port="0" to-layer="6175" to-port="1" /> <edge from-layer="6175" from-port="2" to-layer="6179" to-port="0" /> <edge from-layer="6176" from-port="0" to-layer="6178" to-port="1" /> <edge from-layer="6177" from-port="0" to-layer="6178" to-port="3" /> <edge from-layer="6178" from-port="4" to-layer="6179" to-port="1" /> <edge from-layer="6179" from-port="2" to-layer="6187" to-port="0" /> <edge from-layer="6180" from-port="0" to-layer="6183" to-port="0" /> <edge from-layer="6181" from-port="0" to-layer="6183" to-port="1" /> <edge from-layer="6182" from-port="0" to-layer="6183" to-port="3" /> <edge from-layer="6183" from-port="4" to-layer="6185" to-port="0" /> <edge from-layer="6184" from-port="0" to-layer="6185" to-port="2" /> <edge from-layer="6185" from-port="3" to-layer="6186" to-port="0" /> <edge from-layer="6186" from-port="2" to-layer="6219" to-port="1" /> <edge from-layer="6186" from-port="2" to-layer="6187" to-port="1" /> <edge from-layer="6187" from-port="2" to-layer="6188" to-port="1" /> <edge from-layer="6188" from-port="2" to-layer="6193" to-port="0" /> <edge from-layer="6189" from-port="0" to-layer="6192" to-port="1" /> <edge from-layer="6190" from-port="0" to-layer="6192" to-port="2" /> <edge from-layer="6191" from-port="0" to-layer="6192" to-port="3" /> <edge from-layer="6192" from-port="4" to-layer="6193" to-port="1" /> <edge from-layer="6193" from-port="2" to-layer="6287" to-port="0" /> <edge from-layer="6193" from-port="2" to-layer="6249" to-port="0" /> <edge from-layer="6194" from-port="0" to-layer="6197" to-port="1" /> <edge from-layer="6195" from-port="0" to-layer="6197" to-port="2" /> <edge from-layer="6196" from-port="0" to-layer="6197" to-port="3" /> <edge from-layer="6197" from-port="4" to-layer="6198" to-port="0" /> <edge from-layer="6197" from-port="4" to-layer="6201" to-port="0" /> <edge from-layer="6197" from-port="4" to-layer="6217" to-port="0" /> <edge from-layer="6197" from-port="4" to-layer="6212" to-port="0" /> <edge from-layer="6198" from-port="2" to-layer="6220" to-port="0" /> <edge from-layer="6199" from-port="0" to-layer="6209" to-port="0" /> <edge from-layer="6200" from-port="0" to-layer="6209" to-port="1" /> <edge from-layer="6201" from-port="1" to-layer="6203" to-port="0" /> <edge from-layer="6202" from-port="0" to-layer="6203" to-port="2" /> <edge from-layer="6203" from-port="3" to-layer="6204" to-port="0" /> <edge from-layer="6204" from-port="2" to-layer="6205" to-port="0" /> <edge from-layer="6205" from-port="1" to-layer="6207" to-port="0" /> <edge from-layer="6206" from-port="0" to-layer="6207" to-port="1" /> <edge from-layer="6207" from-port="2" to-layer="6209" to-port="2" /> <edge from-layer="6208" from-port="0" to-layer="6209" to-port="3" /> <edge from-layer="6209" from-port="4" to-layer="6212" to-port="1" /> <edge from-layer="6209" from-port="4" to-layer="6217" to-port="2" /> <edge from-layer="6210" from-port="0" to-layer="6212" to-port="2" /> <edge from-layer="6211" from-port="0" to-layer="6212" to-port="3" /> <edge from-layer="6212" from-port="4" to-layer="6214" to-port="0" /> <edge from-layer="6213" from-port="0" to-layer="6214" to-port="1" /> <edge from-layer="6214" from-port="2" to-layer="6218" to-port="0" /> <edge from-layer="6215" from-port="0" to-layer="6217" to-port="1" /> <edge from-layer="6216" from-port="0" to-layer="6217" to-port="3" /> <edge from-layer="6217" from-port="4" to-layer="6218" to-port="1" /> <edge from-layer="6218" from-port="2" to-layer="6219" to-port="0" /> <edge from-layer="6219" from-port="2" to-layer="6220" to-port="1" /> <edge from-layer="6220" from-port="2" to-layer="6225" to-port="0" /> <edge from-layer="6221" from-port="0" to-layer="6224" to-port="1" /> <edge from-layer="6222" from-port="0" to-layer="6224" to-port="2" /> <edge from-layer="6223" from-port="0" to-layer="6224" to-port="3" /> <edge from-layer="6224" from-port="4" to-layer="6225" to-port="1" /> <edge from-layer="6225" from-port="2" to-layer="6226" to-port="1" /> <edge from-layer="6226" from-port="2" to-layer="6287" to-port="1" /> <edge from-layer="6226" from-port="2" to-layer="10135" to-port="0" /> <edge from-layer="6226" from-port="2" to-layer="6246" to-port="0" /> <edge from-layer="6227" from-port="0" to-layer="6232" to-port="0" /> <edge from-layer="6228" from-port="0" to-layer="6231" to-port="1" /> <edge from-layer="6229" from-port="0" to-layer="6231" to-port="2" /> <edge from-layer="6230" from-port="0" to-layer="6231" to-port="3" /> <edge from-layer="6231" from-port="4" to-layer="6232" to-port="1" /> <edge from-layer="6232" from-port="2" to-layer="6233" to-port="0" /> <edge from-layer="6233" from-port="1" to-layer="6235" to-port="0" /> <edge from-layer="6234" from-port="0" to-layer="6235" to-port="2" /> <edge from-layer="6235" from-port="3" to-layer="6242" to-port="0" /> <edge from-layer="6236" from-port="0" to-layer="6239" to-port="1" /> <edge from-layer="6237" from-port="0" to-layer="6239" to-port="2" /> <edge from-layer="6238" from-port="0" to-layer="6239" to-port="3" /> <edge from-layer="6239" from-port="4" to-layer="6241" to-port="0" /> <edge from-layer="6240" from-port="0" to-layer="6241" to-port="1" /> <edge from-layer="6241" from-port="2" to-layer="6242" to-port="1" /> <edge from-layer="6242" from-port="2" to-layer="10134" to-port="0" /> <edge from-layer="6242" from-port="2" to-layer="6287" to-port="2" /> <edge from-layer="6243" from-port="0" to-layer="6261" to-port="0" /> <edge from-layer="6244" from-port="0" to-layer="6256" to-port="0" /> <edge from-layer="6245" from-port="0" to-layer="6259" to-port="1" /> <edge from-layer="6245" from-port="0" to-layer="6256" to-port="1" /> <edge from-layer="6246" from-port="1" to-layer="6248" to-port="0" /> <edge from-layer="6247" from-port="0" to-layer="6248" to-port="2" /> <edge from-layer="6248" from-port="3" to-layer="6258" to-port="0" /> <edge from-layer="6248" from-port="3" to-layer="6252" to-port="0" /> <edge from-layer="6249" from-port="1" to-layer="6276" to-port="0" /> <edge from-layer="6249" from-port="1" to-layer="6251" to-port="0" /> <edge from-layer="6250" from-port="0" to-layer="6251" to-port="2" /> <edge from-layer="6251" from-port="3" to-layer="6252" to-port="1" /> <edge from-layer="6252" from-port="2" to-layer="6254" to-port="0" /> <edge from-layer="6253" from-port="0" to-layer="6258" to-port="1" /> <edge from-layer="6253" from-port="0" to-layer="6254" to-port="1" /> <edge from-layer="6254" from-port="2" to-layer="6256" to-port="2" /> <edge from-layer="6255" from-port="0" to-layer="6256" to-port="3" /> <edge from-layer="6255" from-port="0" to-layer="6259" to-port="3" /> <edge from-layer="6256" from-port="4" to-layer="6261" to-port="1" /> <edge from-layer="6257" from-port="0" to-layer="6259" to-port="0" /> <edge from-layer="6258" from-port="2" to-layer="6259" to-port="2" /> <edge from-layer="6258" from-port="2" to-layer="6266" to-port="2" /> <edge from-layer="6259" from-port="4" to-layer="6261" to-port="2" /> <edge from-layer="6260" from-port="0" to-layer="6261" to-port="3" /> <edge from-layer="6261" from-port="4" to-layer="6268" to-port="0" /> <edge from-layer="6262" from-port="0" to-layer="6268" to-port="1" /> <edge from-layer="6263" from-port="0" to-layer="6266" to-port="0" /> <edge from-layer="6264" from-port="0" to-layer="6266" to-port="1" /> <edge from-layer="6265" from-port="0" to-layer="6266" to-port="3" /> <edge from-layer="6266" from-port="4" to-layer="6268" to-port="2" /> <edge from-layer="6267" from-port="0" to-layer="6268" to-port="3" /> <edge from-layer="6268" from-port="4" to-layer="6271" to-port="0" /> <edge from-layer="6269" from-port="0" to-layer="6271" to-port="1" /> <edge from-layer="6270" from-port="0" to-layer="6271" to-port="2" /> <edge from-layer="6271" from-port="3" to-layer="6272" to-port="0" /> <edge from-layer="6272" from-port="2" to-layer="6286" to-port="0" /> <edge from-layer="6272" from-port="2" to-layer="6273" to-port="0" /> <edge from-layer="6273" from-port="1" to-layer="6285" to-port="0" /> <edge from-layer="6274" from-port="0" to-layer="6276" to-port="1" /> <edge from-layer="6275" from-port="0" to-layer="6276" to-port="2" /> <edge from-layer="6276" from-port="3" to-layer="6280" to-port="0" /> <edge from-layer="6277" from-port="0" to-layer="6280" to-port="1" /> <edge from-layer="6278" from-port="0" to-layer="6280" to-port="2" /> <edge from-layer="6279" from-port="0" to-layer="6280" to-port="3" /> <edge from-layer="6280" from-port="4" to-layer="6284" to-port="2" /> <edge from-layer="6280" from-port="4" to-layer="6282" to-port="0" /> <edge from-layer="6281" from-port="0" to-layer="6282" to-port="1" /> <edge from-layer="6282" from-port="2" to-layer="6284" to-port="0" /> <edge from-layer="6283" from-port="0" to-layer="6284" to-port="1" /> <edge from-layer="6284" from-port="3" to-layer="6285" to-port="1" /> <edge from-layer="6285" from-port="2" to-layer="6286" to-port="1" /> <edge from-layer="6286" from-port="2" to-layer="6287" to-port="3" /> <edge from-layer="6287" from-port="4" to-layer="6289" to-port="0" /> <edge from-layer="6288" from-port="0" to-layer="6289" to-port="1" /> <edge from-layer="6289" from-port="2" to-layer="6291" to-port="0" /> <edge from-layer="6290" from-port="0" to-layer="6291" to-port="1" /> <edge from-layer="6291" from-port="2" to-layer="6299" to-port="0" /> <edge from-layer="6292" from-port="0" to-layer="6293" to-port="0" /> <edge from-layer="6293" from-port="1" to-layer="6296" to-port="0" /> <edge from-layer="6294" from-port="0" to-layer="6295" to-port="0" /> <edge from-layer="6295" from-port="1" to-layer="6296" to-port="1" /> <edge from-layer="6296" from-port="2" to-layer="6298" to-port="0" /> <edge from-layer="6297" from-port="0" to-layer="6298" to-port="1" /> <edge from-layer="6298" from-port="2" to-layer="6299" to-port="1" /> <edge from-layer="6299" from-port="2" to-layer="6301" to-port="0" /> <edge from-layer="6300" from-port="0" to-layer="6301" to-port="1" /> <edge from-layer="6301" from-port="2" to-layer="6302" to-port="1" /> <edge from-layer="6302" from-port="2" to-layer="6303" to-port="0" /> <edge from-layer="6303" from-port="2" to-layer="6305" to-port="0" /> <edge from-layer="6303" from-port="2" to-layer="6332" to-port="0" /> <edge from-layer="6303" from-port="2" to-layer="6534" to-port="1" /> <edge from-layer="6304" from-port="0" to-layer="6305" to-port="1" /> <edge from-layer="6305" from-port="2" to-layer="6307" to-port="0" /> <edge from-layer="6306" from-port="0" to-layer="6307" to-port="1" /> <edge from-layer="6307" from-port="2" to-layer="6309" to-port="0" /> <edge from-layer="6308" from-port="0" to-layer="6309" to-port="1" /> <edge from-layer="6309" from-port="2" to-layer="6317" to-port="0" /> <edge from-layer="6310" from-port="0" to-layer="6311" to-port="0" /> <edge from-layer="6311" from-port="1" to-layer="6314" to-port="0" /> <edge from-layer="6312" from-port="0" to-layer="6313" to-port="0" /> <edge from-layer="6313" from-port="1" to-layer="6314" to-port="1" /> <edge from-layer="6314" from-port="2" to-layer="6316" to-port="0" /> <edge from-layer="6315" from-port="0" to-layer="6316" to-port="1" /> <edge from-layer="6316" from-port="2" to-layer="6317" to-port="1" /> <edge from-layer="6317" from-port="2" to-layer="6319" to-port="0" /> <edge from-layer="6318" from-port="0" to-layer="6319" to-port="1" /> <edge from-layer="6319" from-port="2" to-layer="6320" to-port="0" /> <edge from-layer="6320" from-port="1" to-layer="6328" to-port="0" /> <edge from-layer="6321" from-port="0" to-layer="6322" to-port="0" /> <edge from-layer="6322" from-port="1" to-layer="6325" to-port="0" /> <edge from-layer="6323" from-port="0" to-layer="6324" to-port="0" /> <edge from-layer="6324" from-port="1" to-layer="6325" to-port="1" /> <edge from-layer="6325" from-port="2" to-layer="6327" to-port="0" /> <edge from-layer="6326" from-port="0" to-layer="6327" to-port="1" /> <edge from-layer="6327" from-port="2" to-layer="6328" to-port="1" /> <edge from-layer="6328" from-port="2" to-layer="6330" to-port="0" /> <edge from-layer="6329" from-port="0" to-layer="6330" to-port="1" /> <edge from-layer="6330" from-port="2" to-layer="6533" to-port="0" /> <edge from-layer="6331" from-port="0" to-layer="6332" to-port="1" /> <edge from-layer="6332" from-port="2" to-layer="6334" to-port="0" /> <edge from-layer="6333" from-port="0" to-layer="6334" to-port="1" /> <edge from-layer="6334" from-port="2" to-layer="6336" to-port="0" /> <edge from-layer="6335" from-port="0" to-layer="6336" to-port="1" /> <edge from-layer="6336" from-port="2" to-layer="6344" to-port="0" /> <edge from-layer="6337" from-port="0" to-layer="6338" to-port="0" /> <edge from-layer="6338" from-port="1" to-layer="6341" to-port="0" /> <edge from-layer="6339" from-port="0" to-layer="6340" to-port="0" /> <edge from-layer="6340" from-port="1" to-layer="6341" to-port="1" /> <edge from-layer="6341" from-port="2" to-layer="6343" to-port="0" /> <edge from-layer="6342" from-port="0" to-layer="6343" to-port="1" /> <edge from-layer="6343" from-port="2" to-layer="6344" to-port="1" /> <edge from-layer="6344" from-port="2" to-layer="6346" to-port="0" /> <edge from-layer="6345" from-port="0" to-layer="6346" to-port="1" /> <edge from-layer="6346" from-port="2" to-layer="6348" to-port="0" /> <edge from-layer="6347" from-port="0" to-layer="6348" to-port="1" /> <edge from-layer="6348" from-port="2" to-layer="6364" to-port="0" /> <edge from-layer="6348" from-port="2" to-layer="6352" to-port="0" /> <edge from-layer="6348" from-port="2" to-layer="6470" to-port="0" /> <edge from-layer="6349" from-port="0" to-layer="6352" to-port="1" /> <edge from-layer="6350" from-port="0" to-layer="6352" to-port="2" /> <edge from-layer="6351" from-port="0" to-layer="6352" to-port="3" /> <edge from-layer="6352" from-port="4" to-layer="6354" to-port="0" /> <edge from-layer="6353" from-port="0" to-layer="6354" to-port="1" /> <edge from-layer="6354" from-port="2" to-layer="6358" to-port="0" /> <edge from-layer="6354" from-port="2" to-layer="6423" to-port="0" /> <edge from-layer="6355" from-port="0" to-layer="6358" to-port="1" /> <edge from-layer="6356" from-port="0" to-layer="6358" to-port="2" /> <edge from-layer="6357" from-port="0" to-layer="6358" to-port="3" /> <edge from-layer="6358" from-port="4" to-layer="6404" to-port="0" /> <edge from-layer="6358" from-port="4" to-layer="6409" to-port="0" /> <edge from-layer="6358" from-port="4" to-layer="6393" to-port="0" /> <edge from-layer="6358" from-port="4" to-layer="6390" to-port="0" /> <edge from-layer="6359" from-port="0" to-layer="6386" to-port="0" /> <edge from-layer="6360" from-port="0" to-layer="6386" to-port="1" /> <edge from-layer="6361" from-port="0" to-layer="6364" to-port="1" /> <edge from-layer="6362" from-port="0" to-layer="6364" to-port="2" /> <edge from-layer="6363" from-port="0" to-layer="6364" to-port="3" /> <edge from-layer="6364" from-port="4" to-layer="6366" to-port="0" /> <edge from-layer="6365" from-port="0" to-layer="6366" to-port="1" /> <edge from-layer="6366" from-port="2" to-layer="6428" to-port="0" /> <edge from-layer="6366" from-port="2" to-layer="6455" to-port="0" /> <edge from-layer="6366" from-port="2" to-layer="6367" to-port="0" /> <edge from-layer="6367" from-port="1" to-layer="6369" to-port="0" /> <edge from-layer="6368" from-port="0" to-layer="6369" to-port="2" /> <edge from-layer="6369" from-port="3" to-layer="6382" to-port="0" /> <edge from-layer="6370" from-port="0" to-layer="6375" to-port="0" /> <edge from-layer="6371" from-port="0" to-layer="6374" to-port="1" /> <edge from-layer="6372" from-port="0" to-layer="6374" to-port="2" /> <edge from-layer="6373" from-port="0" to-layer="6374" to-port="3" /> <edge from-layer="6374" from-port="4" to-layer="6375" to-port="1" /> <edge from-layer="6375" from-port="2" to-layer="6376" to-port="0" /> <edge from-layer="6376" from-port="1" to-layer="6378" to-port="0" /> <edge from-layer="6377" from-port="0" to-layer="6378" to-port="2" /> <edge from-layer="6378" from-port="3" to-layer="6379" to-port="0" /> <edge from-layer="6378" from-port="3" to-layer="6457" to-port="0" /> <edge from-layer="6379" from-port="1" to-layer="6381" to-port="0" /> <edge from-layer="6380" from-port="0" to-layer="6381" to-port="2" /> <edge from-layer="6381" from-port="3" to-layer="6382" to-port="1" /> <edge from-layer="6382" from-port="2" to-layer="6384" to-port="0" /> <edge from-layer="6383" from-port="0" to-layer="6384" to-port="1" /> <edge from-layer="6384" from-port="2" to-layer="6414" to-port="2" /> <edge from-layer="6384" from-port="2" to-layer="6386" to-port="2" /> <edge from-layer="6385" from-port="0" to-layer="6386" to-port="3" /> <edge from-layer="6386" from-port="4" to-layer="6388" to-port="0" /> <edge from-layer="6387" from-port="0" to-layer="6388" to-port="2" /> <edge from-layer="6388" from-port="3" to-layer="6389" to-port="0" /> <edge from-layer="6389" from-port="2" to-layer="6429" to-port="1" /> <edge from-layer="6389" from-port="2" to-layer="6390" to-port="1" /> <edge from-layer="6390" from-port="2" to-layer="6419" to-port="0" /> <edge from-layer="6391" from-port="0" to-layer="6401" to-port="0" /> <edge from-layer="6392" from-port="0" to-layer="6401" to-port="1" /> <edge from-layer="6393" from-port="1" to-layer="6395" to-port="0" /> <edge from-layer="6394" from-port="0" to-layer="6395" to-port="2" /> <edge from-layer="6395" from-port="3" to-layer="6396" to-port="0" /> <edge from-layer="6396" from-port="2" to-layer="6397" to-port="0" /> <edge from-layer="6397" from-port="1" to-layer="6399" to-port="0" /> <edge from-layer="6398" from-port="0" to-layer="6399" to-port="1" /> <edge from-layer="6399" from-port="2" to-layer="6401" to-port="2" /> <edge from-layer="6400" from-port="0" to-layer="6401" to-port="3" /> <edge from-layer="6401" from-port="4" to-layer="6404" to-port="1" /> <edge from-layer="6401" from-port="4" to-layer="6409" to-port="2" /> <edge from-layer="6402" from-port="0" to-layer="6404" to-port="2" /> <edge from-layer="6403" from-port="0" to-layer="6404" to-port="3" /> <edge from-layer="6404" from-port="4" to-layer="6406" to-port="0" /> <edge from-layer="6405" from-port="0" to-layer="6406" to-port="1" /> <edge from-layer="6406" from-port="2" to-layer="6410" to-port="0" /> <edge from-layer="6407" from-port="0" to-layer="6409" to-port="1" /> <edge from-layer="6408" from-port="0" to-layer="6409" to-port="3" /> <edge from-layer="6409" from-port="4" to-layer="6410" to-port="1" /> <edge from-layer="6410" from-port="2" to-layer="6418" to-port="0" /> <edge from-layer="6411" from-port="0" to-layer="6414" to-port="0" /> <edge from-layer="6412" from-port="0" to-layer="6414" to-port="1" /> <edge from-layer="6413" from-port="0" to-layer="6414" to-port="3" /> <edge from-layer="6414" from-port="4" to-layer="6416" to-port="0" /> <edge from-layer="6415" from-port="0" to-layer="6416" to-port="2" /> <edge from-layer="6416" from-port="3" to-layer="6417" to-port="0" /> <edge from-layer="6417" from-port="2" to-layer="6418" to-port="1" /> <edge from-layer="6417" from-port="2" to-layer="6450" to-port="1" /> <edge from-layer="6418" from-port="2" to-layer="6419" to-port="1" /> <edge from-layer="6419" from-port="2" to-layer="6424" to-port="0" /> <edge from-layer="6420" from-port="0" to-layer="6423" to-port="1" /> <edge from-layer="6421" from-port="0" to-layer="6423" to-port="2" /> <edge from-layer="6422" from-port="0" to-layer="6423" to-port="3" /> <edge from-layer="6423" from-port="4" to-layer="6424" to-port="1" /> <edge from-layer="6424" from-port="2" to-layer="6480" to-port="0" /> <edge from-layer="6424" from-port="2" to-layer="6518" to-port="0" /> <edge from-layer="6425" from-port="0" to-layer="6428" to-port="1" /> <edge from-layer="6426" from-port="0" to-layer="6428" to-port="2" /> <edge from-layer="6427" from-port="0" to-layer="6428" to-port="3" /> <edge from-layer="6428" from-port="4" to-layer="6443" to-port="0" /> <edge from-layer="6428" from-port="4" to-layer="6429" to-port="0" /> <edge from-layer="6428" from-port="4" to-layer="6432" to-port="0" /> <edge from-layer="6428" from-port="4" to-layer="6448" to-port="0" /> <edge from-layer="6429" from-port="2" to-layer="6451" to-port="0" /> <edge from-layer="6430" from-port="0" to-layer="6440" to-port="0" /> <edge from-layer="6431" from-port="0" to-layer="6440" to-port="1" /> <edge from-layer="6432" from-port="1" to-layer="6434" to-port="0" /> <edge from-layer="6433" from-port="0" to-layer="6434" to-port="2" /> <edge from-layer="6434" from-port="3" to-layer="6435" to-port="0" /> <edge from-layer="6435" from-port="2" to-layer="6436" to-port="0" /> <edge from-layer="6436" from-port="1" to-layer="6438" to-port="0" /> <edge from-layer="6437" from-port="0" to-layer="6438" to-port="1" /> <edge from-layer="6438" from-port="2" to-layer="6440" to-port="2" /> <edge from-layer="6439" from-port="0" to-layer="6440" to-port="3" /> <edge from-layer="6440" from-port="4" to-layer="6448" to-port="2" /> <edge from-layer="6440" from-port="4" to-layer="6443" to-port="1" /> <edge from-layer="6441" from-port="0" to-layer="6443" to-port="2" /> <edge from-layer="6442" from-port="0" to-layer="6443" to-port="3" /> <edge from-layer="6443" from-port="4" to-layer="6445" to-port="0" /> <edge from-layer="6444" from-port="0" to-layer="6445" to-port="1" /> <edge from-layer="6445" from-port="2" to-layer="6449" to-port="0" /> <edge from-layer="6446" from-port="0" to-layer="6448" to-port="1" /> <edge from-layer="6447" from-port="0" to-layer="6448" to-port="3" /> <edge from-layer="6448" from-port="4" to-layer="6449" to-port="1" /> <edge from-layer="6449" from-port="2" to-layer="6450" to-port="0" /> <edge from-layer="6450" from-port="2" to-layer="6451" to-port="1" /> <edge from-layer="6451" from-port="2" to-layer="6456" to-port="0" /> <edge from-layer="6452" from-port="0" to-layer="6455" to-port="1" /> <edge from-layer="6453" from-port="0" to-layer="6455" to-port="2" /> <edge from-layer="6454" from-port="0" to-layer="6455" to-port="3" /> <edge from-layer="6455" from-port="4" to-layer="6456" to-port="1" /> <edge from-layer="6456" from-port="2" to-layer="6457" to-port="1" /> <edge from-layer="6457" from-port="2" to-layer="10133" to-port="0" /> <edge from-layer="6457" from-port="2" to-layer="6518" to-port="1" /> <edge from-layer="6457" from-port="2" to-layer="6477" to-port="0" /> <edge from-layer="6458" from-port="0" to-layer="6463" to-port="0" /> <edge from-layer="6459" from-port="0" to-layer="6462" to-port="1" /> <edge from-layer="6460" from-port="0" to-layer="6462" to-port="2" /> <edge from-layer="6461" from-port="0" to-layer="6462" to-port="3" /> <edge from-layer="6462" from-port="4" to-layer="6463" to-port="1" /> <edge from-layer="6463" from-port="2" to-layer="6464" to-port="0" /> <edge from-layer="6464" from-port="1" to-layer="6466" to-port="0" /> <edge from-layer="6465" from-port="0" to-layer="6466" to-port="2" /> <edge from-layer="6466" from-port="3" to-layer="6473" to-port="0" /> <edge from-layer="6467" from-port="0" to-layer="6470" to-port="1" /> <edge from-layer="6468" from-port="0" to-layer="6470" to-port="2" /> <edge from-layer="6469" from-port="0" to-layer="6470" to-port="3" /> <edge from-layer="6470" from-port="4" to-layer="6472" to-port="0" /> <edge from-layer="6471" from-port="0" to-layer="6472" to-port="1" /> <edge from-layer="6472" from-port="2" to-layer="6473" to-port="1" /> <edge from-layer="6473" from-port="2" to-layer="10132" to-port="0" /> <edge from-layer="6473" from-port="2" to-layer="6518" to-port="2" /> <edge from-layer="6474" from-port="0" to-layer="6492" to-port="0" /> <edge from-layer="6475" from-port="0" to-layer="6487" to-port="0" /> <edge from-layer="6476" from-port="0" to-layer="6490" to-port="1" /> <edge from-layer="6476" from-port="0" to-layer="6487" to-port="1" /> <edge from-layer="6477" from-port="1" to-layer="6479" to-port="0" /> <edge from-layer="6478" from-port="0" to-layer="6479" to-port="2" /> <edge from-layer="6479" from-port="3" to-layer="6489" to-port="0" /> <edge from-layer="6479" from-port="3" to-layer="6483" to-port="0" /> <edge from-layer="6480" from-port="1" to-layer="6482" to-port="0" /> <edge from-layer="6480" from-port="1" to-layer="6507" to-port="0" /> <edge from-layer="6481" from-port="0" to-layer="6482" to-port="2" /> <edge from-layer="6482" from-port="3" to-layer="6483" to-port="1" /> <edge from-layer="6483" from-port="2" to-layer="6485" to-port="0" /> <edge from-layer="6484" from-port="0" to-layer="6489" to-port="1" /> <edge from-layer="6484" from-port="0" to-layer="6485" to-port="1" /> <edge from-layer="6485" from-port="2" to-layer="6487" to-port="2" /> <edge from-layer="6486" from-port="0" to-layer="6487" to-port="3" /> <edge from-layer="6486" from-port="0" to-layer="6490" to-port="3" /> <edge from-layer="6487" from-port="4" to-layer="6492" to-port="1" /> <edge from-layer="6488" from-port="0" to-layer="6490" to-port="0" /> <edge from-layer="6489" from-port="2" to-layer="6490" to-port="2" /> <edge from-layer="6489" from-port="2" to-layer="6497" to-port="2" /> <edge from-layer="6490" from-port="4" to-layer="6492" to-port="2" /> <edge from-layer="6491" from-port="0" to-layer="6492" to-port="3" /> <edge from-layer="6492" from-port="4" to-layer="6499" to-port="0" /> <edge from-layer="6493" from-port="0" to-layer="6499" to-port="1" /> <edge from-layer="6494" from-port="0" to-layer="6497" to-port="0" /> <edge from-layer="6495" from-port="0" to-layer="6497" to-port="1" /> <edge from-layer="6496" from-port="0" to-layer="6497" to-port="3" /> <edge from-layer="6497" from-port="4" to-layer="6499" to-port="2" /> <edge from-layer="6498" from-port="0" to-layer="6499" to-port="3" /> <edge from-layer="6499" from-port="4" to-layer="6502" to-port="0" /> <edge from-layer="6500" from-port="0" to-layer="6502" to-port="1" /> <edge from-layer="6501" from-port="0" to-layer="6502" to-port="2" /> <edge from-layer="6502" from-port="3" to-layer="6503" to-port="0" /> <edge from-layer="6503" from-port="2" to-layer="6517" to-port="0" /> <edge from-layer="6503" from-port="2" to-layer="6504" to-port="0" /> <edge from-layer="6504" from-port="1" to-layer="6516" to-port="0" /> <edge from-layer="6505" from-port="0" to-layer="6507" to-port="1" /> <edge from-layer="6506" from-port="0" to-layer="6507" to-port="2" /> <edge from-layer="6507" from-port="3" to-layer="6511" to-port="0" /> <edge from-layer="6508" from-port="0" to-layer="6511" to-port="1" /> <edge from-layer="6509" from-port="0" to-layer="6511" to-port="2" /> <edge from-layer="6510" from-port="0" to-layer="6511" to-port="3" /> <edge from-layer="6511" from-port="4" to-layer="6515" to-port="2" /> <edge from-layer="6511" from-port="4" to-layer="6513" to-port="0" /> <edge from-layer="6512" from-port="0" to-layer="6513" to-port="1" /> <edge from-layer="6513" from-port="2" to-layer="6515" to-port="0" /> <edge from-layer="6514" from-port="0" to-layer="6515" to-port="1" /> <edge from-layer="6515" from-port="3" to-layer="6516" to-port="1" /> <edge from-layer="6516" from-port="2" to-layer="6517" to-port="1" /> <edge from-layer="6517" from-port="2" to-layer="6518" to-port="3" /> <edge from-layer="6518" from-port="4" to-layer="6520" to-port="0" /> <edge from-layer="6519" from-port="0" to-layer="6520" to-port="1" /> <edge from-layer="6520" from-port="2" to-layer="6522" to-port="0" /> <edge from-layer="6521" from-port="0" to-layer="6522" to-port="1" /> <edge from-layer="6522" from-port="2" to-layer="6530" to-port="0" /> <edge from-layer="6523" from-port="0" to-layer="6524" to-port="0" /> <edge from-layer="6524" from-port="1" to-layer="6527" to-port="0" /> <edge from-layer="6525" from-port="0" to-layer="6526" to-port="0" /> <edge from-layer="6526" from-port="1" to-layer="6527" to-port="1" /> <edge from-layer="6527" from-port="2" to-layer="6529" to-port="0" /> <edge from-layer="6528" from-port="0" to-layer="6529" to-port="1" /> <edge from-layer="6529" from-port="2" to-layer="6530" to-port="1" /> <edge from-layer="6530" from-port="2" to-layer="6532" to-port="0" /> <edge from-layer="6531" from-port="0" to-layer="6532" to-port="1" /> <edge from-layer="6532" from-port="2" to-layer="6533" to-port="1" /> <edge from-layer="6533" from-port="2" to-layer="6534" to-port="0" /> <edge from-layer="6534" from-port="2" to-layer="6765" to-port="1" /> <edge from-layer="6534" from-port="2" to-layer="6536" to-port="0" /> <edge from-layer="6534" from-port="2" to-layer="6563" to-port="0" /> <edge from-layer="6535" from-port="0" to-layer="6536" to-port="1" /> <edge from-layer="6536" from-port="2" to-layer="6538" to-port="0" /> <edge from-layer="6537" from-port="0" to-layer="6538" to-port="1" /> <edge from-layer="6538" from-port="2" to-layer="6540" to-port="0" /> <edge from-layer="6539" from-port="0" to-layer="6540" to-port="1" /> <edge from-layer="6540" from-port="2" to-layer="6548" to-port="0" /> <edge from-layer="6541" from-port="0" to-layer="6542" to-port="0" /> <edge from-layer="6542" from-port="1" to-layer="6545" to-port="0" /> <edge from-layer="6543" from-port="0" to-layer="6544" to-port="0" /> <edge from-layer="6544" from-port="1" to-layer="6545" to-port="1" /> <edge from-layer="6545" from-port="2" to-layer="6547" to-port="0" /> <edge from-layer="6546" from-port="0" to-layer="6547" to-port="1" /> <edge from-layer="6547" from-port="2" to-layer="6548" to-port="1" /> <edge from-layer="6548" from-port="2" to-layer="6550" to-port="0" /> <edge from-layer="6549" from-port="0" to-layer="6550" to-port="1" /> <edge from-layer="6550" from-port="2" to-layer="6551" to-port="0" /> <edge from-layer="6551" from-port="1" to-layer="6559" to-port="0" /> <edge from-layer="6552" from-port="0" to-layer="6553" to-port="0" /> <edge from-layer="6553" from-port="1" to-layer="6556" to-port="0" /> <edge from-layer="6554" from-port="0" to-layer="6555" to-port="0" /> <edge from-layer="6555" from-port="1" to-layer="6556" to-port="1" /> <edge from-layer="6556" from-port="2" to-layer="6558" to-port="0" /> <edge from-layer="6557" from-port="0" to-layer="6558" to-port="1" /> <edge from-layer="6558" from-port="2" to-layer="6559" to-port="1" /> <edge from-layer="6559" from-port="2" to-layer="6561" to-port="0" /> <edge from-layer="6560" from-port="0" to-layer="6561" to-port="1" /> <edge from-layer="6561" from-port="2" to-layer="6764" to-port="0" /> <edge from-layer="6562" from-port="0" to-layer="6563" to-port="1" /> <edge from-layer="6563" from-port="2" to-layer="6565" to-port="0" /> <edge from-layer="6564" from-port="0" to-layer="6565" to-port="1" /> <edge from-layer="6565" from-port="2" to-layer="6567" to-port="0" /> <edge from-layer="6566" from-port="0" to-layer="6567" to-port="1" /> <edge from-layer="6567" from-port="2" to-layer="6575" to-port="0" /> <edge from-layer="6568" from-port="0" to-layer="6569" to-port="0" /> <edge from-layer="6569" from-port="1" to-layer="6572" to-port="0" /> <edge from-layer="6570" from-port="0" to-layer="6571" to-port="0" /> <edge from-layer="6571" from-port="1" to-layer="6572" to-port="1" /> <edge from-layer="6572" from-port="2" to-layer="6574" to-port="0" /> <edge from-layer="6573" from-port="0" to-layer="6574" to-port="1" /> <edge from-layer="6574" from-port="2" to-layer="6575" to-port="1" /> <edge from-layer="6575" from-port="2" to-layer="6577" to-port="0" /> <edge from-layer="6576" from-port="0" to-layer="6577" to-port="1" /> <edge from-layer="6577" from-port="2" to-layer="6579" to-port="0" /> <edge from-layer="6578" from-port="0" to-layer="6579" to-port="1" /> <edge from-layer="6579" from-port="2" to-layer="6701" to-port="0" /> <edge from-layer="6579" from-port="2" to-layer="6595" to-port="0" /> <edge from-layer="6579" from-port="2" to-layer="6583" to-port="0" /> <edge from-layer="6580" from-port="0" to-layer="6583" to-port="1" /> <edge from-layer="6581" from-port="0" to-layer="6583" to-port="2" /> <edge from-layer="6582" from-port="0" to-layer="6583" to-port="3" /> <edge from-layer="6583" from-port="4" to-layer="6585" to-port="0" /> <edge from-layer="6584" from-port="0" to-layer="6585" to-port="1" /> <edge from-layer="6585" from-port="2" to-layer="6589" to-port="0" /> <edge from-layer="6585" from-port="2" to-layer="6654" to-port="0" /> <edge from-layer="6586" from-port="0" to-layer="6589" to-port="1" /> <edge from-layer="6587" from-port="0" to-layer="6589" to-port="2" /> <edge from-layer="6588" from-port="0" to-layer="6589" to-port="3" /> <edge from-layer="6589" from-port="4" to-layer="6621" to-port="0" /> <edge from-layer="6589" from-port="4" to-layer="6624" to-port="0" /> <edge from-layer="6589" from-port="4" to-layer="6635" to-port="0" /> <edge from-layer="6589" from-port="4" to-layer="6640" to-port="0" /> <edge from-layer="6590" from-port="0" to-layer="6617" to-port="0" /> <edge from-layer="6591" from-port="0" to-layer="6617" to-port="1" /> <edge from-layer="6592" from-port="0" to-layer="6595" to-port="1" /> <edge from-layer="6593" from-port="0" to-layer="6595" to-port="2" /> <edge from-layer="6594" from-port="0" to-layer="6595" to-port="3" /> <edge from-layer="6595" from-port="4" to-layer="6597" to-port="0" /> <edge from-layer="6596" from-port="0" to-layer="6597" to-port="1" /> <edge from-layer="6597" from-port="2" to-layer="6598" to-port="0" /> <edge from-layer="6597" from-port="2" to-layer="6686" to-port="0" /> <edge from-layer="6597" from-port="2" to-layer="6659" to-port="0" /> <edge from-layer="6598" from-port="1" to-layer="6600" to-port="0" /> <edge from-layer="6599" from-port="0" to-layer="6600" to-port="2" /> <edge from-layer="6600" from-port="3" to-layer="6613" to-port="0" /> <edge from-layer="6601" from-port="0" to-layer="6606" to-port="0" /> <edge from-layer="6602" from-port="0" to-layer="6605" to-port="1" /> <edge from-layer="6603" from-port="0" to-layer="6605" to-port="2" /> <edge from-layer="6604" from-port="0" to-layer="6605" to-port="3" /> <edge from-layer="6605" from-port="4" to-layer="6606" to-port="1" /> <edge from-layer="6606" from-port="2" to-layer="6607" to-port="0" /> <edge from-layer="6607" from-port="1" to-layer="6609" to-port="0" /> <edge from-layer="6608" from-port="0" to-layer="6609" to-port="2" /> <edge from-layer="6609" from-port="3" to-layer="6688" to-port="0" /> <edge from-layer="6609" from-port="3" to-layer="6610" to-port="0" /> <edge from-layer="6610" from-port="1" to-layer="6612" to-port="0" /> <edge from-layer="6611" from-port="0" to-layer="6612" to-port="2" /> <edge from-layer="6612" from-port="3" to-layer="6613" to-port="1" /> <edge from-layer="6613" from-port="2" to-layer="6615" to-port="0" /> <edge from-layer="6614" from-port="0" to-layer="6615" to-port="1" /> <edge from-layer="6615" from-port="2" to-layer="6645" to-port="2" /> <edge from-layer="6615" from-port="2" to-layer="6617" to-port="2" /> <edge from-layer="6616" from-port="0" to-layer="6617" to-port="3" /> <edge from-layer="6617" from-port="4" to-layer="6619" to-port="0" /> <edge from-layer="6618" from-port="0" to-layer="6619" to-port="2" /> <edge from-layer="6619" from-port="3" to-layer="6620" to-port="0" /> <edge from-layer="6620" from-port="2" to-layer="6621" to-port="1" /> <edge from-layer="6620" from-port="2" to-layer="6660" to-port="1" /> <edge from-layer="6621" from-port="2" to-layer="6650" to-port="0" /> <edge from-layer="6622" from-port="0" to-layer="6632" to-port="0" /> <edge from-layer="6623" from-port="0" to-layer="6632" to-port="1" /> <edge from-layer="6624" from-port="1" to-layer="6626" to-port="0" /> <edge from-layer="6625" from-port="0" to-layer="6626" to-port="2" /> <edge from-layer="6626" from-port="3" to-layer="6627" to-port="0" /> <edge from-layer="6627" from-port="2" to-layer="6628" to-port="0" /> <edge from-layer="6628" from-port="1" to-layer="6630" to-port="0" /> <edge from-layer="6629" from-port="0" to-layer="6630" to-port="1" /> <edge from-layer="6630" from-port="2" to-layer="6632" to-port="2" /> <edge from-layer="6631" from-port="0" to-layer="6632" to-port="3" /> <edge from-layer="6632" from-port="4" to-layer="6635" to-port="1" /> <edge from-layer="6632" from-port="4" to-layer="6640" to-port="2" /> <edge from-layer="6633" from-port="0" to-layer="6635" to-port="2" /> <edge from-layer="6634" from-port="0" to-layer="6635" to-port="3" /> <edge from-layer="6635" from-port="4" to-layer="6637" to-port="0" /> <edge from-layer="6636" from-port="0" to-layer="6637" to-port="1" /> <edge from-layer="6637" from-port="2" to-layer="6641" to-port="0" /> <edge from-layer="6638" from-port="0" to-layer="6640" to-port="1" /> <edge from-layer="6639" from-port="0" to-layer="6640" to-port="3" /> <edge from-layer="6640" from-port="4" to-layer="6641" to-port="1" /> <edge from-layer="6641" from-port="2" to-layer="6649" to-port="0" /> <edge from-layer="6642" from-port="0" to-layer="6645" to-port="0" /> <edge from-layer="6643" from-port="0" to-layer="6645" to-port="1" /> <edge from-layer="6644" from-port="0" to-layer="6645" to-port="3" /> <edge from-layer="6645" from-port="4" to-layer="6647" to-port="0" /> <edge from-layer="6646" from-port="0" to-layer="6647" to-port="2" /> <edge from-layer="6647" from-port="3" to-layer="6648" to-port="0" /> <edge from-layer="6648" from-port="2" to-layer="6681" to-port="1" /> <edge from-layer="6648" from-port="2" to-layer="6649" to-port="1" /> <edge from-layer="6649" from-port="2" to-layer="6650" to-port="1" /> <edge from-layer="6650" from-port="2" to-layer="6655" to-port="0" /> <edge from-layer="6651" from-port="0" to-layer="6654" to-port="1" /> <edge from-layer="6652" from-port="0" to-layer="6654" to-port="2" /> <edge from-layer="6653" from-port="0" to-layer="6654" to-port="3" /> <edge from-layer="6654" from-port="4" to-layer="6655" to-port="1" /> <edge from-layer="6655" from-port="2" to-layer="6749" to-port="0" /> <edge from-layer="6655" from-port="2" to-layer="6711" to-port="0" /> <edge from-layer="6656" from-port="0" to-layer="6659" to-port="1" /> <edge from-layer="6657" from-port="0" to-layer="6659" to-port="2" /> <edge from-layer="6658" from-port="0" to-layer="6659" to-port="3" /> <edge from-layer="6659" from-port="4" to-layer="6660" to-port="0" /> <edge from-layer="6659" from-port="4" to-layer="6663" to-port="0" /> <edge from-layer="6659" from-port="4" to-layer="6679" to-port="0" /> <edge from-layer="6659" from-port="4" to-layer="6674" to-port="0" /> <edge from-layer="6660" from-port="2" to-layer="6682" to-port="0" /> <edge from-layer="6661" from-port="0" to-layer="6671" to-port="0" /> <edge from-layer="6662" from-port="0" to-layer="6671" to-port="1" /> <edge from-layer="6663" from-port="1" to-layer="6665" to-port="0" /> <edge from-layer="6664" from-port="0" to-layer="6665" to-port="2" /> <edge from-layer="6665" from-port="3" to-layer="6666" to-port="0" /> <edge from-layer="6666" from-port="2" to-layer="6667" to-port="0" /> <edge from-layer="6667" from-port="1" to-layer="6669" to-port="0" /> <edge from-layer="6668" from-port="0" to-layer="6669" to-port="1" /> <edge from-layer="6669" from-port="2" to-layer="6671" to-port="2" /> <edge from-layer="6670" from-port="0" to-layer="6671" to-port="3" /> <edge from-layer="6671" from-port="4" to-layer="6679" to-port="2" /> <edge from-layer="6671" from-port="4" to-layer="6674" to-port="1" /> <edge from-layer="6672" from-port="0" to-layer="6674" to-port="2" /> <edge from-layer="6673" from-port="0" to-layer="6674" to-port="3" /> <edge from-layer="6674" from-port="4" to-layer="6676" to-port="0" /> <edge from-layer="6675" from-port="0" to-layer="6676" to-port="1" /> <edge from-layer="6676" from-port="2" to-layer="6680" to-port="0" /> <edge from-layer="6677" from-port="0" to-layer="6679" to-port="1" /> <edge from-layer="6678" from-port="0" to-layer="6679" to-port="3" /> <edge from-layer="6679" from-port="4" to-layer="6680" to-port="1" /> <edge from-layer="6680" from-port="2" to-layer="6681" to-port="0" /> <edge from-layer="6681" from-port="2" to-layer="6682" to-port="1" /> <edge from-layer="6682" from-port="2" to-layer="6687" to-port="0" /> <edge from-layer="6683" from-port="0" to-layer="6686" to-port="1" /> <edge from-layer="6684" from-port="0" to-layer="6686" to-port="2" /> <edge from-layer="6685" from-port="0" to-layer="6686" to-port="3" /> <edge from-layer="6686" from-port="4" to-layer="6687" to-port="1" /> <edge from-layer="6687" from-port="2" to-layer="6688" to-port="1" /> <edge from-layer="6688" from-port="2" to-layer="10131" to-port="0" /> <edge from-layer="6688" from-port="2" to-layer="6749" to-port="1" /> <edge from-layer="6688" from-port="2" to-layer="6708" to-port="0" /> <edge from-layer="6689" from-port="0" to-layer="6694" to-port="0" /> <edge from-layer="6690" from-port="0" to-layer="6693" to-port="1" /> <edge from-layer="6691" from-port="0" to-layer="6693" to-port="2" /> <edge from-layer="6692" from-port="0" to-layer="6693" to-port="3" /> <edge from-layer="6693" from-port="4" to-layer="6694" to-port="1" /> <edge from-layer="6694" from-port="2" to-layer="6695" to-port="0" /> <edge from-layer="6695" from-port="1" to-layer="6697" to-port="0" /> <edge from-layer="6696" from-port="0" to-layer="6697" to-port="2" /> <edge from-layer="6697" from-port="3" to-layer="6704" to-port="0" /> <edge from-layer="6698" from-port="0" to-layer="6701" to-port="1" /> <edge from-layer="6699" from-port="0" to-layer="6701" to-port="2" /> <edge from-layer="6700" from-port="0" to-layer="6701" to-port="3" /> <edge from-layer="6701" from-port="4" to-layer="6703" to-port="0" /> <edge from-layer="6702" from-port="0" to-layer="6703" to-port="1" /> <edge from-layer="6703" from-port="2" to-layer="6704" to-port="1" /> <edge from-layer="6704" from-port="2" to-layer="10130" to-port="0" /> <edge from-layer="6704" from-port="2" to-layer="6749" to-port="2" /> <edge from-layer="6705" from-port="0" to-layer="6723" to-port="0" /> <edge from-layer="6706" from-port="0" to-layer="6718" to-port="0" /> <edge from-layer="6707" from-port="0" to-layer="6718" to-port="1" /> <edge from-layer="6707" from-port="0" to-layer="6721" to-port="1" /> <edge from-layer="6708" from-port="1" to-layer="6710" to-port="0" /> <edge from-layer="6709" from-port="0" to-layer="6710" to-port="2" /> <edge from-layer="6710" from-port="3" to-layer="6714" to-port="0" /> <edge from-layer="6710" from-port="3" to-layer="6720" to-port="0" /> <edge from-layer="6711" from-port="1" to-layer="6738" to-port="0" /> <edge from-layer="6711" from-port="1" to-layer="6713" to-port="0" /> <edge from-layer="6712" from-port="0" to-layer="6713" to-port="2" /> <edge from-layer="6713" from-port="3" to-layer="6714" to-port="1" /> <edge from-layer="6714" from-port="2" to-layer="6716" to-port="0" /> <edge from-layer="6715" from-port="0" to-layer="6720" to-port="1" /> <edge from-layer="6715" from-port="0" to-layer="6716" to-port="1" /> <edge from-layer="6716" from-port="2" to-layer="6718" to-port="2" /> <edge from-layer="6717" from-port="0" to-layer="6721" to-port="3" /> <edge from-layer="6717" from-port="0" to-layer="6718" to-port="3" /> <edge from-layer="6718" from-port="4" to-layer="6723" to-port="1" /> <edge from-layer="6719" from-port="0" to-layer="6721" to-port="0" /> <edge from-layer="6720" from-port="2" to-layer="6721" to-port="2" /> <edge from-layer="6720" from-port="2" to-layer="6728" to-port="2" /> <edge from-layer="6721" from-port="4" to-layer="6723" to-port="2" /> <edge from-layer="6722" from-port="0" to-layer="6723" to-port="3" /> <edge from-layer="6723" from-port="4" to-layer="6730" to-port="0" /> <edge from-layer="6724" from-port="0" to-layer="6730" to-port="1" /> <edge from-layer="6725" from-port="0" to-layer="6728" to-port="0" /> <edge from-layer="6726" from-port="0" to-layer="6728" to-port="1" /> <edge from-layer="6727" from-port="0" to-layer="6728" to-port="3" /> <edge from-layer="6728" from-port="4" to-layer="6730" to-port="2" /> <edge from-layer="6729" from-port="0" to-layer="6730" to-port="3" /> <edge from-layer="6730" from-port="4" to-layer="6733" to-port="0" /> <edge from-layer="6731" from-port="0" to-layer="6733" to-port="1" /> <edge from-layer="6732" from-port="0" to-layer="6733" to-port="2" /> <edge from-layer="6733" from-port="3" to-layer="6734" to-port="0" /> <edge from-layer="6734" from-port="2" to-layer="6735" to-port="0" /> <edge from-layer="6734" from-port="2" to-layer="6748" to-port="0" /> <edge from-layer="6735" from-port="1" to-layer="6747" to-port="0" /> <edge from-layer="6736" from-port="0" to-layer="6738" to-port="1" /> <edge from-layer="6737" from-port="0" to-layer="6738" to-port="2" /> <edge from-layer="6738" from-port="3" to-layer="6742" to-port="0" /> <edge from-layer="6739" from-port="0" to-layer="6742" to-port="1" /> <edge from-layer="6740" from-port="0" to-layer="6742" to-port="2" /> <edge from-layer="6741" from-port="0" to-layer="6742" to-port="3" /> <edge from-layer="6742" from-port="4" to-layer="6744" to-port="0" /> <edge from-layer="6742" from-port="4" to-layer="6746" to-port="2" /> <edge from-layer="6743" from-port="0" to-layer="6744" to-port="1" /> <edge from-layer="6744" from-port="2" to-layer="6746" to-port="0" /> <edge from-layer="6745" from-port="0" to-layer="6746" to-port="1" /> <edge from-layer="6746" from-port="3" to-layer="6747" to-port="1" /> <edge from-layer="6747" from-port="2" to-layer="6748" to-port="1" /> <edge from-layer="6748" from-port="2" to-layer="6749" to-port="3" /> <edge from-layer="6749" from-port="4" to-layer="6751" to-port="0" /> <edge from-layer="6750" from-port="0" to-layer="6751" to-port="1" /> <edge from-layer="6751" from-port="2" to-layer="6753" to-port="0" /> <edge from-layer="6752" from-port="0" to-layer="6753" to-port="1" /> <edge from-layer="6753" from-port="2" to-layer="6761" to-port="0" /> <edge from-layer="6754" from-port="0" to-layer="6755" to-port="0" /> <edge from-layer="6755" from-port="1" to-layer="6758" to-port="0" /> <edge from-layer="6756" from-port="0" to-layer="6757" to-port="0" /> <edge from-layer="6757" from-port="1" to-layer="6758" to-port="1" /> <edge from-layer="6758" from-port="2" to-layer="6760" to-port="0" /> <edge from-layer="6759" from-port="0" to-layer="6760" to-port="1" /> <edge from-layer="6760" from-port="2" to-layer="6761" to-port="1" /> <edge from-layer="6761" from-port="2" to-layer="6763" to-port="0" /> <edge from-layer="6762" from-port="0" to-layer="6763" to-port="1" /> <edge from-layer="6763" from-port="2" to-layer="6764" to-port="1" /> <edge from-layer="6764" from-port="2" to-layer="6765" to-port="0" /> <edge from-layer="6765" from-port="2" to-layer="6996" to-port="1" /> <edge from-layer="6765" from-port="2" to-layer="6794" to-port="0" /> <edge from-layer="6765" from-port="2" to-layer="6767" to-port="0" /> <edge from-layer="6766" from-port="0" to-layer="6767" to-port="1" /> <edge from-layer="6767" from-port="2" to-layer="6769" to-port="0" /> <edge from-layer="6768" from-port="0" to-layer="6769" to-port="1" /> <edge from-layer="6769" from-port="2" to-layer="6771" to-port="0" /> <edge from-layer="6770" from-port="0" to-layer="6771" to-port="1" /> <edge from-layer="6771" from-port="2" to-layer="6779" to-port="0" /> <edge from-layer="6772" from-port="0" to-layer="6773" to-port="0" /> <edge from-layer="6773" from-port="1" to-layer="6776" to-port="0" /> <edge from-layer="6774" from-port="0" to-layer="6775" to-port="0" /> <edge from-layer="6775" from-port="1" to-layer="6776" to-port="1" /> <edge from-layer="6776" from-port="2" to-layer="6778" to-port="0" /> <edge from-layer="6777" from-port="0" to-layer="6778" to-port="1" /> <edge from-layer="6778" from-port="2" to-layer="6779" to-port="1" /> <edge from-layer="6779" from-port="2" to-layer="6781" to-port="0" /> <edge from-layer="6780" from-port="0" to-layer="6781" to-port="1" /> <edge from-layer="6781" from-port="2" to-layer="6782" to-port="0" /> <edge from-layer="6782" from-port="1" to-layer="6790" to-port="0" /> <edge from-layer="6783" from-port="0" to-layer="6784" to-port="0" /> <edge from-layer="6784" from-port="1" to-layer="6787" to-port="0" /> <edge from-layer="6785" from-port="0" to-layer="6786" to-port="0" /> <edge from-layer="6786" from-port="1" to-layer="6787" to-port="1" /> <edge from-layer="6787" from-port="2" to-layer="6789" to-port="0" /> <edge from-layer="6788" from-port="0" to-layer="6789" to-port="1" /> <edge from-layer="6789" from-port="2" to-layer="6790" to-port="1" /> <edge from-layer="6790" from-port="2" to-layer="6792" to-port="0" /> <edge from-layer="6791" from-port="0" to-layer="6792" to-port="1" /> <edge from-layer="6792" from-port="2" to-layer="6995" to-port="0" /> <edge from-layer="6793" from-port="0" to-layer="6794" to-port="1" /> <edge from-layer="6794" from-port="2" to-layer="6796" to-port="0" /> <edge from-layer="6795" from-port="0" to-layer="6796" to-port="1" /> <edge from-layer="6796" from-port="2" to-layer="6798" to-port="0" /> <edge from-layer="6797" from-port="0" to-layer="6798" to-port="1" /> <edge from-layer="6798" from-port="2" to-layer="6806" to-port="0" /> <edge from-layer="6799" from-port="0" to-layer="6800" to-port="0" /> <edge from-layer="6800" from-port="1" to-layer="6803" to-port="0" /> <edge from-layer="6801" from-port="0" to-layer="6802" to-port="0" /> <edge from-layer="6802" from-port="1" to-layer="6803" to-port="1" /> <edge from-layer="6803" from-port="2" to-layer="6805" to-port="0" /> <edge from-layer="6804" from-port="0" to-layer="6805" to-port="1" /> <edge from-layer="6805" from-port="2" to-layer="6806" to-port="1" /> <edge from-layer="6806" from-port="2" to-layer="6808" to-port="0" /> <edge from-layer="6807" from-port="0" to-layer="6808" to-port="1" /> <edge from-layer="6808" from-port="2" to-layer="6810" to-port="0" /> <edge from-layer="6809" from-port="0" to-layer="6810" to-port="1" /> <edge from-layer="6810" from-port="2" to-layer="6826" to-port="0" /> <edge from-layer="6810" from-port="2" to-layer="6814" to-port="0" /> <edge from-layer="6810" from-port="2" to-layer="6932" to-port="0" /> <edge from-layer="6811" from-port="0" to-layer="6814" to-port="1" /> <edge from-layer="6812" from-port="0" to-layer="6814" to-port="2" /> <edge from-layer="6813" from-port="0" to-layer="6814" to-port="3" /> <edge from-layer="6814" from-port="4" to-layer="6816" to-port="0" /> <edge from-layer="6815" from-port="0" to-layer="6816" to-port="1" /> <edge from-layer="6816" from-port="2" to-layer="6885" to-port="0" /> <edge from-layer="6816" from-port="2" to-layer="6820" to-port="0" /> <edge from-layer="6817" from-port="0" to-layer="6820" to-port="1" /> <edge from-layer="6818" from-port="0" to-layer="6820" to-port="2" /> <edge from-layer="6819" from-port="0" to-layer="6820" to-port="3" /> <edge from-layer="6820" from-port="4" to-layer="6871" to-port="0" /> <edge from-layer="6820" from-port="4" to-layer="6866" to-port="0" /> <edge from-layer="6820" from-port="4" to-layer="6852" to-port="0" /> <edge from-layer="6820" from-port="4" to-layer="6855" to-port="0" /> <edge from-layer="6821" from-port="0" to-layer="6848" to-port="0" /> <edge from-layer="6822" from-port="0" to-layer="6848" to-port="1" /> <edge from-layer="6823" from-port="0" to-layer="6826" to-port="1" /> <edge from-layer="6824" from-port="0" to-layer="6826" to-port="2" /> <edge from-layer="6825" from-port="0" to-layer="6826" to-port="3" /> <edge from-layer="6826" from-port="4" to-layer="6828" to-port="0" /> <edge from-layer="6827" from-port="0" to-layer="6828" to-port="1" /> <edge from-layer="6828" from-port="2" to-layer="6829" to-port="0" /> <edge from-layer="6828" from-port="2" to-layer="6890" to-port="0" /> <edge from-layer="6828" from-port="2" to-layer="6917" to-port="0" /> <edge from-layer="6829" from-port="1" to-layer="6831" to-port="0" /> <edge from-layer="6830" from-port="0" to-layer="6831" to-port="2" /> <edge from-layer="6831" from-port="3" to-layer="6844" to-port="0" /> <edge from-layer="6832" from-port="0" to-layer="6837" to-port="0" /> <edge from-layer="6833" from-port="0" to-layer="6836" to-port="1" /> <edge from-layer="6834" from-port="0" to-layer="6836" to-port="2" /> <edge from-layer="6835" from-port="0" to-layer="6836" to-port="3" /> <edge from-layer="6836" from-port="4" to-layer="6837" to-port="1" /> <edge from-layer="6837" from-port="2" to-layer="6838" to-port="0" /> <edge from-layer="6838" from-port="1" to-layer="6840" to-port="0" /> <edge from-layer="6839" from-port="0" to-layer="6840" to-port="2" /> <edge from-layer="6840" from-port="3" to-layer="6841" to-port="0" /> <edge from-layer="6840" from-port="3" to-layer="6919" to-port="0" /> <edge from-layer="6841" from-port="1" to-layer="6843" to-port="0" /> <edge from-layer="6842" from-port="0" to-layer="6843" to-port="2" /> <edge from-layer="6843" from-port="3" to-layer="6844" to-port="1" /> <edge from-layer="6844" from-port="2" to-layer="6846" to-port="0" /> <edge from-layer="6845" from-port="0" to-layer="6846" to-port="1" /> <edge from-layer="6846" from-port="2" to-layer="6876" to-port="2" /> <edge from-layer="6846" from-port="2" to-layer="6848" to-port="2" /> <edge from-layer="6847" from-port="0" to-layer="6848" to-port="3" /> <edge from-layer="6848" from-port="4" to-layer="6850" to-port="0" /> <edge from-layer="6849" from-port="0" to-layer="6850" to-port="2" /> <edge from-layer="6850" from-port="3" to-layer="6851" to-port="0" /> <edge from-layer="6851" from-port="2" to-layer="6852" to-port="1" /> <edge from-layer="6851" from-port="2" to-layer="6891" to-port="1" /> <edge from-layer="6852" from-port="2" to-layer="6881" to-port="0" /> <edge from-layer="6853" from-port="0" to-layer="6863" to-port="0" /> <edge from-layer="6854" from-port="0" to-layer="6863" to-port="1" /> <edge from-layer="6855" from-port="1" to-layer="6857" to-port="0" /> <edge from-layer="6856" from-port="0" to-layer="6857" to-port="2" /> <edge from-layer="6857" from-port="3" to-layer="6858" to-port="0" /> <edge from-layer="6858" from-port="2" to-layer="6859" to-port="0" /> <edge from-layer="6859" from-port="1" to-layer="6861" to-port="0" /> <edge from-layer="6860" from-port="0" to-layer="6861" to-port="1" /> <edge from-layer="6861" from-port="2" to-layer="6863" to-port="2" /> <edge from-layer="6862" from-port="0" to-layer="6863" to-port="3" /> <edge from-layer="6863" from-port="4" to-layer="6866" to-port="1" /> <edge from-layer="6863" from-port="4" to-layer="6871" to-port="2" /> <edge from-layer="6864" from-port="0" to-layer="6866" to-port="2" /> <edge from-layer="6865" from-port="0" to-layer="6866" to-port="3" /> <edge from-layer="6866" from-port="4" to-layer="6868" to-port="0" /> <edge from-layer="6867" from-port="0" to-layer="6868" to-port="1" /> <edge from-layer="6868" from-port="2" to-layer="6872" to-port="0" /> <edge from-layer="6869" from-port="0" to-layer="6871" to-port="1" /> <edge from-layer="6870" from-port="0" to-layer="6871" to-port="3" /> <edge from-layer="6871" from-port="4" to-layer="6872" to-port="1" /> <edge from-layer="6872" from-port="2" to-layer="6880" to-port="0" /> <edge from-layer="6873" from-port="0" to-layer="6876" to-port="0" /> <edge from-layer="6874" from-port="0" to-layer="6876" to-port="1" /> <edge from-layer="6875" from-port="0" to-layer="6876" to-port="3" /> <edge from-layer="6876" from-port="4" to-layer="6878" to-port="0" /> <edge from-layer="6877" from-port="0" to-layer="6878" to-port="2" /> <edge from-layer="6878" from-port="3" to-layer="6879" to-port="0" /> <edge from-layer="6879" from-port="2" to-layer="6912" to-port="1" /> <edge from-layer="6879" from-port="2" to-layer="6880" to-port="1" /> <edge from-layer="6880" from-port="2" to-layer="6881" to-port="1" /> <edge from-layer="6881" from-port="2" to-layer="6886" to-port="0" /> <edge from-layer="6882" from-port="0" to-layer="6885" to-port="1" /> <edge from-layer="6883" from-port="0" to-layer="6885" to-port="2" /> <edge from-layer="6884" from-port="0" to-layer="6885" to-port="3" /> <edge from-layer="6885" from-port="4" to-layer="6886" to-port="1" /> <edge from-layer="6886" from-port="2" to-layer="6980" to-port="0" /> <edge from-layer="6886" from-port="2" to-layer="6942" to-port="0" /> <edge from-layer="6887" from-port="0" to-layer="6890" to-port="1" /> <edge from-layer="6888" from-port="0" to-layer="6890" to-port="2" /> <edge from-layer="6889" from-port="0" to-layer="6890" to-port="3" /> <edge from-layer="6890" from-port="4" to-layer="6894" to-port="0" /> <edge from-layer="6890" from-port="4" to-layer="6910" to-port="0" /> <edge from-layer="6890" from-port="4" to-layer="6905" to-port="0" /> <edge from-layer="6890" from-port="4" to-layer="6891" to-port="0" /> <edge from-layer="6891" from-port="2" to-layer="6913" to-port="0" /> <edge from-layer="6892" from-port="0" to-layer="6902" to-port="0" /> <edge from-layer="6893" from-port="0" to-layer="6902" to-port="1" /> <edge from-layer="6894" from-port="1" to-layer="6896" to-port="0" /> <edge from-layer="6895" from-port="0" to-layer="6896" to-port="2" /> <edge from-layer="6896" from-port="3" to-layer="6897" to-port="0" /> <edge from-layer="6897" from-port="2" to-layer="6898" to-port="0" /> <edge from-layer="6898" from-port="1" to-layer="6900" to-port="0" /> <edge from-layer="6899" from-port="0" to-layer="6900" to-port="1" /> <edge from-layer="6900" from-port="2" to-layer="6902" to-port="2" /> <edge from-layer="6901" from-port="0" to-layer="6902" to-port="3" /> <edge from-layer="6902" from-port="4" to-layer="6905" to-port="1" /> <edge from-layer="6902" from-port="4" to-layer="6910" to-port="2" /> <edge from-layer="6903" from-port="0" to-layer="6905" to-port="2" /> <edge from-layer="6904" from-port="0" to-layer="6905" to-port="3" /> <edge from-layer="6905" from-port="4" to-layer="6907" to-port="0" /> <edge from-layer="6906" from-port="0" to-layer="6907" to-port="1" /> <edge from-layer="6907" from-port="2" to-layer="6911" to-port="0" /> <edge from-layer="6908" from-port="0" to-layer="6910" to-port="1" /> <edge from-layer="6909" from-port="0" to-layer="6910" to-port="3" /> <edge from-layer="6910" from-port="4" to-layer="6911" to-port="1" /> <edge from-layer="6911" from-port="2" to-layer="6912" to-port="0" /> <edge from-layer="6912" from-port="2" to-layer="6913" to-port="1" /> <edge from-layer="6913" from-port="2" to-layer="6918" to-port="0" /> <edge from-layer="6914" from-port="0" to-layer="6917" to-port="1" /> <edge from-layer="6915" from-port="0" to-layer="6917" to-port="2" /> <edge from-layer="6916" from-port="0" to-layer="6917" to-port="3" /> <edge from-layer="6917" from-port="4" to-layer="6918" to-port="1" /> <edge from-layer="6918" from-port="2" to-layer="6919" to-port="1" /> <edge from-layer="6919" from-port="2" to-layer="6939" to-port="0" /> <edge from-layer="6919" from-port="2" to-layer="6980" to-port="1" /> <edge from-layer="6919" from-port="2" to-layer="10129" to-port="0" /> <edge from-layer="6920" from-port="0" to-layer="6925" to-port="0" /> <edge from-layer="6921" from-port="0" to-layer="6924" to-port="1" /> <edge from-layer="6922" from-port="0" to-layer="6924" to-port="2" /> <edge from-layer="6923" from-port="0" to-layer="6924" to-port="3" /> <edge from-layer="6924" from-port="4" to-layer="6925" to-port="1" /> <edge from-layer="6925" from-port="2" to-layer="6926" to-port="0" /> <edge from-layer="6926" from-port="1" to-layer="6928" to-port="0" /> <edge from-layer="6927" from-port="0" to-layer="6928" to-port="2" /> <edge from-layer="6928" from-port="3" to-layer="6935" to-port="0" /> <edge from-layer="6929" from-port="0" to-layer="6932" to-port="1" /> <edge from-layer="6930" from-port="0" to-layer="6932" to-port="2" /> <edge from-layer="6931" from-port="0" to-layer="6932" to-port="3" /> <edge from-layer="6932" from-port="4" to-layer="6934" to-port="0" /> <edge from-layer="6933" from-port="0" to-layer="6934" to-port="1" /> <edge from-layer="6934" from-port="2" to-layer="6935" to-port="1" /> <edge from-layer="6935" from-port="2" to-layer="6980" to-port="2" /> <edge from-layer="6935" from-port="2" to-layer="10128" to-port="0" /> <edge from-layer="6936" from-port="0" to-layer="6954" to-port="0" /> <edge from-layer="6937" from-port="0" to-layer="6949" to-port="0" /> <edge from-layer="6938" from-port="0" to-layer="6952" to-port="1" /> <edge from-layer="6938" from-port="0" to-layer="6949" to-port="1" /> <edge from-layer="6939" from-port="1" to-layer="6941" to-port="0" /> <edge from-layer="6940" from-port="0" to-layer="6941" to-port="2" /> <edge from-layer="6941" from-port="3" to-layer="6945" to-port="0" /> <edge from-layer="6941" from-port="3" to-layer="6951" to-port="0" /> <edge from-layer="6942" from-port="1" to-layer="6944" to-port="0" /> <edge from-layer="6942" from-port="1" to-layer="6969" to-port="0" /> <edge from-layer="6943" from-port="0" to-layer="6944" to-port="2" /> <edge from-layer="6944" from-port="3" to-layer="6945" to-port="1" /> <edge from-layer="6945" from-port="2" to-layer="6947" to-port="0" /> <edge from-layer="6946" from-port="0" to-layer="6947" to-port="1" /> <edge from-layer="6946" from-port="0" to-layer="6951" to-port="1" /> <edge from-layer="6947" from-port="2" to-layer="6949" to-port="2" /> <edge from-layer="6948" from-port="0" to-layer="6949" to-port="3" /> <edge from-layer="6948" from-port="0" to-layer="6952" to-port="3" /> <edge from-layer="6949" from-port="4" to-layer="6954" to-port="1" /> <edge from-layer="6950" from-port="0" to-layer="6952" to-port="0" /> <edge from-layer="6951" from-port="2" to-layer="6959" to-port="2" /> <edge from-layer="6951" from-port="2" to-layer="6952" to-port="2" /> <edge from-layer="6952" from-port="4" to-layer="6954" to-port="2" /> <edge from-layer="6953" from-port="0" to-layer="6954" to-port="3" /> <edge from-layer="6954" from-port="4" to-layer="6961" to-port="0" /> <edge from-layer="6955" from-port="0" to-layer="6961" to-port="1" /> <edge from-layer="6956" from-port="0" to-layer="6959" to-port="0" /> <edge from-layer="6957" from-port="0" to-layer="6959" to-port="1" /> <edge from-layer="6958" from-port="0" to-layer="6959" to-port="3" /> <edge from-layer="6959" from-port="4" to-layer="6961" to-port="2" /> <edge from-layer="6960" from-port="0" to-layer="6961" to-port="3" /> <edge from-layer="6961" from-port="4" to-layer="6964" to-port="0" /> <edge from-layer="6962" from-port="0" to-layer="6964" to-port="1" /> <edge from-layer="6963" from-port="0" to-layer="6964" to-port="2" /> <edge from-layer="6964" from-port="3" to-layer="6965" to-port="0" /> <edge from-layer="6965" from-port="2" to-layer="6966" to-port="0" /> <edge from-layer="6965" from-port="2" to-layer="6979" to-port="0" /> <edge from-layer="6966" from-port="1" to-layer="6978" to-port="0" /> <edge from-layer="6967" from-port="0" to-layer="6969" to-port="1" /> <edge from-layer="6968" from-port="0" to-layer="6969" to-port="2" /> <edge from-layer="6969" from-port="3" to-layer="6973" to-port="0" /> <edge from-layer="6970" from-port="0" to-layer="6973" to-port="1" /> <edge from-layer="6971" from-port="0" to-layer="6973" to-port="2" /> <edge from-layer="6972" from-port="0" to-layer="6973" to-port="3" /> <edge from-layer="6973" from-port="4" to-layer="6975" to-port="0" /> <edge from-layer="6973" from-port="4" to-layer="6977" to-port="2" /> <edge from-layer="6974" from-port="0" to-layer="6975" to-port="1" /> <edge from-layer="6975" from-port="2" to-layer="6977" to-port="0" /> <edge from-layer="6976" from-port="0" to-layer="6977" to-port="1" /> <edge from-layer="6977" from-port="3" to-layer="6978" to-port="1" /> <edge from-layer="6978" from-port="2" to-layer="6979" to-port="1" /> <edge from-layer="6979" from-port="2" to-layer="6980" to-port="3" /> <edge from-layer="6980" from-port="4" to-layer="6982" to-port="0" /> <edge from-layer="6981" from-port="0" to-layer="6982" to-port="1" /> <edge from-layer="6982" from-port="2" to-layer="6984" to-port="0" /> <edge from-layer="6983" from-port="0" to-layer="6984" to-port="1" /> <edge from-layer="6984" from-port="2" to-layer="6992" to-port="0" /> <edge from-layer="6985" from-port="0" to-layer="6986" to-port="0" /> <edge from-layer="6986" from-port="1" to-layer="6989" to-port="0" /> <edge from-layer="6987" from-port="0" to-layer="6988" to-port="0" /> <edge from-layer="6988" from-port="1" to-layer="6989" to-port="1" /> <edge from-layer="6989" from-port="2" to-layer="6991" to-port="0" /> <edge from-layer="6990" from-port="0" to-layer="6991" to-port="1" /> <edge from-layer="6991" from-port="2" to-layer="6992" to-port="1" /> <edge from-layer="6992" from-port="2" to-layer="6994" to-port="0" /> <edge from-layer="6993" from-port="0" to-layer="6994" to-port="1" /> <edge from-layer="6994" from-port="2" to-layer="6995" to-port="1" /> <edge from-layer="6995" from-port="2" to-layer="6996" to-port="0" /> <edge from-layer="6996" from-port="2" to-layer="7025" to-port="0" /> <edge from-layer="6996" from-port="2" to-layer="6998" to-port="0" /> <edge from-layer="6996" from-port="2" to-layer="7227" to-port="1" /> <edge from-layer="6997" from-port="0" to-layer="6998" to-port="1" /> <edge from-layer="6998" from-port="2" to-layer="7000" to-port="0" /> <edge from-layer="6999" from-port="0" to-layer="7000" to-port="1" /> <edge from-layer="7000" from-port="2" to-layer="7002" to-port="0" /> <edge from-layer="7001" from-port="0" to-layer="7002" to-port="1" /> <edge from-layer="7002" from-port="2" to-layer="7010" to-port="0" /> <edge from-layer="7003" from-port="0" to-layer="7004" to-port="0" /> <edge from-layer="7004" from-port="1" to-layer="7007" to-port="0" /> <edge from-layer="7005" from-port="0" to-layer="7006" to-port="0" /> <edge from-layer="7006" from-port="1" to-layer="7007" to-port="1" /> <edge from-layer="7007" from-port="2" to-layer="7009" to-port="0" /> <edge from-layer="7008" from-port="0" to-layer="7009" to-port="1" /> <edge from-layer="7009" from-port="2" to-layer="7010" to-port="1" /> <edge from-layer="7010" from-port="2" to-layer="7012" to-port="0" /> <edge from-layer="7011" from-port="0" to-layer="7012" to-port="1" /> <edge from-layer="7012" from-port="2" to-layer="7013" to-port="0" /> <edge from-layer="7013" from-port="1" to-layer="7021" to-port="0" /> <edge from-layer="7014" from-port="0" to-layer="7015" to-port="0" /> <edge from-layer="7015" from-port="1" to-layer="7018" to-port="0" /> <edge from-layer="7016" from-port="0" to-layer="7017" to-port="0" /> <edge from-layer="7017" from-port="1" to-layer="7018" to-port="1" /> <edge from-layer="7018" from-port="2" to-layer="7020" to-port="0" /> <edge from-layer="7019" from-port="0" to-layer="7020" to-port="1" /> <edge from-layer="7020" from-port="2" to-layer="7021" to-port="1" /> <edge from-layer="7021" from-port="2" to-layer="7023" to-port="0" /> <edge from-layer="7022" from-port="0" to-layer="7023" to-port="1" /> <edge from-layer="7023" from-port="2" to-layer="7226" to-port="0" /> <edge from-layer="7024" from-port="0" to-layer="7025" to-port="1" /> <edge from-layer="7025" from-port="2" to-layer="7027" to-port="0" /> <edge from-layer="7026" from-port="0" to-layer="7027" to-port="1" /> <edge from-layer="7027" from-port="2" to-layer="7029" to-port="0" /> <edge from-layer="7028" from-port="0" to-layer="7029" to-port="1" /> <edge from-layer="7029" from-port="2" to-layer="7037" to-port="0" /> <edge from-layer="7030" from-port="0" to-layer="7031" to-port="0" /> <edge from-layer="7031" from-port="1" to-layer="7034" to-port="0" /> <edge from-layer="7032" from-port="0" to-layer="7033" to-port="0" /> <edge from-layer="7033" from-port="1" to-layer="7034" to-port="1" /> <edge from-layer="7034" from-port="2" to-layer="7036" to-port="0" /> <edge from-layer="7035" from-port="0" to-layer="7036" to-port="1" /> <edge from-layer="7036" from-port="2" to-layer="7037" to-port="1" /> <edge from-layer="7037" from-port="2" to-layer="7039" to-port="0" /> <edge from-layer="7038" from-port="0" to-layer="7039" to-port="1" /> <edge from-layer="7039" from-port="2" to-layer="7041" to-port="0" /> <edge from-layer="7040" from-port="0" to-layer="7041" to-port="1" /> <edge from-layer="7041" from-port="2" to-layer="7057" to-port="0" /> <edge from-layer="7041" from-port="2" to-layer="7163" to-port="0" /> <edge from-layer="7041" from-port="2" to-layer="7045" to-port="0" /> <edge from-layer="7042" from-port="0" to-layer="7045" to-port="1" /> <edge from-layer="7043" from-port="0" to-layer="7045" to-port="2" /> <edge from-layer="7044" from-port="0" to-layer="7045" to-port="3" /> <edge from-layer="7045" from-port="4" to-layer="7047" to-port="0" /> <edge from-layer="7046" from-port="0" to-layer="7047" to-port="1" /> <edge from-layer="7047" from-port="2" to-layer="7116" to-port="0" /> <edge from-layer="7047" from-port="2" to-layer="7051" to-port="0" /> <edge from-layer="7048" from-port="0" to-layer="7051" to-port="1" /> <edge from-layer="7049" from-port="0" to-layer="7051" to-port="2" /> <edge from-layer="7050" from-port="0" to-layer="7051" to-port="3" /> <edge from-layer="7051" from-port="4" to-layer="7086" to-port="0" /> <edge from-layer="7051" from-port="4" to-layer="7097" to-port="0" /> <edge from-layer="7051" from-port="4" to-layer="7083" to-port="0" /> <edge from-layer="7051" from-port="4" to-layer="7102" to-port="0" /> <edge from-layer="7052" from-port="0" to-layer="7079" to-port="0" /> <edge from-layer="7053" from-port="0" to-layer="7079" to-port="1" /> <edge from-layer="7054" from-port="0" to-layer="7057" to-port="1" /> <edge from-layer="7055" from-port="0" to-layer="7057" to-port="2" /> <edge from-layer="7056" from-port="0" to-layer="7057" to-port="3" /> <edge from-layer="7057" from-port="4" to-layer="7059" to-port="0" /> <edge from-layer="7058" from-port="0" to-layer="7059" to-port="1" /> <edge from-layer="7059" from-port="2" to-layer="7060" to-port="0" /> <edge from-layer="7059" from-port="2" to-layer="7148" to-port="0" /> <edge from-layer="7059" from-port="2" to-layer="7121" to-port="0" /> <edge from-layer="7060" from-port="1" to-layer="7062" to-port="0" /> <edge from-layer="7061" from-port="0" to-layer="7062" to-port="2" /> <edge from-layer="7062" from-port="3" to-layer="7075" to-port="0" /> <edge from-layer="7063" from-port="0" to-layer="7068" to-port="0" /> <edge from-layer="7064" from-port="0" to-layer="7067" to-port="1" /> <edge from-layer="7065" from-port="0" to-layer="7067" to-port="2" /> <edge from-layer="7066" from-port="0" to-layer="7067" to-port="3" /> <edge from-layer="7067" from-port="4" to-layer="7068" to-port="1" /> <edge from-layer="7068" from-port="2" to-layer="7069" to-port="0" /> <edge from-layer="7069" from-port="1" to-layer="7071" to-port="0" /> <edge from-layer="7070" from-port="0" to-layer="7071" to-port="2" /> <edge from-layer="7071" from-port="3" to-layer="7072" to-port="0" /> <edge from-layer="7071" from-port="3" to-layer="7150" to-port="0" /> <edge from-layer="7072" from-port="1" to-layer="7074" to-port="0" /> <edge from-layer="7073" from-port="0" to-layer="7074" to-port="2" /> <edge from-layer="7074" from-port="3" to-layer="7075" to-port="1" /> <edge from-layer="7075" from-port="2" to-layer="7077" to-port="0" /> <edge from-layer="7076" from-port="0" to-layer="7077" to-port="1" /> <edge from-layer="7077" from-port="2" to-layer="7079" to-port="2" /> <edge from-layer="7077" from-port="2" to-layer="7107" to-port="2" /> <edge from-layer="7078" from-port="0" to-layer="7079" to-port="3" /> <edge from-layer="7079" from-port="4" to-layer="7081" to-port="0" /> <edge from-layer="7080" from-port="0" to-layer="7081" to-port="2" /> <edge from-layer="7081" from-port="3" to-layer="7082" to-port="0" /> <edge from-layer="7082" from-port="2" to-layer="7122" to-port="1" /> <edge from-layer="7082" from-port="2" to-layer="7083" to-port="1" /> <edge from-layer="7083" from-port="2" to-layer="7112" to-port="0" /> <edge from-layer="7084" from-port="0" to-layer="7094" to-port="0" /> <edge from-layer="7085" from-port="0" to-layer="7094" to-port="1" /> <edge from-layer="7086" from-port="1" to-layer="7088" to-port="0" /> <edge from-layer="7087" from-port="0" to-layer="7088" to-port="2" /> <edge from-layer="7088" from-port="3" to-layer="7089" to-port="0" /> <edge from-layer="7089" from-port="2" to-layer="7090" to-port="0" /> <edge from-layer="7090" from-port="1" to-layer="7092" to-port="0" /> <edge from-layer="7091" from-port="0" to-layer="7092" to-port="1" /> <edge from-layer="7092" from-port="2" to-layer="7094" to-port="2" /> <edge from-layer="7093" from-port="0" to-layer="7094" to-port="3" /> <edge from-layer="7094" from-port="4" to-layer="7097" to-port="1" /> <edge from-layer="7094" from-port="4" to-layer="7102" to-port="2" /> <edge from-layer="7095" from-port="0" to-layer="7097" to-port="2" /> <edge from-layer="7096" from-port="0" to-layer="7097" to-port="3" /> <edge from-layer="7097" from-port="4" to-layer="7099" to-port="0" /> <edge from-layer="7098" from-port="0" to-layer="7099" to-port="1" /> <edge from-layer="7099" from-port="2" to-layer="7103" to-port="0" /> <edge from-layer="7100" from-port="0" to-layer="7102" to-port="1" /> <edge from-layer="7101" from-port="0" to-layer="7102" to-port="3" /> <edge from-layer="7102" from-port="4" to-layer="7103" to-port="1" /> <edge from-layer="7103" from-port="2" to-layer="7111" to-port="0" /> <edge from-layer="7104" from-port="0" to-layer="7107" to-port="0" /> <edge from-layer="7105" from-port="0" to-layer="7107" to-port="1" /> <edge from-layer="7106" from-port="0" to-layer="7107" to-port="3" /> <edge from-layer="7107" from-port="4" to-layer="7109" to-port="0" /> <edge from-layer="7108" from-port="0" to-layer="7109" to-port="2" /> <edge from-layer="7109" from-port="3" to-layer="7110" to-port="0" /> <edge from-layer="7110" from-port="2" to-layer="7143" to-port="1" /> <edge from-layer="7110" from-port="2" to-layer="7111" to-port="1" /> <edge from-layer="7111" from-port="2" to-layer="7112" to-port="1" /> <edge from-layer="7112" from-port="2" to-layer="7117" to-port="0" /> <edge from-layer="7113" from-port="0" to-layer="7116" to-port="1" /> <edge from-layer="7114" from-port="0" to-layer="7116" to-port="2" /> <edge from-layer="7115" from-port="0" to-layer="7116" to-port="3" /> <edge from-layer="7116" from-port="4" to-layer="7117" to-port="1" /> <edge from-layer="7117" from-port="2" to-layer="7211" to-port="0" /> <edge from-layer="7117" from-port="2" to-layer="7173" to-port="0" /> <edge from-layer="7118" from-port="0" to-layer="7121" to-port="1" /> <edge from-layer="7119" from-port="0" to-layer="7121" to-port="2" /> <edge from-layer="7120" from-port="0" to-layer="7121" to-port="3" /> <edge from-layer="7121" from-port="4" to-layer="7141" to-port="0" /> <edge from-layer="7121" from-port="4" to-layer="7136" to-port="0" /> <edge from-layer="7121" from-port="4" to-layer="7122" to-port="0" /> <edge from-layer="7121" from-port="4" to-layer="7125" to-port="0" /> <edge from-layer="7122" from-port="2" to-layer="7144" to-port="0" /> <edge from-layer="7123" from-port="0" to-layer="7133" to-port="0" /> <edge from-layer="7124" from-port="0" to-layer="7133" to-port="1" /> <edge from-layer="7125" from-port="1" to-layer="7127" to-port="0" /> <edge from-layer="7126" from-port="0" to-layer="7127" to-port="2" /> <edge from-layer="7127" from-port="3" to-layer="7128" to-port="0" /> <edge from-layer="7128" from-port="2" to-layer="7129" to-port="0" /> <edge from-layer="7129" from-port="1" to-layer="7131" to-port="0" /> <edge from-layer="7130" from-port="0" to-layer="7131" to-port="1" /> <edge from-layer="7131" from-port="2" to-layer="7133" to-port="2" /> <edge from-layer="7132" from-port="0" to-layer="7133" to-port="3" /> <edge from-layer="7133" from-port="4" to-layer="7141" to-port="2" /> <edge from-layer="7133" from-port="4" to-layer="7136" to-port="1" /> <edge from-layer="7134" from-port="0" to-layer="7136" to-port="2" /> <edge from-layer="7135" from-port="0" to-layer="7136" to-port="3" /> <edge from-layer="7136" from-port="4" to-layer="7138" to-port="0" /> <edge from-layer="7137" from-port="0" to-layer="7138" to-port="1" /> <edge from-layer="7138" from-port="2" to-layer="7142" to-port="0" /> <edge from-layer="7139" from-port="0" to-layer="7141" to-port="1" /> <edge from-layer="7140" from-port="0" to-layer="7141" to-port="3" /> <edge from-layer="7141" from-port="4" to-layer="7142" to-port="1" /> <edge from-layer="7142" from-port="2" to-layer="7143" to-port="0" /> <edge from-layer="7143" from-port="2" to-layer="7144" to-port="1" /> <edge from-layer="7144" from-port="2" to-layer="7149" to-port="0" /> <edge from-layer="7145" from-port="0" to-layer="7148" to-port="1" /> <edge from-layer="7146" from-port="0" to-layer="7148" to-port="2" /> <edge from-layer="7147" from-port="0" to-layer="7148" to-port="3" /> <edge from-layer="7148" from-port="4" to-layer="7149" to-port="1" /> <edge from-layer="7149" from-port="2" to-layer="7150" to-port="1" /> <edge from-layer="7150" from-port="2" to-layer="7211" to-port="1" /> <edge from-layer="7150" from-port="2" to-layer="7170" to-port="0" /> <edge from-layer="7150" from-port="2" to-layer="10125" to-port="0" /> <edge from-layer="7151" from-port="0" to-layer="7156" to-port="0" /> <edge from-layer="7152" from-port="0" to-layer="7155" to-port="1" /> <edge from-layer="7153" from-port="0" to-layer="7155" to-port="2" /> <edge from-layer="7154" from-port="0" to-layer="7155" to-port="3" /> <edge from-layer="7155" from-port="4" to-layer="7156" to-port="1" /> <edge from-layer="7156" from-port="2" to-layer="7157" to-port="0" /> <edge from-layer="7157" from-port="1" to-layer="7159" to-port="0" /> <edge from-layer="7158" from-port="0" to-layer="7159" to-port="2" /> <edge from-layer="7159" from-port="3" to-layer="7166" to-port="0" /> <edge from-layer="7160" from-port="0" to-layer="7163" to-port="1" /> <edge from-layer="7161" from-port="0" to-layer="7163" to-port="2" /> <edge from-layer="7162" from-port="0" to-layer="7163" to-port="3" /> <edge from-layer="7163" from-port="4" to-layer="7165" to-port="0" /> <edge from-layer="7164" from-port="0" to-layer="7165" to-port="1" /> <edge from-layer="7165" from-port="2" to-layer="7166" to-port="1" /> <edge from-layer="7166" from-port="2" to-layer="7211" to-port="2" /> <edge from-layer="7166" from-port="2" to-layer="10124" to-port="0" /> <edge from-layer="7167" from-port="0" to-layer="7185" to-port="0" /> <edge from-layer="7168" from-port="0" to-layer="7180" to-port="0" /> <edge from-layer="7169" from-port="0" to-layer="7180" to-port="1" /> <edge from-layer="7169" from-port="0" to-layer="7183" to-port="1" /> <edge from-layer="7170" from-port="1" to-layer="7172" to-port="0" /> <edge from-layer="7171" from-port="0" to-layer="7172" to-port="2" /> <edge from-layer="7172" from-port="3" to-layer="7182" to-port="0" /> <edge from-layer="7172" from-port="3" to-layer="7176" to-port="0" /> <edge from-layer="7173" from-port="1" to-layer="7200" to-port="0" /> <edge from-layer="7173" from-port="1" to-layer="7175" to-port="0" /> <edge from-layer="7174" from-port="0" to-layer="7175" to-port="2" /> <edge from-layer="7175" from-port="3" to-layer="7176" to-port="1" /> <edge from-layer="7176" from-port="2" to-layer="7178" to-port="0" /> <edge from-layer="7177" from-port="0" to-layer="7178" to-port="1" /> <edge from-layer="7177" from-port="0" to-layer="7182" to-port="1" /> <edge from-layer="7178" from-port="2" to-layer="7180" to-port="2" /> <edge from-layer="7179" from-port="0" to-layer="7183" to-port="3" /> <edge from-layer="7179" from-port="0" to-layer="7180" to-port="3" /> <edge from-layer="7180" from-port="4" to-layer="7185" to-port="1" /> <edge from-layer="7181" from-port="0" to-layer="7183" to-port="0" /> <edge from-layer="7182" from-port="2" to-layer="7183" to-port="2" /> <edge from-layer="7182" from-port="2" to-layer="7190" to-port="2" /> <edge from-layer="7183" from-port="4" to-layer="7185" to-port="2" /> <edge from-layer="7184" from-port="0" to-layer="7185" to-port="3" /> <edge from-layer="7185" from-port="4" to-layer="7192" to-port="0" /> <edge from-layer="7186" from-port="0" to-layer="7192" to-port="1" /> <edge from-layer="7187" from-port="0" to-layer="7190" to-port="0" /> <edge from-layer="7188" from-port="0" to-layer="7190" to-port="1" /> <edge from-layer="7189" from-port="0" to-layer="7190" to-port="3" /> <edge from-layer="7190" from-port="4" to-layer="7192" to-port="2" /> <edge from-layer="7191" from-port="0" to-layer="7192" to-port="3" /> <edge from-layer="7192" from-port="4" to-layer="7195" to-port="0" /> <edge from-layer="7193" from-port="0" to-layer="7195" to-port="1" /> <edge from-layer="7194" from-port="0" to-layer="7195" to-port="2" /> <edge from-layer="7195" from-port="3" to-layer="7196" to-port="0" /> <edge from-layer="7196" from-port="2" to-layer="7210" to-port="0" /> <edge from-layer="7196" from-port="2" to-layer="7197" to-port="0" /> <edge from-layer="7197" from-port="1" to-layer="7209" to-port="0" /> <edge from-layer="7198" from-port="0" to-layer="7200" to-port="1" /> <edge from-layer="7199" from-port="0" to-layer="7200" to-port="2" /> <edge from-layer="7200" from-port="3" to-layer="7204" to-port="0" /> <edge from-layer="7201" from-port="0" to-layer="7204" to-port="1" /> <edge from-layer="7202" from-port="0" to-layer="7204" to-port="2" /> <edge from-layer="7203" from-port="0" to-layer="7204" to-port="3" /> <edge from-layer="7204" from-port="4" to-layer="7208" to-port="2" /> <edge from-layer="7204" from-port="4" to-layer="7206" to-port="0" /> <edge from-layer="7205" from-port="0" to-layer="7206" to-port="1" /> <edge from-layer="7206" from-port="2" to-layer="7208" to-port="0" /> <edge from-layer="7207" from-port="0" to-layer="7208" to-port="1" /> <edge from-layer="7208" from-port="3" to-layer="7209" to-port="1" /> <edge from-layer="7209" from-port="2" to-layer="7210" to-port="1" /> <edge from-layer="7210" from-port="2" to-layer="7211" to-port="3" /> <edge from-layer="7211" from-port="4" to-layer="7213" to-port="0" /> <edge from-layer="7212" from-port="0" to-layer="7213" to-port="1" /> <edge from-layer="7213" from-port="2" to-layer="7215" to-port="0" /> <edge from-layer="7214" from-port="0" to-layer="7215" to-port="1" /> <edge from-layer="7215" from-port="2" to-layer="7223" to-port="0" /> <edge from-layer="7216" from-port="0" to-layer="7217" to-port="0" /> <edge from-layer="7217" from-port="1" to-layer="7220" to-port="0" /> <edge from-layer="7218" from-port="0" to-layer="7219" to-port="0" /> <edge from-layer="7219" from-port="1" to-layer="7220" to-port="1" /> <edge from-layer="7220" from-port="2" to-layer="7222" to-port="0" /> <edge from-layer="7221" from-port="0" to-layer="7222" to-port="1" /> <edge from-layer="7222" from-port="2" to-layer="7223" to-port="1" /> <edge from-layer="7223" from-port="2" to-layer="7225" to-port="0" /> <edge from-layer="7224" from-port="0" to-layer="7225" to-port="1" /> <edge from-layer="7225" from-port="2" to-layer="7226" to-port="1" /> <edge from-layer="7226" from-port="2" to-layer="7227" to-port="0" /> <edge from-layer="7227" from-port="2" to-layer="7458" to-port="1" /> <edge from-layer="7227" from-port="2" to-layer="7256" to-port="0" /> <edge from-layer="7227" from-port="2" to-layer="7229" to-port="0" /> <edge from-layer="7228" from-port="0" to-layer="7229" to-port="1" /> <edge from-layer="7229" from-port="2" to-layer="7231" to-port="0" /> <edge from-layer="7230" from-port="0" to-layer="7231" to-port="1" /> <edge from-layer="7231" from-port="2" to-layer="7233" to-port="0" /> <edge from-layer="7232" from-port="0" to-layer="7233" to-port="1" /> <edge from-layer="7233" from-port="2" to-layer="7241" to-port="0" /> <edge from-layer="7234" from-port="0" to-layer="7235" to-port="0" /> <edge from-layer="7235" from-port="1" to-layer="7238" to-port="0" /> <edge from-layer="7236" from-port="0" to-layer="7237" to-port="0" /> <edge from-layer="7237" from-port="1" to-layer="7238" to-port="1" /> <edge from-layer="7238" from-port="2" to-layer="7240" to-port="0" /> <edge from-layer="7239" from-port="0" to-layer="7240" to-port="1" /> <edge from-layer="7240" from-port="2" to-layer="7241" to-port="1" /> <edge from-layer="7241" from-port="2" to-layer="7243" to-port="0" /> <edge from-layer="7242" from-port="0" to-layer="7243" to-port="1" /> <edge from-layer="7243" from-port="2" to-layer="7244" to-port="0" /> <edge from-layer="7244" from-port="1" to-layer="7252" to-port="0" /> <edge from-layer="7245" from-port="0" to-layer="7246" to-port="0" /> <edge from-layer="7246" from-port="1" to-layer="7249" to-port="0" /> <edge from-layer="7247" from-port="0" to-layer="7248" to-port="0" /> <edge from-layer="7248" from-port="1" to-layer="7249" to-port="1" /> <edge from-layer="7249" from-port="2" to-layer="7251" to-port="0" /> <edge from-layer="7250" from-port="0" to-layer="7251" to-port="1" /> <edge from-layer="7251" from-port="2" to-layer="7252" to-port="1" /> <edge from-layer="7252" from-port="2" to-layer="7254" to-port="0" /> <edge from-layer="7253" from-port="0" to-layer="7254" to-port="1" /> <edge from-layer="7254" from-port="2" to-layer="7457" to-port="0" /> <edge from-layer="7255" from-port="0" to-layer="7256" to-port="1" /> <edge from-layer="7256" from-port="2" to-layer="7258" to-port="0" /> <edge from-layer="7257" from-port="0" to-layer="7258" to-port="1" /> <edge from-layer="7258" from-port="2" to-layer="7260" to-port="0" /> <edge from-layer="7259" from-port="0" to-layer="7260" to-port="1" /> <edge from-layer="7260" from-port="2" to-layer="7268" to-port="0" /> <edge from-layer="7261" from-port="0" to-layer="7262" to-port="0" /> <edge from-layer="7262" from-port="1" to-layer="7265" to-port="0" /> <edge from-layer="7263" from-port="0" to-layer="7264" to-port="0" /> <edge from-layer="7264" from-port="1" to-layer="7265" to-port="1" /> <edge from-layer="7265" from-port="2" to-layer="7267" to-port="0" /> <edge from-layer="7266" from-port="0" to-layer="7267" to-port="1" /> <edge from-layer="7267" from-port="2" to-layer="7268" to-port="1" /> <edge from-layer="7268" from-port="2" to-layer="7270" to-port="0" /> <edge from-layer="7269" from-port="0" to-layer="7270" to-port="1" /> <edge from-layer="7270" from-port="2" to-layer="7272" to-port="0" /> <edge from-layer="7271" from-port="0" to-layer="7272" to-port="1" /> <edge from-layer="7272" from-port="2" to-layer="7276" to-port="0" /> <edge from-layer="7272" from-port="2" to-layer="7288" to-port="0" /> <edge from-layer="7272" from-port="2" to-layer="7394" to-port="0" /> <edge from-layer="7273" from-port="0" to-layer="7276" to-port="1" /> <edge from-layer="7274" from-port="0" to-layer="7276" to-port="2" /> <edge from-layer="7275" from-port="0" to-layer="7276" to-port="3" /> <edge from-layer="7276" from-port="4" to-layer="7278" to-port="0" /> <edge from-layer="7277" from-port="0" to-layer="7278" to-port="1" /> <edge from-layer="7278" from-port="2" to-layer="7347" to-port="0" /> <edge from-layer="7278" from-port="2" to-layer="7282" to-port="0" /> <edge from-layer="7279" from-port="0" to-layer="7282" to-port="1" /> <edge from-layer="7280" from-port="0" to-layer="7282" to-port="2" /> <edge from-layer="7281" from-port="0" to-layer="7282" to-port="3" /> <edge from-layer="7282" from-port="4" to-layer="7314" to-port="0" /> <edge from-layer="7282" from-port="4" to-layer="7317" to-port="0" /> <edge from-layer="7282" from-port="4" to-layer="7333" to-port="0" /> <edge from-layer="7282" from-port="4" to-layer="7328" to-port="0" /> <edge from-layer="7283" from-port="0" to-layer="7310" to-port="0" /> <edge from-layer="7284" from-port="0" to-layer="7310" to-port="1" /> <edge from-layer="7285" from-port="0" to-layer="7288" to-port="1" /> <edge from-layer="7286" from-port="0" to-layer="7288" to-port="2" /> <edge from-layer="7287" from-port="0" to-layer="7288" to-port="3" /> <edge from-layer="7288" from-port="4" to-layer="7290" to-port="0" /> <edge from-layer="7289" from-port="0" to-layer="7290" to-port="1" /> <edge from-layer="7290" from-port="2" to-layer="7379" to-port="0" /> <edge from-layer="7290" from-port="2" to-layer="7352" to-port="0" /> <edge from-layer="7290" from-port="2" to-layer="7291" to-port="0" /> <edge from-layer="7291" from-port="1" to-layer="7293" to-port="0" /> <edge from-layer="7292" from-port="0" to-layer="7293" to-port="2" /> <edge from-layer="7293" from-port="3" to-layer="7306" to-port="0" /> <edge from-layer="7294" from-port="0" to-layer="7299" to-port="0" /> <edge from-layer="7295" from-port="0" to-layer="7298" to-port="1" /> <edge from-layer="7296" from-port="0" to-layer="7298" to-port="2" /> <edge from-layer="7297" from-port="0" to-layer="7298" to-port="3" /> <edge from-layer="7298" from-port="4" to-layer="7299" to-port="1" /> <edge from-layer="7299" from-port="2" to-layer="7300" to-port="0" /> <edge from-layer="7300" from-port="1" to-layer="7302" to-port="0" /> <edge from-layer="7301" from-port="0" to-layer="7302" to-port="2" /> <edge from-layer="7302" from-port="3" to-layer="7303" to-port="0" /> <edge from-layer="7302" from-port="3" to-layer="7381" to-port="0" /> <edge from-layer="7303" from-port="1" to-layer="7305" to-port="0" /> <edge from-layer="7304" from-port="0" to-layer="7305" to-port="2" /> <edge from-layer="7305" from-port="3" to-layer="7306" to-port="1" /> <edge from-layer="7306" from-port="2" to-layer="7308" to-port="0" /> <edge from-layer="7307" from-port="0" to-layer="7308" to-port="1" /> <edge from-layer="7308" from-port="2" to-layer="7310" to-port="2" /> <edge from-layer="7308" from-port="2" to-layer="7338" to-port="2" /> <edge from-layer="7309" from-port="0" to-layer="7310" to-port="3" /> <edge from-layer="7310" from-port="4" to-layer="7312" to-port="0" /> <edge from-layer="7311" from-port="0" to-layer="7312" to-port="2" /> <edge from-layer="7312" from-port="3" to-layer="7313" to-port="0" /> <edge from-layer="7313" from-port="2" to-layer="7314" to-port="1" /> <edge from-layer="7313" from-port="2" to-layer="7353" to-port="1" /> <edge from-layer="7314" from-port="2" to-layer="7343" to-port="0" /> <edge from-layer="7315" from-port="0" to-layer="7325" to-port="0" /> <edge from-layer="7316" from-port="0" to-layer="7325" to-port="1" /> <edge from-layer="7317" from-port="1" to-layer="7319" to-port="0" /> <edge from-layer="7318" from-port="0" to-layer="7319" to-port="2" /> <edge from-layer="7319" from-port="3" to-layer="7320" to-port="0" /> <edge from-layer="7320" from-port="2" to-layer="7321" to-port="0" /> <edge from-layer="7321" from-port="1" to-layer="7323" to-port="0" /> <edge from-layer="7322" from-port="0" to-layer="7323" to-port="1" /> <edge from-layer="7323" from-port="2" to-layer="7325" to-port="2" /> <edge from-layer="7324" from-port="0" to-layer="7325" to-port="3" /> <edge from-layer="7325" from-port="4" to-layer="7328" to-port="1" /> <edge from-layer="7325" from-port="4" to-layer="7333" to-port="2" /> <edge from-layer="7326" from-port="0" to-layer="7328" to-port="2" /> <edge from-layer="7327" from-port="0" to-layer="7328" to-port="3" /> <edge from-layer="7328" from-port="4" to-layer="7330" to-port="0" /> <edge from-layer="7329" from-port="0" to-layer="7330" to-port="1" /> <edge from-layer="7330" from-port="2" to-layer="7334" to-port="0" /> <edge from-layer="7331" from-port="0" to-layer="7333" to-port="1" /> <edge from-layer="7332" from-port="0" to-layer="7333" to-port="3" /> <edge from-layer="7333" from-port="4" to-layer="7334" to-port="1" /> <edge from-layer="7334" from-port="2" to-layer="7342" to-port="0" /> <edge from-layer="7335" from-port="0" to-layer="7338" to-port="0" /> <edge from-layer="7336" from-port="0" to-layer="7338" to-port="1" /> <edge from-layer="7337" from-port="0" to-layer="7338" to-port="3" /> <edge from-layer="7338" from-port="4" to-layer="7340" to-port="0" /> <edge from-layer="7339" from-port="0" to-layer="7340" to-port="2" /> <edge from-layer="7340" from-port="3" to-layer="7341" to-port="0" /> <edge from-layer="7341" from-port="2" to-layer="7342" to-port="1" /> <edge from-layer="7341" from-port="2" to-layer="7374" to-port="1" /> <edge from-layer="7342" from-port="2" to-layer="7343" to-port="1" /> <edge from-layer="7343" from-port="2" to-layer="7348" to-port="0" /> <edge from-layer="7344" from-port="0" to-layer="7347" to-port="1" /> <edge from-layer="7345" from-port="0" to-layer="7347" to-port="2" /> <edge from-layer="7346" from-port="0" to-layer="7347" to-port="3" /> <edge from-layer="7347" from-port="4" to-layer="7348" to-port="1" /> <edge from-layer="7348" from-port="2" to-layer="7442" to-port="0" /> <edge from-layer="7348" from-port="2" to-layer="7404" to-port="0" /> <edge from-layer="7349" from-port="0" to-layer="7352" to-port="1" /> <edge from-layer="7350" from-port="0" to-layer="7352" to-port="2" /> <edge from-layer="7351" from-port="0" to-layer="7352" to-port="3" /> <edge from-layer="7352" from-port="4" to-layer="7353" to-port="0" /> <edge from-layer="7352" from-port="4" to-layer="7356" to-port="0" /> <edge from-layer="7352" from-port="4" to-layer="7367" to-port="0" /> <edge from-layer="7352" from-port="4" to-layer="7372" to-port="0" /> <edge from-layer="7353" from-port="2" to-layer="7375" to-port="0" /> <edge from-layer="7354" from-port="0" to-layer="7364" to-port="0" /> <edge from-layer="7355" from-port="0" to-layer="7364" to-port="1" /> <edge from-layer="7356" from-port="1" to-layer="7358" to-port="0" /> <edge from-layer="7357" from-port="0" to-layer="7358" to-port="2" /> <edge from-layer="7358" from-port="3" to-layer="7359" to-port="0" /> <edge from-layer="7359" from-port="2" to-layer="7360" to-port="0" /> <edge from-layer="7360" from-port="1" to-layer="7362" to-port="0" /> <edge from-layer="7361" from-port="0" to-layer="7362" to-port="1" /> <edge from-layer="7362" from-port="2" to-layer="7364" to-port="2" /> <edge from-layer="7363" from-port="0" to-layer="7364" to-port="3" /> <edge from-layer="7364" from-port="4" to-layer="7372" to-port="2" /> <edge from-layer="7364" from-port="4" to-layer="7367" to-port="1" /> <edge from-layer="7365" from-port="0" to-layer="7367" to-port="2" /> <edge from-layer="7366" from-port="0" to-layer="7367" to-port="3" /> <edge from-layer="7367" from-port="4" to-layer="7369" to-port="0" /> <edge from-layer="7368" from-port="0" to-layer="7369" to-port="1" /> <edge from-layer="7369" from-port="2" to-layer="7373" to-port="0" /> <edge from-layer="7370" from-port="0" to-layer="7372" to-port="1" /> <edge from-layer="7371" from-port="0" to-layer="7372" to-port="3" /> <edge from-layer="7372" from-port="4" to-layer="7373" to-port="1" /> <edge from-layer="7373" from-port="2" to-layer="7374" to-port="0" /> <edge from-layer="7374" from-port="2" to-layer="7375" to-port="1" /> <edge from-layer="7375" from-port="2" to-layer="7380" to-port="0" /> <edge from-layer="7376" from-port="0" to-layer="7379" to-port="1" /> <edge from-layer="7377" from-port="0" to-layer="7379" to-port="2" /> <edge from-layer="7378" from-port="0" to-layer="7379" to-port="3" /> <edge from-layer="7379" from-port="4" to-layer="7380" to-port="1" /> <edge from-layer="7380" from-port="2" to-layer="7381" to-port="1" /> <edge from-layer="7381" from-port="2" to-layer="10123" to-port="0" /> <edge from-layer="7381" from-port="2" to-layer="7442" to-port="1" /> <edge from-layer="7381" from-port="2" to-layer="7401" to-port="0" /> <edge from-layer="7382" from-port="0" to-layer="7387" to-port="0" /> <edge from-layer="7383" from-port="0" to-layer="7386" to-port="1" /> <edge from-layer="7384" from-port="0" to-layer="7386" to-port="2" /> <edge from-layer="7385" from-port="0" to-layer="7386" to-port="3" /> <edge from-layer="7386" from-port="4" to-layer="7387" to-port="1" /> <edge from-layer="7387" from-port="2" to-layer="7388" to-port="0" /> <edge from-layer="7388" from-port="1" to-layer="7390" to-port="0" /> <edge from-layer="7389" from-port="0" to-layer="7390" to-port="2" /> <edge from-layer="7390" from-port="3" to-layer="7397" to-port="0" /> <edge from-layer="7391" from-port="0" to-layer="7394" to-port="1" /> <edge from-layer="7392" from-port="0" to-layer="7394" to-port="2" /> <edge from-layer="7393" from-port="0" to-layer="7394" to-port="3" /> <edge from-layer="7394" from-port="4" to-layer="7396" to-port="0" /> <edge from-layer="7395" from-port="0" to-layer="7396" to-port="1" /> <edge from-layer="7396" from-port="2" to-layer="7397" to-port="1" /> <edge from-layer="7397" from-port="2" to-layer="10122" to-port="0" /> <edge from-layer="7397" from-port="2" to-layer="7442" to-port="2" /> <edge from-layer="7398" from-port="0" to-layer="7416" to-port="0" /> <edge from-layer="7399" from-port="0" to-layer="7411" to-port="0" /> <edge from-layer="7400" from-port="0" to-layer="7411" to-port="1" /> <edge from-layer="7400" from-port="0" to-layer="7414" to-port="1" /> <edge from-layer="7401" from-port="1" to-layer="7403" to-port="0" /> <edge from-layer="7402" from-port="0" to-layer="7403" to-port="2" /> <edge from-layer="7403" from-port="3" to-layer="7407" to-port="0" /> <edge from-layer="7403" from-port="3" to-layer="7413" to-port="0" /> <edge from-layer="7404" from-port="1" to-layer="7406" to-port="0" /> <edge from-layer="7404" from-port="1" to-layer="7431" to-port="0" /> <edge from-layer="7405" from-port="0" to-layer="7406" to-port="2" /> <edge from-layer="7406" from-port="3" to-layer="7407" to-port="1" /> <edge from-layer="7407" from-port="2" to-layer="7409" to-port="0" /> <edge from-layer="7408" from-port="0" to-layer="7413" to-port="1" /> <edge from-layer="7408" from-port="0" to-layer="7409" to-port="1" /> <edge from-layer="7409" from-port="2" to-layer="7411" to-port="2" /> <edge from-layer="7410" from-port="0" to-layer="7414" to-port="3" /> <edge from-layer="7410" from-port="0" to-layer="7411" to-port="3" /> <edge from-layer="7411" from-port="4" to-layer="7416" to-port="1" /> <edge from-layer="7412" from-port="0" to-layer="7414" to-port="0" /> <edge from-layer="7413" from-port="2" to-layer="7421" to-port="2" /> <edge from-layer="7413" from-port="2" to-layer="7414" to-port="2" /> <edge from-layer="7414" from-port="4" to-layer="7416" to-port="2" /> <edge from-layer="7415" from-port="0" to-layer="7416" to-port="3" /> <edge from-layer="7416" from-port="4" to-layer="7423" to-port="0" /> <edge from-layer="7417" from-port="0" to-layer="7423" to-port="1" /> <edge from-layer="7418" from-port="0" to-layer="7421" to-port="0" /> <edge from-layer="7419" from-port="0" to-layer="7421" to-port="1" /> <edge from-layer="7420" from-port="0" to-layer="7421" to-port="3" /> <edge from-layer="7421" from-port="4" to-layer="7423" to-port="2" /> <edge from-layer="7422" from-port="0" to-layer="7423" to-port="3" /> <edge from-layer="7423" from-port="4" to-layer="7426" to-port="0" /> <edge from-layer="7424" from-port="0" to-layer="7426" to-port="1" /> <edge from-layer="7425" from-port="0" to-layer="7426" to-port="2" /> <edge from-layer="7426" from-port="3" to-layer="7427" to-port="0" /> <edge from-layer="7427" from-port="2" to-layer="7428" to-port="0" /> <edge from-layer="7427" from-port="2" to-layer="7441" to-port="0" /> <edge from-layer="7428" from-port="1" to-layer="7440" to-port="0" /> <edge from-layer="7429" from-port="0" to-layer="7431" to-port="1" /> <edge from-layer="7430" from-port="0" to-layer="7431" to-port="2" /> <edge from-layer="7431" from-port="3" to-layer="7435" to-port="0" /> <edge from-layer="7432" from-port="0" to-layer="7435" to-port="1" /> <edge from-layer="7433" from-port="0" to-layer="7435" to-port="2" /> <edge from-layer="7434" from-port="0" to-layer="7435" to-port="3" /> <edge from-layer="7435" from-port="4" to-layer="7437" to-port="0" /> <edge from-layer="7435" from-port="4" to-layer="7439" to-port="2" /> <edge from-layer="7436" from-port="0" to-layer="7437" to-port="1" /> <edge from-layer="7437" from-port="2" to-layer="7439" to-port="0" /> <edge from-layer="7438" from-port="0" to-layer="7439" to-port="1" /> <edge from-layer="7439" from-port="3" to-layer="7440" to-port="1" /> <edge from-layer="7440" from-port="2" to-layer="7441" to-port="1" /> <edge from-layer="7441" from-port="2" to-layer="7442" to-port="3" /> <edge from-layer="7442" from-port="4" to-layer="7444" to-port="0" /> <edge from-layer="7443" from-port="0" to-layer="7444" to-port="1" /> <edge from-layer="7444" from-port="2" to-layer="7446" to-port="0" /> <edge from-layer="7445" from-port="0" to-layer="7446" to-port="1" /> <edge from-layer="7446" from-port="2" to-layer="7454" to-port="0" /> <edge from-layer="7447" from-port="0" to-layer="7448" to-port="0" /> <edge from-layer="7448" from-port="1" to-layer="7451" to-port="0" /> <edge from-layer="7449" from-port="0" to-layer="7450" to-port="0" /> <edge from-layer="7450" from-port="1" to-layer="7451" to-port="1" /> <edge from-layer="7451" from-port="2" to-layer="7453" to-port="0" /> <edge from-layer="7452" from-port="0" to-layer="7453" to-port="1" /> <edge from-layer="7453" from-port="2" to-layer="7454" to-port="1" /> <edge from-layer="7454" from-port="2" to-layer="7456" to-port="0" /> <edge from-layer="7455" from-port="0" to-layer="7456" to-port="1" /> <edge from-layer="7456" from-port="2" to-layer="7457" to-port="1" /> <edge from-layer="7457" from-port="2" to-layer="7458" to-port="0" /> <edge from-layer="7458" from-port="2" to-layer="7487" to-port="0" /> <edge from-layer="7458" from-port="2" to-layer="7460" to-port="0" /> <edge from-layer="7458" from-port="2" to-layer="7689" to-port="1" /> <edge from-layer="7459" from-port="0" to-layer="7460" to-port="1" /> <edge from-layer="7460" from-port="2" to-layer="7462" to-port="0" /> <edge from-layer="7461" from-port="0" to-layer="7462" to-port="1" /> <edge from-layer="7462" from-port="2" to-layer="7464" to-port="0" /> <edge from-layer="7463" from-port="0" to-layer="7464" to-port="1" /> <edge from-layer="7464" from-port="2" to-layer="7472" to-port="0" /> <edge from-layer="7465" from-port="0" to-layer="7466" to-port="0" /> <edge from-layer="7466" from-port="1" to-layer="7469" to-port="0" /> <edge from-layer="7467" from-port="0" to-layer="7468" to-port="0" /> <edge from-layer="7468" from-port="1" to-layer="7469" to-port="1" /> <edge from-layer="7469" from-port="2" to-layer="7471" to-port="0" /> <edge from-layer="7470" from-port="0" to-layer="7471" to-port="1" /> <edge from-layer="7471" from-port="2" to-layer="7472" to-port="1" /> <edge from-layer="7472" from-port="2" to-layer="7474" to-port="0" /> <edge from-layer="7473" from-port="0" to-layer="7474" to-port="1" /> <edge from-layer="7474" from-port="2" to-layer="7475" to-port="0" /> <edge from-layer="7475" from-port="1" to-layer="7483" to-port="0" /> <edge from-layer="7476" from-port="0" to-layer="7477" to-port="0" /> <edge from-layer="7477" from-port="1" to-layer="7480" to-port="0" /> <edge from-layer="7478" from-port="0" to-layer="7479" to-port="0" /> <edge from-layer="7479" from-port="1" to-layer="7480" to-port="1" /> <edge from-layer="7480" from-port="2" to-layer="7482" to-port="0" /> <edge from-layer="7481" from-port="0" to-layer="7482" to-port="1" /> <edge from-layer="7482" from-port="2" to-layer="7483" to-port="1" /> <edge from-layer="7483" from-port="2" to-layer="7485" to-port="0" /> <edge from-layer="7484" from-port="0" to-layer="7485" to-port="1" /> <edge from-layer="7485" from-port="2" to-layer="7688" to-port="0" /> <edge from-layer="7486" from-port="0" to-layer="7487" to-port="1" /> <edge from-layer="7487" from-port="2" to-layer="7489" to-port="0" /> <edge from-layer="7488" from-port="0" to-layer="7489" to-port="1" /> <edge from-layer="7489" from-port="2" to-layer="7491" to-port="0" /> <edge from-layer="7490" from-port="0" to-layer="7491" to-port="1" /> <edge from-layer="7491" from-port="2" to-layer="7499" to-port="0" /> <edge from-layer="7492" from-port="0" to-layer="7493" to-port="0" /> <edge from-layer="7493" from-port="1" to-layer="7496" to-port="0" /> <edge from-layer="7494" from-port="0" to-layer="7495" to-port="0" /> <edge from-layer="7495" from-port="1" to-layer="7496" to-port="1" /> <edge from-layer="7496" from-port="2" to-layer="7498" to-port="0" /> <edge from-layer="7497" from-port="0" to-layer="7498" to-port="1" /> <edge from-layer="7498" from-port="2" to-layer="7499" to-port="1" /> <edge from-layer="7499" from-port="2" to-layer="7501" to-port="0" /> <edge from-layer="7500" from-port="0" to-layer="7501" to-port="1" /> <edge from-layer="7501" from-port="2" to-layer="7503" to-port="0" /> <edge from-layer="7502" from-port="0" to-layer="7503" to-port="1" /> <edge from-layer="7503" from-port="2" to-layer="7625" to-port="0" /> <edge from-layer="7503" from-port="2" to-layer="7507" to-port="0" /> <edge from-layer="7503" from-port="2" to-layer="7519" to-port="0" /> <edge from-layer="7504" from-port="0" to-layer="7507" to-port="1" /> <edge from-layer="7505" from-port="0" to-layer="7507" to-port="2" /> <edge from-layer="7506" from-port="0" to-layer="7507" to-port="3" /> <edge from-layer="7507" from-port="4" to-layer="7509" to-port="0" /> <edge from-layer="7508" from-port="0" to-layer="7509" to-port="1" /> <edge from-layer="7509" from-port="2" to-layer="7513" to-port="0" /> <edge from-layer="7509" from-port="2" to-layer="7578" to-port="0" /> <edge from-layer="7510" from-port="0" to-layer="7513" to-port="1" /> <edge from-layer="7511" from-port="0" to-layer="7513" to-port="2" /> <edge from-layer="7512" from-port="0" to-layer="7513" to-port="3" /> <edge from-layer="7513" from-port="4" to-layer="7548" to-port="0" /> <edge from-layer="7513" from-port="4" to-layer="7559" to-port="0" /> <edge from-layer="7513" from-port="4" to-layer="7545" to-port="0" /> <edge from-layer="7513" from-port="4" to-layer="7564" to-port="0" /> <edge from-layer="7514" from-port="0" to-layer="7541" to-port="0" /> <edge from-layer="7515" from-port="0" to-layer="7541" to-port="1" /> <edge from-layer="7516" from-port="0" to-layer="7519" to-port="1" /> <edge from-layer="7517" from-port="0" to-layer="7519" to-port="2" /> <edge from-layer="7518" from-port="0" to-layer="7519" to-port="3" /> <edge from-layer="7519" from-port="4" to-layer="7521" to-port="0" /> <edge from-layer="7520" from-port="0" to-layer="7521" to-port="1" /> <edge from-layer="7521" from-port="2" to-layer="7610" to-port="0" /> <edge from-layer="7521" from-port="2" to-layer="7583" to-port="0" /> <edge from-layer="7521" from-port="2" to-layer="7522" to-port="0" /> <edge from-layer="7522" from-port="1" to-layer="7524" to-port="0" /> <edge from-layer="7523" from-port="0" to-layer="7524" to-port="2" /> <edge from-layer="7524" from-port="3" to-layer="7537" to-port="0" /> <edge from-layer="7525" from-port="0" to-layer="7530" to-port="0" /> <edge from-layer="7526" from-port="0" to-layer="7529" to-port="1" /> <edge from-layer="7527" from-port="0" to-layer="7529" to-port="2" /> <edge from-layer="7528" from-port="0" to-layer="7529" to-port="3" /> <edge from-layer="7529" from-port="4" to-layer="7530" to-port="1" /> <edge from-layer="7530" from-port="2" to-layer="7531" to-port="0" /> <edge from-layer="7531" from-port="1" to-layer="7533" to-port="0" /> <edge from-layer="7532" from-port="0" to-layer="7533" to-port="2" /> <edge from-layer="7533" from-port="3" to-layer="7534" to-port="0" /> <edge from-layer="7533" from-port="3" to-layer="7612" to-port="0" /> <edge from-layer="7534" from-port="1" to-layer="7536" to-port="0" /> <edge from-layer="7535" from-port="0" to-layer="7536" to-port="2" /> <edge from-layer="7536" from-port="3" to-layer="7537" to-port="1" /> <edge from-layer="7537" from-port="2" to-layer="7539" to-port="0" /> <edge from-layer="7538" from-port="0" to-layer="7539" to-port="1" /> <edge from-layer="7539" from-port="2" to-layer="7569" to-port="2" /> <edge from-layer="7539" from-port="2" to-layer="7541" to-port="2" /> <edge from-layer="7540" from-port="0" to-layer="7541" to-port="3" /> <edge from-layer="7541" from-port="4" to-layer="7543" to-port="0" /> <edge from-layer="7542" from-port="0" to-layer="7543" to-port="2" /> <edge from-layer="7543" from-port="3" to-layer="7544" to-port="0" /> <edge from-layer="7544" from-port="2" to-layer="7584" to-port="1" /> <edge from-layer="7544" from-port="2" to-layer="7545" to-port="1" /> <edge from-layer="7545" from-port="2" to-layer="7574" to-port="0" /> <edge from-layer="7546" from-port="0" to-layer="7556" to-port="0" /> <edge from-layer="7547" from-port="0" to-layer="7556" to-port="1" /> <edge from-layer="7548" from-port="1" to-layer="7550" to-port="0" /> <edge from-layer="7549" from-port="0" to-layer="7550" to-port="2" /> <edge from-layer="7550" from-port="3" to-layer="7551" to-port="0" /> <edge from-layer="7551" from-port="2" to-layer="7552" to-port="0" /> <edge from-layer="7552" from-port="1" to-layer="7554" to-port="0" /> <edge from-layer="7553" from-port="0" to-layer="7554" to-port="1" /> <edge from-layer="7554" from-port="2" to-layer="7556" to-port="2" /> <edge from-layer="7555" from-port="0" to-layer="7556" to-port="3" /> <edge from-layer="7556" from-port="4" to-layer="7564" to-port="2" /> <edge from-layer="7556" from-port="4" to-layer="7559" to-port="1" /> <edge from-layer="7557" from-port="0" to-layer="7559" to-port="2" /> <edge from-layer="7558" from-port="0" to-layer="7559" to-port="3" /> <edge from-layer="7559" from-port="4" to-layer="7561" to-port="0" /> <edge from-layer="7560" from-port="0" to-layer="7561" to-port="1" /> <edge from-layer="7561" from-port="2" to-layer="7565" to-port="0" /> <edge from-layer="7562" from-port="0" to-layer="7564" to-port="1" /> <edge from-layer="7563" from-port="0" to-layer="7564" to-port="3" /> <edge from-layer="7564" from-port="4" to-layer="7565" to-port="1" /> <edge from-layer="7565" from-port="2" to-layer="7573" to-port="0" /> <edge from-layer="7566" from-port="0" to-layer="7569" to-port="0" /> <edge from-layer="7567" from-port="0" to-layer="7569" to-port="1" /> <edge from-layer="7568" from-port="0" to-layer="7569" to-port="3" /> <edge from-layer="7569" from-port="4" to-layer="7571" to-port="0" /> <edge from-layer="7570" from-port="0" to-layer="7571" to-port="2" /> <edge from-layer="7571" from-port="3" to-layer="7572" to-port="0" /> <edge from-layer="7572" from-port="2" to-layer="7573" to-port="1" /> <edge from-layer="7572" from-port="2" to-layer="7605" to-port="1" /> <edge from-layer="7573" from-port="2" to-layer="7574" to-port="1" /> <edge from-layer="7574" from-port="2" to-layer="7579" to-port="0" /> <edge from-layer="7575" from-port="0" to-layer="7578" to-port="1" /> <edge from-layer="7576" from-port="0" to-layer="7578" to-port="2" /> <edge from-layer="7577" from-port="0" to-layer="7578" to-port="3" /> <edge from-layer="7578" from-port="4" to-layer="7579" to-port="1" /> <edge from-layer="7579" from-port="2" to-layer="7673" to-port="0" /> <edge from-layer="7579" from-port="2" to-layer="7635" to-port="0" /> <edge from-layer="7580" from-port="0" to-layer="7583" to-port="1" /> <edge from-layer="7581" from-port="0" to-layer="7583" to-port="2" /> <edge from-layer="7582" from-port="0" to-layer="7583" to-port="3" /> <edge from-layer="7583" from-port="4" to-layer="7603" to-port="0" /> <edge from-layer="7583" from-port="4" to-layer="7598" to-port="0" /> <edge from-layer="7583" from-port="4" to-layer="7584" to-port="0" /> <edge from-layer="7583" from-port="4" to-layer="7587" to-port="0" /> <edge from-layer="7584" from-port="2" to-layer="7606" to-port="0" /> <edge from-layer="7585" from-port="0" to-layer="7595" to-port="0" /> <edge from-layer="7586" from-port="0" to-layer="7595" to-port="1" /> <edge from-layer="7587" from-port="1" to-layer="7589" to-port="0" /> <edge from-layer="7588" from-port="0" to-layer="7589" to-port="2" /> <edge from-layer="7589" from-port="3" to-layer="7590" to-port="0" /> <edge from-layer="7590" from-port="2" to-layer="7591" to-port="0" /> <edge from-layer="7591" from-port="1" to-layer="7593" to-port="0" /> <edge from-layer="7592" from-port="0" to-layer="7593" to-port="1" /> <edge from-layer="7593" from-port="2" to-layer="7595" to-port="2" /> <edge from-layer="7594" from-port="0" to-layer="7595" to-port="3" /> <edge from-layer="7595" from-port="4" to-layer="7603" to-port="2" /> <edge from-layer="7595" from-port="4" to-layer="7598" to-port="1" /> <edge from-layer="7596" from-port="0" to-layer="7598" to-port="2" /> <edge from-layer="7597" from-port="0" to-layer="7598" to-port="3" /> <edge from-layer="7598" from-port="4" to-layer="7600" to-port="0" /> <edge from-layer="7599" from-port="0" to-layer="7600" to-port="1" /> <edge from-layer="7600" from-port="2" to-layer="7604" to-port="0" /> <edge from-layer="7601" from-port="0" to-layer="7603" to-port="1" /> <edge from-layer="7602" from-port="0" to-layer="7603" to-port="3" /> <edge from-layer="7603" from-port="4" to-layer="7604" to-port="1" /> <edge from-layer="7604" from-port="2" to-layer="7605" to-port="0" /> <edge from-layer="7605" from-port="2" to-layer="7606" to-port="1" /> <edge from-layer="7606" from-port="2" to-layer="7611" to-port="0" /> <edge from-layer="7607" from-port="0" to-layer="7610" to-port="1" /> <edge from-layer="7608" from-port="0" to-layer="7610" to-port="2" /> <edge from-layer="7609" from-port="0" to-layer="7610" to-port="3" /> <edge from-layer="7610" from-port="4" to-layer="7611" to-port="1" /> <edge from-layer="7611" from-port="2" to-layer="7612" to-port="1" /> <edge from-layer="7612" from-port="2" to-layer="10121" to-port="0" /> <edge from-layer="7612" from-port="2" to-layer="7673" to-port="1" /> <edge from-layer="7612" from-port="2" to-layer="7632" to-port="0" /> <edge from-layer="7613" from-port="0" to-layer="7618" to-port="0" /> <edge from-layer="7614" from-port="0" to-layer="7617" to-port="1" /> <edge from-layer="7615" from-port="0" to-layer="7617" to-port="2" /> <edge from-layer="7616" from-port="0" to-layer="7617" to-port="3" /> <edge from-layer="7617" from-port="4" to-layer="7618" to-port="1" /> <edge from-layer="7618" from-port="2" to-layer="7619" to-port="0" /> <edge from-layer="7619" from-port="1" to-layer="7621" to-port="0" /> <edge from-layer="7620" from-port="0" to-layer="7621" to-port="2" /> <edge from-layer="7621" from-port="3" to-layer="7628" to-port="0" /> <edge from-layer="7622" from-port="0" to-layer="7625" to-port="1" /> <edge from-layer="7623" from-port="0" to-layer="7625" to-port="2" /> <edge from-layer="7624" from-port="0" to-layer="7625" to-port="3" /> <edge from-layer="7625" from-port="4" to-layer="7627" to-port="0" /> <edge from-layer="7626" from-port="0" to-layer="7627" to-port="1" /> <edge from-layer="7627" from-port="2" to-layer="7628" to-port="1" /> <edge from-layer="7628" from-port="2" to-layer="10120" to-port="0" /> <edge from-layer="7628" from-port="2" to-layer="7673" to-port="2" /> <edge from-layer="7629" from-port="0" to-layer="7647" to-port="0" /> <edge from-layer="7630" from-port="0" to-layer="7642" to-port="0" /> <edge from-layer="7631" from-port="0" to-layer="7645" to-port="1" /> <edge from-layer="7631" from-port="0" to-layer="7642" to-port="1" /> <edge from-layer="7632" from-port="1" to-layer="7634" to-port="0" /> <edge from-layer="7633" from-port="0" to-layer="7634" to-port="2" /> <edge from-layer="7634" from-port="3" to-layer="7638" to-port="0" /> <edge from-layer="7634" from-port="3" to-layer="7644" to-port="0" /> <edge from-layer="7635" from-port="1" to-layer="7662" to-port="0" /> <edge from-layer="7635" from-port="1" to-layer="7637" to-port="0" /> <edge from-layer="7636" from-port="0" to-layer="7637" to-port="2" /> <edge from-layer="7637" from-port="3" to-layer="7638" to-port="1" /> <edge from-layer="7638" from-port="2" to-layer="7640" to-port="0" /> <edge from-layer="7639" from-port="0" to-layer="7644" to-port="1" /> <edge from-layer="7639" from-port="0" to-layer="7640" to-port="1" /> <edge from-layer="7640" from-port="2" to-layer="7642" to-port="2" /> <edge from-layer="7641" from-port="0" to-layer="7645" to-port="3" /> <edge from-layer="7641" from-port="0" to-layer="7642" to-port="3" /> <edge from-layer="7642" from-port="4" to-layer="7647" to-port="1" /> <edge from-layer="7643" from-port="0" to-layer="7645" to-port="0" /> <edge from-layer="7644" from-port="2" to-layer="7652" to-port="2" /> <edge from-layer="7644" from-port="2" to-layer="7645" to-port="2" /> <edge from-layer="7645" from-port="4" to-layer="7647" to-port="2" /> <edge from-layer="7646" from-port="0" to-layer="7647" to-port="3" /> <edge from-layer="7647" from-port="4" to-layer="7654" to-port="0" /> <edge from-layer="7648" from-port="0" to-layer="7654" to-port="1" /> <edge from-layer="7649" from-port="0" to-layer="7652" to-port="0" /> <edge from-layer="7650" from-port="0" to-layer="7652" to-port="1" /> <edge from-layer="7651" from-port="0" to-layer="7652" to-port="3" /> <edge from-layer="7652" from-port="4" to-layer="7654" to-port="2" /> <edge from-layer="7653" from-port="0" to-layer="7654" to-port="3" /> <edge from-layer="7654" from-port="4" to-layer="7657" to-port="0" /> <edge from-layer="7655" from-port="0" to-layer="7657" to-port="1" /> <edge from-layer="7656" from-port="0" to-layer="7657" to-port="2" /> <edge from-layer="7657" from-port="3" to-layer="7658" to-port="0" /> <edge from-layer="7658" from-port="2" to-layer="7672" to-port="0" /> <edge from-layer="7658" from-port="2" to-layer="7659" to-port="0" /> <edge from-layer="7659" from-port="1" to-layer="7671" to-port="0" /> <edge from-layer="7660" from-port="0" to-layer="7662" to-port="1" /> <edge from-layer="7661" from-port="0" to-layer="7662" to-port="2" /> <edge from-layer="7662" from-port="3" to-layer="7666" to-port="0" /> <edge from-layer="7663" from-port="0" to-layer="7666" to-port="1" /> <edge from-layer="7664" from-port="0" to-layer="7666" to-port="2" /> <edge from-layer="7665" from-port="0" to-layer="7666" to-port="3" /> <edge from-layer="7666" from-port="4" to-layer="7668" to-port="0" /> <edge from-layer="7666" from-port="4" to-layer="7670" to-port="2" /> <edge from-layer="7667" from-port="0" to-layer="7668" to-port="1" /> <edge from-layer="7668" from-port="2" to-layer="7670" to-port="0" /> <edge from-layer="7669" from-port="0" to-layer="7670" to-port="1" /> <edge from-layer="7670" from-port="3" to-layer="7671" to-port="1" /> <edge from-layer="7671" from-port="2" to-layer="7672" to-port="1" /> <edge from-layer="7672" from-port="2" to-layer="7673" to-port="3" /> <edge from-layer="7673" from-port="4" to-layer="7675" to-port="0" /> <edge from-layer="7674" from-port="0" to-layer="7675" to-port="1" /> <edge from-layer="7675" from-port="2" to-layer="7677" to-port="0" /> <edge from-layer="7676" from-port="0" to-layer="7677" to-port="1" /> <edge from-layer="7677" from-port="2" to-layer="7685" to-port="0" /> <edge from-layer="7678" from-port="0" to-layer="7679" to-port="0" /> <edge from-layer="7679" from-port="1" to-layer="7682" to-port="0" /> <edge from-layer="7680" from-port="0" to-layer="7681" to-port="0" /> <edge from-layer="7681" from-port="1" to-layer="7682" to-port="1" /> <edge from-layer="7682" from-port="2" to-layer="7684" to-port="0" /> <edge from-layer="7683" from-port="0" to-layer="7684" to-port="1" /> <edge from-layer="7684" from-port="2" to-layer="7685" to-port="1" /> <edge from-layer="7685" from-port="2" to-layer="7687" to-port="0" /> <edge from-layer="7686" from-port="0" to-layer="7687" to-port="1" /> <edge from-layer="7687" from-port="2" to-layer="7688" to-port="1" /> <edge from-layer="7688" from-port="2" to-layer="7689" to-port="0" /> <edge from-layer="7689" from-port="2" to-layer="7691" to-port="0" /> <edge from-layer="7689" from-port="2" to-layer="7920" to-port="1" /> <edge from-layer="7689" from-port="2" to-layer="7718" to-port="0" /> <edge from-layer="7690" from-port="0" to-layer="7691" to-port="1" /> <edge from-layer="7691" from-port="2" to-layer="7693" to-port="0" /> <edge from-layer="7692" from-port="0" to-layer="7693" to-port="1" /> <edge from-layer="7693" from-port="2" to-layer="7695" to-port="0" /> <edge from-layer="7694" from-port="0" to-layer="7695" to-port="1" /> <edge from-layer="7695" from-port="2" to-layer="7703" to-port="0" /> <edge from-layer="7696" from-port="0" to-layer="7697" to-port="0" /> <edge from-layer="7697" from-port="1" to-layer="7700" to-port="0" /> <edge from-layer="7698" from-port="0" to-layer="7699" to-port="0" /> <edge from-layer="7699" from-port="1" to-layer="7700" to-port="1" /> <edge from-layer="7700" from-port="2" to-layer="7702" to-port="0" /> <edge from-layer="7701" from-port="0" to-layer="7702" to-port="1" /> <edge from-layer="7702" from-port="2" to-layer="7703" to-port="1" /> <edge from-layer="7703" from-port="2" to-layer="7705" to-port="0" /> <edge from-layer="7704" from-port="0" to-layer="7705" to-port="1" /> <edge from-layer="7705" from-port="2" to-layer="7706" to-port="0" /> <edge from-layer="7706" from-port="1" to-layer="7714" to-port="0" /> <edge from-layer="7707" from-port="0" to-layer="7708" to-port="0" /> <edge from-layer="7708" from-port="1" to-layer="7711" to-port="0" /> <edge from-layer="7709" from-port="0" to-layer="7710" to-port="0" /> <edge from-layer="7710" from-port="1" to-layer="7711" to-port="1" /> <edge from-layer="7711" from-port="2" to-layer="7713" to-port="0" /> <edge from-layer="7712" from-port="0" to-layer="7713" to-port="1" /> <edge from-layer="7713" from-port="2" to-layer="7714" to-port="1" /> <edge from-layer="7714" from-port="2" to-layer="7716" to-port="0" /> <edge from-layer="7715" from-port="0" to-layer="7716" to-port="1" /> <edge from-layer="7716" from-port="2" to-layer="7919" to-port="0" /> <edge from-layer="7717" from-port="0" to-layer="7718" to-port="1" /> <edge from-layer="7718" from-port="2" to-layer="7720" to-port="0" /> <edge from-layer="7719" from-port="0" to-layer="7720" to-port="1" /> <edge from-layer="7720" from-port="2" to-layer="7722" to-port="0" /> <edge from-layer="7721" from-port="0" to-layer="7722" to-port="1" /> <edge from-layer="7722" from-port="2" to-layer="7730" to-port="0" /> <edge from-layer="7723" from-port="0" to-layer="7724" to-port="0" /> <edge from-layer="7724" from-port="1" to-layer="7727" to-port="0" /> <edge from-layer="7725" from-port="0" to-layer="7726" to-port="0" /> <edge from-layer="7726" from-port="1" to-layer="7727" to-port="1" /> <edge from-layer="7727" from-port="2" to-layer="7729" to-port="0" /> <edge from-layer="7728" from-port="0" to-layer="7729" to-port="1" /> <edge from-layer="7729" from-port="2" to-layer="7730" to-port="1" /> <edge from-layer="7730" from-port="2" to-layer="7732" to-port="0" /> <edge from-layer="7731" from-port="0" to-layer="7732" to-port="1" /> <edge from-layer="7732" from-port="2" to-layer="7734" to-port="0" /> <edge from-layer="7733" from-port="0" to-layer="7734" to-port="1" /> <edge from-layer="7734" from-port="2" to-layer="7750" to-port="0" /> <edge from-layer="7734" from-port="2" to-layer="7856" to-port="0" /> <edge from-layer="7734" from-port="2" to-layer="7738" to-port="0" /> <edge from-layer="7735" from-port="0" to-layer="7738" to-port="1" /> <edge from-layer="7736" from-port="0" to-layer="7738" to-port="2" /> <edge from-layer="7737" from-port="0" to-layer="7738" to-port="3" /> <edge from-layer="7738" from-port="4" to-layer="7740" to-port="0" /> <edge from-layer="7739" from-port="0" to-layer="7740" to-port="1" /> <edge from-layer="7740" from-port="2" to-layer="7809" to-port="0" /> <edge from-layer="7740" from-port="2" to-layer="7744" to-port="0" /> <edge from-layer="7741" from-port="0" to-layer="7744" to-port="1" /> <edge from-layer="7742" from-port="0" to-layer="7744" to-port="2" /> <edge from-layer="7743" from-port="0" to-layer="7744" to-port="3" /> <edge from-layer="7744" from-port="4" to-layer="7795" to-port="0" /> <edge from-layer="7744" from-port="4" to-layer="7790" to-port="0" /> <edge from-layer="7744" from-port="4" to-layer="7779" to-port="0" /> <edge from-layer="7744" from-port="4" to-layer="7776" to-port="0" /> <edge from-layer="7745" from-port="0" to-layer="7772" to-port="0" /> <edge from-layer="7746" from-port="0" to-layer="7772" to-port="1" /> <edge from-layer="7747" from-port="0" to-layer="7750" to-port="1" /> <edge from-layer="7748" from-port="0" to-layer="7750" to-port="2" /> <edge from-layer="7749" from-port="0" to-layer="7750" to-port="3" /> <edge from-layer="7750" from-port="4" to-layer="7752" to-port="0" /> <edge from-layer="7751" from-port="0" to-layer="7752" to-port="1" /> <edge from-layer="7752" from-port="2" to-layer="7753" to-port="0" /> <edge from-layer="7752" from-port="2" to-layer="7814" to-port="0" /> <edge from-layer="7752" from-port="2" to-layer="7841" to-port="0" /> <edge from-layer="7753" from-port="1" to-layer="7755" to-port="0" /> <edge from-layer="7754" from-port="0" to-layer="7755" to-port="2" /> <edge from-layer="7755" from-port="3" to-layer="7768" to-port="0" /> <edge from-layer="7756" from-port="0" to-layer="7761" to-port="0" /> <edge from-layer="7757" from-port="0" to-layer="7760" to-port="1" /> <edge from-layer="7758" from-port="0" to-layer="7760" to-port="2" /> <edge from-layer="7759" from-port="0" to-layer="7760" to-port="3" /> <edge from-layer="7760" from-port="4" to-layer="7761" to-port="1" /> <edge from-layer="7761" from-port="2" to-layer="7762" to-port="0" /> <edge from-layer="7762" from-port="1" to-layer="7764" to-port="0" /> <edge from-layer="7763" from-port="0" to-layer="7764" to-port="2" /> <edge from-layer="7764" from-port="3" to-layer="7765" to-port="0" /> <edge from-layer="7764" from-port="3" to-layer="7843" to-port="0" /> <edge from-layer="7765" from-port="1" to-layer="7767" to-port="0" /> <edge from-layer="7766" from-port="0" to-layer="7767" to-port="2" /> <edge from-layer="7767" from-port="3" to-layer="7768" to-port="1" /> <edge from-layer="7768" from-port="2" to-layer="7770" to-port="0" /> <edge from-layer="7769" from-port="0" to-layer="7770" to-port="1" /> <edge from-layer="7770" from-port="2" to-layer="7800" to-port="2" /> <edge from-layer="7770" from-port="2" to-layer="7772" to-port="2" /> <edge from-layer="7771" from-port="0" to-layer="7772" to-port="3" /> <edge from-layer="7772" from-port="4" to-layer="7774" to-port="0" /> <edge from-layer="7773" from-port="0" to-layer="7774" to-port="2" /> <edge from-layer="7774" from-port="3" to-layer="7775" to-port="0" /> <edge from-layer="7775" from-port="2" to-layer="7815" to-port="1" /> <edge from-layer="7775" from-port="2" to-layer="7776" to-port="1" /> <edge from-layer="7776" from-port="2" to-layer="7805" to-port="0" /> <edge from-layer="7777" from-port="0" to-layer="7787" to-port="0" /> <edge from-layer="7778" from-port="0" to-layer="7787" to-port="1" /> <edge from-layer="7779" from-port="1" to-layer="7781" to-port="0" /> <edge from-layer="7780" from-port="0" to-layer="7781" to-port="2" /> <edge from-layer="7781" from-port="3" to-layer="7782" to-port="0" /> <edge from-layer="7782" from-port="2" to-layer="7783" to-port="0" /> <edge from-layer="7783" from-port="1" to-layer="7785" to-port="0" /> <edge from-layer="7784" from-port="0" to-layer="7785" to-port="1" /> <edge from-layer="7785" from-port="2" to-layer="7787" to-port="2" /> <edge from-layer="7786" from-port="0" to-layer="7787" to-port="3" /> <edge from-layer="7787" from-port="4" to-layer="7790" to-port="1" /> <edge from-layer="7787" from-port="4" to-layer="7795" to-port="2" /> <edge from-layer="7788" from-port="0" to-layer="7790" to-port="2" /> <edge from-layer="7789" from-port="0" to-layer="7790" to-port="3" /> <edge from-layer="7790" from-port="4" to-layer="7792" to-port="0" /> <edge from-layer="7791" from-port="0" to-layer="7792" to-port="1" /> <edge from-layer="7792" from-port="2" to-layer="7796" to-port="0" /> <edge from-layer="7793" from-port="0" to-layer="7795" to-port="1" /> <edge from-layer="7794" from-port="0" to-layer="7795" to-port="3" /> <edge from-layer="7795" from-port="4" to-layer="7796" to-port="1" /> <edge from-layer="7796" from-port="2" to-layer="7804" to-port="0" /> <edge from-layer="7797" from-port="0" to-layer="7800" to-port="0" /> <edge from-layer="7798" from-port="0" to-layer="7800" to-port="1" /> <edge from-layer="7799" from-port="0" to-layer="7800" to-port="3" /> <edge from-layer="7800" from-port="4" to-layer="7802" to-port="0" /> <edge from-layer="7801" from-port="0" to-layer="7802" to-port="2" /> <edge from-layer="7802" from-port="3" to-layer="7803" to-port="0" /> <edge from-layer="7803" from-port="2" to-layer="7836" to-port="1" /> <edge from-layer="7803" from-port="2" to-layer="7804" to-port="1" /> <edge from-layer="7804" from-port="2" to-layer="7805" to-port="1" /> <edge from-layer="7805" from-port="2" to-layer="7810" to-port="0" /> <edge from-layer="7806" from-port="0" to-layer="7809" to-port="1" /> <edge from-layer="7807" from-port="0" to-layer="7809" to-port="2" /> <edge from-layer="7808" from-port="0" to-layer="7809" to-port="3" /> <edge from-layer="7809" from-port="4" to-layer="7810" to-port="1" /> <edge from-layer="7810" from-port="2" to-layer="7866" to-port="0" /> <edge from-layer="7810" from-port="2" to-layer="7904" to-port="0" /> <edge from-layer="7811" from-port="0" to-layer="7814" to-port="1" /> <edge from-layer="7812" from-port="0" to-layer="7814" to-port="2" /> <edge from-layer="7813" from-port="0" to-layer="7814" to-port="3" /> <edge from-layer="7814" from-port="4" to-layer="7815" to-port="0" /> <edge from-layer="7814" from-port="4" to-layer="7818" to-port="0" /> <edge from-layer="7814" from-port="4" to-layer="7834" to-port="0" /> <edge from-layer="7814" from-port="4" to-layer="7829" to-port="0" /> <edge from-layer="7815" from-port="2" to-layer="7837" to-port="0" /> <edge from-layer="7816" from-port="0" to-layer="7826" to-port="0" /> <edge from-layer="7817" from-port="0" to-layer="7826" to-port="1" /> <edge from-layer="7818" from-port="1" to-layer="7820" to-port="0" /> <edge from-layer="7819" from-port="0" to-layer="7820" to-port="2" /> <edge from-layer="7820" from-port="3" to-layer="7821" to-port="0" /> <edge from-layer="7821" from-port="2" to-layer="7822" to-port="0" /> <edge from-layer="7822" from-port="1" to-layer="7824" to-port="0" /> <edge from-layer="7823" from-port="0" to-layer="7824" to-port="1" /> <edge from-layer="7824" from-port="2" to-layer="7826" to-port="2" /> <edge from-layer="7825" from-port="0" to-layer="7826" to-port="3" /> <edge from-layer="7826" from-port="4" to-layer="7829" to-port="1" /> <edge from-layer="7826" from-port="4" to-layer="7834" to-port="2" /> <edge from-layer="7827" from-port="0" to-layer="7829" to-port="2" /> <edge from-layer="7828" from-port="0" to-layer="7829" to-port="3" /> <edge from-layer="7829" from-port="4" to-layer="7831" to-port="0" /> <edge from-layer="7830" from-port="0" to-layer="7831" to-port="1" /> <edge from-layer="7831" from-port="2" to-layer="7835" to-port="0" /> <edge from-layer="7832" from-port="0" to-layer="7834" to-port="1" /> <edge from-layer="7833" from-port="0" to-layer="7834" to-port="3" /> <edge from-layer="7834" from-port="4" to-layer="7835" to-port="1" /> <edge from-layer="7835" from-port="2" to-layer="7836" to-port="0" /> <edge from-layer="7836" from-port="2" to-layer="7837" to-port="1" /> <edge from-layer="7837" from-port="2" to-layer="7842" to-port="0" /> <edge from-layer="7838" from-port="0" to-layer="7841" to-port="1" /> <edge from-layer="7839" from-port="0" to-layer="7841" to-port="2" /> <edge from-layer="7840" from-port="0" to-layer="7841" to-port="3" /> <edge from-layer="7841" from-port="4" to-layer="7842" to-port="1" /> <edge from-layer="7842" from-port="2" to-layer="7843" to-port="1" /> <edge from-layer="7843" from-port="2" to-layer="10119" to-port="0" /> <edge from-layer="7843" from-port="2" to-layer="7904" to-port="1" /> <edge from-layer="7843" from-port="2" to-layer="7863" to-port="0" /> <edge from-layer="7844" from-port="0" to-layer="7849" to-port="0" /> <edge from-layer="7845" from-port="0" to-layer="7848" to-port="1" /> <edge from-layer="7846" from-port="0" to-layer="7848" to-port="2" /> <edge from-layer="7847" from-port="0" to-layer="7848" to-port="3" /> <edge from-layer="7848" from-port="4" to-layer="7849" to-port="1" /> <edge from-layer="7849" from-port="2" to-layer="7850" to-port="0" /> <edge from-layer="7850" from-port="1" to-layer="7852" to-port="0" /> <edge from-layer="7851" from-port="0" to-layer="7852" to-port="2" /> <edge from-layer="7852" from-port="3" to-layer="7859" to-port="0" /> <edge from-layer="7853" from-port="0" to-layer="7856" to-port="1" /> <edge from-layer="7854" from-port="0" to-layer="7856" to-port="2" /> <edge from-layer="7855" from-port="0" to-layer="7856" to-port="3" /> <edge from-layer="7856" from-port="4" to-layer="7858" to-port="0" /> <edge from-layer="7857" from-port="0" to-layer="7858" to-port="1" /> <edge from-layer="7858" from-port="2" to-layer="7859" to-port="1" /> <edge from-layer="7859" from-port="2" to-layer="7904" to-port="2" /> <edge from-layer="7859" from-port="2" to-layer="10118" to-port="0" /> <edge from-layer="7860" from-port="0" to-layer="7878" to-port="0" /> <edge from-layer="7861" from-port="0" to-layer="7873" to-port="0" /> <edge from-layer="7862" from-port="0" to-layer="7876" to-port="1" /> <edge from-layer="7862" from-port="0" to-layer="7873" to-port="1" /> <edge from-layer="7863" from-port="1" to-layer="7865" to-port="0" /> <edge from-layer="7864" from-port="0" to-layer="7865" to-port="2" /> <edge from-layer="7865" from-port="3" to-layer="7869" to-port="0" /> <edge from-layer="7865" from-port="3" to-layer="7875" to-port="0" /> <edge from-layer="7866" from-port="1" to-layer="7868" to-port="0" /> <edge from-layer="7866" from-port="1" to-layer="7893" to-port="0" /> <edge from-layer="7867" from-port="0" to-layer="7868" to-port="2" /> <edge from-layer="7868" from-port="3" to-layer="7869" to-port="1" /> <edge from-layer="7869" from-port="2" to-layer="7871" to-port="0" /> <edge from-layer="7870" from-port="0" to-layer="7871" to-port="1" /> <edge from-layer="7870" from-port="0" to-layer="7875" to-port="1" /> <edge from-layer="7871" from-port="2" to-layer="7873" to-port="2" /> <edge from-layer="7872" from-port="0" to-layer="7876" to-port="3" /> <edge from-layer="7872" from-port="0" to-layer="7873" to-port="3" /> <edge from-layer="7873" from-port="4" to-layer="7878" to-port="1" /> <edge from-layer="7874" from-port="0" to-layer="7876" to-port="0" /> <edge from-layer="7875" from-port="2" to-layer="7876" to-port="2" /> <edge from-layer="7875" from-port="2" to-layer="7883" to-port="2" /> <edge from-layer="7876" from-port="4" to-layer="7878" to-port="2" /> <edge from-layer="7877" from-port="0" to-layer="7878" to-port="3" /> <edge from-layer="7878" from-port="4" to-layer="7885" to-port="0" /> <edge from-layer="7879" from-port="0" to-layer="7885" to-port="1" /> <edge from-layer="7880" from-port="0" to-layer="7883" to-port="0" /> <edge from-layer="7881" from-port="0" to-layer="7883" to-port="1" /> <edge from-layer="7882" from-port="0" to-layer="7883" to-port="3" /> <edge from-layer="7883" from-port="4" to-layer="7885" to-port="2" /> <edge from-layer="7884" from-port="0" to-layer="7885" to-port="3" /> <edge from-layer="7885" from-port="4" to-layer="7888" to-port="0" /> <edge from-layer="7886" from-port="0" to-layer="7888" to-port="1" /> <edge from-layer="7887" from-port="0" to-layer="7888" to-port="2" /> <edge from-layer="7888" from-port="3" to-layer="7889" to-port="0" /> <edge from-layer="7889" from-port="2" to-layer="7903" to-port="0" /> <edge from-layer="7889" from-port="2" to-layer="7890" to-port="0" /> <edge from-layer="7890" from-port="1" to-layer="7902" to-port="0" /> <edge from-layer="7891" from-port="0" to-layer="7893" to-port="1" /> <edge from-layer="7892" from-port="0" to-layer="7893" to-port="2" /> <edge from-layer="7893" from-port="3" to-layer="7897" to-port="0" /> <edge from-layer="7894" from-port="0" to-layer="7897" to-port="1" /> <edge from-layer="7895" from-port="0" to-layer="7897" to-port="2" /> <edge from-layer="7896" from-port="0" to-layer="7897" to-port="3" /> <edge from-layer="7897" from-port="4" to-layer="7899" to-port="0" /> <edge from-layer="7897" from-port="4" to-layer="7901" to-port="2" /> <edge from-layer="7898" from-port="0" to-layer="7899" to-port="1" /> <edge from-layer="7899" from-port="2" to-layer="7901" to-port="0" /> <edge from-layer="7900" from-port="0" to-layer="7901" to-port="1" /> <edge from-layer="7901" from-port="3" to-layer="7902" to-port="1" /> <edge from-layer="7902" from-port="2" to-layer="7903" to-port="1" /> <edge from-layer="7903" from-port="2" to-layer="7904" to-port="3" /> <edge from-layer="7904" from-port="4" to-layer="7906" to-port="0" /> <edge from-layer="7905" from-port="0" to-layer="7906" to-port="1" /> <edge from-layer="7906" from-port="2" to-layer="7908" to-port="0" /> <edge from-layer="7907" from-port="0" to-layer="7908" to-port="1" /> <edge from-layer="7908" from-port="2" to-layer="7916" to-port="0" /> <edge from-layer="7909" from-port="0" to-layer="7910" to-port="0" /> <edge from-layer="7910" from-port="1" to-layer="7913" to-port="0" /> <edge from-layer="7911" from-port="0" to-layer="7912" to-port="0" /> <edge from-layer="7912" from-port="1" to-layer="7913" to-port="1" /> <edge from-layer="7913" from-port="2" to-layer="7915" to-port="0" /> <edge from-layer="7914" from-port="0" to-layer="7915" to-port="1" /> <edge from-layer="7915" from-port="2" to-layer="7916" to-port="1" /> <edge from-layer="7916" from-port="2" to-layer="7918" to-port="0" /> <edge from-layer="7917" from-port="0" to-layer="7918" to-port="1" /> <edge from-layer="7918" from-port="2" to-layer="7919" to-port="1" /> <edge from-layer="7919" from-port="2" to-layer="7920" to-port="0" /> <edge from-layer="7920" from-port="2" to-layer="7922" to-port="0" /> <edge from-layer="7920" from-port="2" to-layer="8151" to-port="1" /> <edge from-layer="7920" from-port="2" to-layer="7949" to-port="0" /> <edge from-layer="7921" from-port="0" to-layer="7922" to-port="1" /> <edge from-layer="7922" from-port="2" to-layer="7924" to-port="0" /> <edge from-layer="7923" from-port="0" to-layer="7924" to-port="1" /> <edge from-layer="7924" from-port="2" to-layer="7926" to-port="0" /> <edge from-layer="7925" from-port="0" to-layer="7926" to-port="1" /> <edge from-layer="7926" from-port="2" to-layer="7934" to-port="0" /> <edge from-layer="7927" from-port="0" to-layer="7928" to-port="0" /> <edge from-layer="7928" from-port="1" to-layer="7931" to-port="0" /> <edge from-layer="7929" from-port="0" to-layer="7930" to-port="0" /> <edge from-layer="7930" from-port="1" to-layer="7931" to-port="1" /> <edge from-layer="7931" from-port="2" to-layer="7933" to-port="0" /> <edge from-layer="7932" from-port="0" to-layer="7933" to-port="1" /> <edge from-layer="7933" from-port="2" to-layer="7934" to-port="1" /> <edge from-layer="7934" from-port="2" to-layer="7936" to-port="0" /> <edge from-layer="7935" from-port="0" to-layer="7936" to-port="1" /> <edge from-layer="7936" from-port="2" to-layer="7937" to-port="0" /> <edge from-layer="7937" from-port="1" to-layer="7945" to-port="0" /> <edge from-layer="7938" from-port="0" to-layer="7939" to-port="0" /> <edge from-layer="7939" from-port="1" to-layer="7942" to-port="0" /> <edge from-layer="7940" from-port="0" to-layer="7941" to-port="0" /> <edge from-layer="7941" from-port="1" to-layer="7942" to-port="1" /> <edge from-layer="7942" from-port="2" to-layer="7944" to-port="0" /> <edge from-layer="7943" from-port="0" to-layer="7944" to-port="1" /> <edge from-layer="7944" from-port="2" to-layer="7945" to-port="1" /> <edge from-layer="7945" from-port="2" to-layer="7947" to-port="0" /> <edge from-layer="7946" from-port="0" to-layer="7947" to-port="1" /> <edge from-layer="7947" from-port="2" to-layer="8150" to-port="0" /> <edge from-layer="7948" from-port="0" to-layer="7949" to-port="1" /> <edge from-layer="7949" from-port="2" to-layer="7951" to-port="0" /> <edge from-layer="7950" from-port="0" to-layer="7951" to-port="1" /> <edge from-layer="7951" from-port="2" to-layer="7953" to-port="0" /> <edge from-layer="7952" from-port="0" to-layer="7953" to-port="1" /> <edge from-layer="7953" from-port="2" to-layer="7961" to-port="0" /> <edge from-layer="7954" from-port="0" to-layer="7955" to-port="0" /> <edge from-layer="7955" from-port="1" to-layer="7958" to-port="0" /> <edge from-layer="7956" from-port="0" to-layer="7957" to-port="0" /> <edge from-layer="7957" from-port="1" to-layer="7958" to-port="1" /> <edge from-layer="7958" from-port="2" to-layer="7960" to-port="0" /> <edge from-layer="7959" from-port="0" to-layer="7960" to-port="1" /> <edge from-layer="7960" from-port="2" to-layer="7961" to-port="1" /> <edge from-layer="7961" from-port="2" to-layer="7963" to-port="0" /> <edge from-layer="7962" from-port="0" to-layer="7963" to-port="1" /> <edge from-layer="7963" from-port="2" to-layer="7965" to-port="0" /> <edge from-layer="7964" from-port="0" to-layer="7965" to-port="1" /> <edge from-layer="7965" from-port="2" to-layer="7969" to-port="0" /> <edge from-layer="7965" from-port="2" to-layer="7981" to-port="0" /> <edge from-layer="7965" from-port="2" to-layer="8087" to-port="0" /> <edge from-layer="7966" from-port="0" to-layer="7969" to-port="1" /> <edge from-layer="7967" from-port="0" to-layer="7969" to-port="2" /> <edge from-layer="7968" from-port="0" to-layer="7969" to-port="3" /> <edge from-layer="7969" from-port="4" to-layer="7971" to-port="0" /> <edge from-layer="7970" from-port="0" to-layer="7971" to-port="1" /> <edge from-layer="7971" from-port="2" to-layer="7975" to-port="0" /> <edge from-layer="7971" from-port="2" to-layer="8040" to-port="0" /> <edge from-layer="7972" from-port="0" to-layer="7975" to-port="1" /> <edge from-layer="7973" from-port="0" to-layer="7975" to-port="2" /> <edge from-layer="7974" from-port="0" to-layer="7975" to-port="3" /> <edge from-layer="7975" from-port="4" to-layer="8026" to-port="0" /> <edge from-layer="7975" from-port="4" to-layer="8021" to-port="0" /> <edge from-layer="7975" from-port="4" to-layer="8010" to-port="0" /> <edge from-layer="7975" from-port="4" to-layer="8007" to-port="0" /> <edge from-layer="7976" from-port="0" to-layer="8003" to-port="0" /> <edge from-layer="7977" from-port="0" to-layer="8003" to-port="1" /> <edge from-layer="7978" from-port="0" to-layer="7981" to-port="1" /> <edge from-layer="7979" from-port="0" to-layer="7981" to-port="2" /> <edge from-layer="7980" from-port="0" to-layer="7981" to-port="3" /> <edge from-layer="7981" from-port="4" to-layer="7983" to-port="0" /> <edge from-layer="7982" from-port="0" to-layer="7983" to-port="1" /> <edge from-layer="7983" from-port="2" to-layer="7984" to-port="0" /> <edge from-layer="7983" from-port="2" to-layer="8072" to-port="0" /> <edge from-layer="7983" from-port="2" to-layer="8045" to-port="0" /> <edge from-layer="7984" from-port="1" to-layer="7986" to-port="0" /> <edge from-layer="7985" from-port="0" to-layer="7986" to-port="2" /> <edge from-layer="7986" from-port="3" to-layer="7999" to-port="0" /> <edge from-layer="7987" from-port="0" to-layer="7992" to-port="0" /> <edge from-layer="7988" from-port="0" to-layer="7991" to-port="1" /> <edge from-layer="7989" from-port="0" to-layer="7991" to-port="2" /> <edge from-layer="7990" from-port="0" to-layer="7991" to-port="3" /> <edge from-layer="7991" from-port="4" to-layer="7992" to-port="1" /> <edge from-layer="7992" from-port="2" to-layer="7993" to-port="0" /> <edge from-layer="7993" from-port="1" to-layer="7995" to-port="0" /> <edge from-layer="7994" from-port="0" to-layer="7995" to-port="2" /> <edge from-layer="7995" from-port="3" to-layer="7996" to-port="0" /> <edge from-layer="7995" from-port="3" to-layer="8074" to-port="0" /> <edge from-layer="7996" from-port="1" to-layer="7998" to-port="0" /> <edge from-layer="7997" from-port="0" to-layer="7998" to-port="2" /> <edge from-layer="7998" from-port="3" to-layer="7999" to-port="1" /> <edge from-layer="7999" from-port="2" to-layer="8001" to-port="0" /> <edge from-layer="8000" from-port="0" to-layer="8001" to-port="1" /> <edge from-layer="8001" from-port="2" to-layer="8031" to-port="2" /> <edge from-layer="8001" from-port="2" to-layer="8003" to-port="2" /> <edge from-layer="8002" from-port="0" to-layer="8003" to-port="3" /> <edge from-layer="8003" from-port="4" to-layer="8005" to-port="0" /> <edge from-layer="8004" from-port="0" to-layer="8005" to-port="2" /> <edge from-layer="8005" from-port="3" to-layer="8006" to-port="0" /> <edge from-layer="8006" from-port="2" to-layer="8007" to-port="1" /> <edge from-layer="8006" from-port="2" to-layer="8046" to-port="1" /> <edge from-layer="8007" from-port="2" to-layer="8036" to-port="0" /> <edge from-layer="8008" from-port="0" to-layer="8018" to-port="0" /> <edge from-layer="8009" from-port="0" to-layer="8018" to-port="1" /> <edge from-layer="8010" from-port="1" to-layer="8012" to-port="0" /> <edge from-layer="8011" from-port="0" to-layer="8012" to-port="2" /> <edge from-layer="8012" from-port="3" to-layer="8013" to-port="0" /> <edge from-layer="8013" from-port="2" to-layer="8014" to-port="0" /> <edge from-layer="8014" from-port="1" to-layer="8016" to-port="0" /> <edge from-layer="8015" from-port="0" to-layer="8016" to-port="1" /> <edge from-layer="8016" from-port="2" to-layer="8018" to-port="2" /> <edge from-layer="8017" from-port="0" to-layer="8018" to-port="3" /> <edge from-layer="8018" from-port="4" to-layer="8021" to-port="1" /> <edge from-layer="8018" from-port="4" to-layer="8026" to-port="2" /> <edge from-layer="8019" from-port="0" to-layer="8021" to-port="2" /> <edge from-layer="8020" from-port="0" to-layer="8021" to-port="3" /> <edge from-layer="8021" from-port="4" to-layer="8023" to-port="0" /> <edge from-layer="8022" from-port="0" to-layer="8023" to-port="1" /> <edge from-layer="8023" from-port="2" to-layer="8027" to-port="0" /> <edge from-layer="8024" from-port="0" to-layer="8026" to-port="1" /> <edge from-layer="8025" from-port="0" to-layer="8026" to-port="3" /> <edge from-layer="8026" from-port="4" to-layer="8027" to-port="1" /> <edge from-layer="8027" from-port="2" to-layer="8035" to-port="0" /> <edge from-layer="8028" from-port="0" to-layer="8031" to-port="0" /> <edge from-layer="8029" from-port="0" to-layer="8031" to-port="1" /> <edge from-layer="8030" from-port="0" to-layer="8031" to-port="3" /> <edge from-layer="8031" from-port="4" to-layer="8033" to-port="0" /> <edge from-layer="8032" from-port="0" to-layer="8033" to-port="2" /> <edge from-layer="8033" from-port="3" to-layer="8034" to-port="0" /> <edge from-layer="8034" from-port="2" to-layer="8067" to-port="1" /> <edge from-layer="8034" from-port="2" to-layer="8035" to-port="1" /> <edge from-layer="8035" from-port="2" to-layer="8036" to-port="1" /> <edge from-layer="8036" from-port="2" to-layer="8041" to-port="0" /> <edge from-layer="8037" from-port="0" to-layer="8040" to-port="1" /> <edge from-layer="8038" from-port="0" to-layer="8040" to-port="2" /> <edge from-layer="8039" from-port="0" to-layer="8040" to-port="3" /> <edge from-layer="8040" from-port="4" to-layer="8041" to-port="1" /> <edge from-layer="8041" from-port="2" to-layer="8097" to-port="0" /> <edge from-layer="8041" from-port="2" to-layer="8135" to-port="0" /> <edge from-layer="8042" from-port="0" to-layer="8045" to-port="1" /> <edge from-layer="8043" from-port="0" to-layer="8045" to-port="2" /> <edge from-layer="8044" from-port="0" to-layer="8045" to-port="3" /> <edge from-layer="8045" from-port="4" to-layer="8065" to-port="0" /> <edge from-layer="8045" from-port="4" to-layer="8046" to-port="0" /> <edge from-layer="8045" from-port="4" to-layer="8060" to-port="0" /> <edge from-layer="8045" from-port="4" to-layer="8049" to-port="0" /> <edge from-layer="8046" from-port="2" to-layer="8068" to-port="0" /> <edge from-layer="8047" from-port="0" to-layer="8057" to-port="0" /> <edge from-layer="8048" from-port="0" to-layer="8057" to-port="1" /> <edge from-layer="8049" from-port="1" to-layer="8051" to-port="0" /> <edge from-layer="8050" from-port="0" to-layer="8051" to-port="2" /> <edge from-layer="8051" from-port="3" to-layer="8052" to-port="0" /> <edge from-layer="8052" from-port="2" to-layer="8053" to-port="0" /> <edge from-layer="8053" from-port="1" to-layer="8055" to-port="0" /> <edge from-layer="8054" from-port="0" to-layer="8055" to-port="1" /> <edge from-layer="8055" from-port="2" to-layer="8057" to-port="2" /> <edge from-layer="8056" from-port="0" to-layer="8057" to-port="3" /> <edge from-layer="8057" from-port="4" to-layer="8065" to-port="2" /> <edge from-layer="8057" from-port="4" to-layer="8060" to-port="1" /> <edge from-layer="8058" from-port="0" to-layer="8060" to-port="2" /> <edge from-layer="8059" from-port="0" to-layer="8060" to-port="3" /> <edge from-layer="8060" from-port="4" to-layer="8062" to-port="0" /> <edge from-layer="8061" from-port="0" to-layer="8062" to-port="1" /> <edge from-layer="8062" from-port="2" to-layer="8066" to-port="0" /> <edge from-layer="8063" from-port="0" to-layer="8065" to-port="1" /> <edge from-layer="8064" from-port="0" to-layer="8065" to-port="3" /> <edge from-layer="8065" from-port="4" to-layer="8066" to-port="1" /> <edge from-layer="8066" from-port="2" to-layer="8067" to-port="0" /> <edge from-layer="8067" from-port="2" to-layer="8068" to-port="1" /> <edge from-layer="8068" from-port="2" to-layer="8073" to-port="0" /> <edge from-layer="8069" from-port="0" to-layer="8072" to-port="1" /> <edge from-layer="8070" from-port="0" to-layer="8072" to-port="2" /> <edge from-layer="8071" from-port="0" to-layer="8072" to-port="3" /> <edge from-layer="8072" from-port="4" to-layer="8073" to-port="1" /> <edge from-layer="8073" from-port="2" to-layer="8074" to-port="1" /> <edge from-layer="8074" from-port="2" to-layer="10117" to-port="0" /> <edge from-layer="8074" from-port="2" to-layer="8094" to-port="0" /> <edge from-layer="8074" from-port="2" to-layer="8135" to-port="1" /> <edge from-layer="8075" from-port="0" to-layer="8080" to-port="0" /> <edge from-layer="8076" from-port="0" to-layer="8079" to-port="1" /> <edge from-layer="8077" from-port="0" to-layer="8079" to-port="2" /> <edge from-layer="8078" from-port="0" to-layer="8079" to-port="3" /> <edge from-layer="8079" from-port="4" to-layer="8080" to-port="1" /> <edge from-layer="8080" from-port="2" to-layer="8081" to-port="0" /> <edge from-layer="8081" from-port="1" to-layer="8083" to-port="0" /> <edge from-layer="8082" from-port="0" to-layer="8083" to-port="2" /> <edge from-layer="8083" from-port="3" to-layer="8090" to-port="0" /> <edge from-layer="8084" from-port="0" to-layer="8087" to-port="1" /> <edge from-layer="8085" from-port="0" to-layer="8087" to-port="2" /> <edge from-layer="8086" from-port="0" to-layer="8087" to-port="3" /> <edge from-layer="8087" from-port="4" to-layer="8089" to-port="0" /> <edge from-layer="8088" from-port="0" to-layer="8089" to-port="1" /> <edge from-layer="8089" from-port="2" to-layer="8090" to-port="1" /> <edge from-layer="8090" from-port="2" to-layer="10116" to-port="0" /> <edge from-layer="8090" from-port="2" to-layer="8135" to-port="2" /> <edge from-layer="8091" from-port="0" to-layer="8109" to-port="0" /> <edge from-layer="8092" from-port="0" to-layer="8104" to-port="0" /> <edge from-layer="8093" from-port="0" to-layer="8104" to-port="1" /> <edge from-layer="8093" from-port="0" to-layer="8107" to-port="1" /> <edge from-layer="8094" from-port="1" to-layer="8096" to-port="0" /> <edge from-layer="8095" from-port="0" to-layer="8096" to-port="2" /> <edge from-layer="8096" from-port="3" to-layer="8100" to-port="0" /> <edge from-layer="8096" from-port="3" to-layer="8106" to-port="0" /> <edge from-layer="8097" from-port="1" to-layer="8124" to-port="0" /> <edge from-layer="8097" from-port="1" to-layer="8099" to-port="0" /> <edge from-layer="8098" from-port="0" to-layer="8099" to-port="2" /> <edge from-layer="8099" from-port="3" to-layer="8100" to-port="1" /> <edge from-layer="8100" from-port="2" to-layer="8102" to-port="0" /> <edge from-layer="8101" from-port="0" to-layer="8102" to-port="1" /> <edge from-layer="8101" from-port="0" to-layer="8106" to-port="1" /> <edge from-layer="8102" from-port="2" to-layer="8104" to-port="2" /> <edge from-layer="8103" from-port="0" to-layer="8104" to-port="3" /> <edge from-layer="8103" from-port="0" to-layer="8107" to-port="3" /> <edge from-layer="8104" from-port="4" to-layer="8109" to-port="1" /> <edge from-layer="8105" from-port="0" to-layer="8107" to-port="0" /> <edge from-layer="8106" from-port="2" to-layer="8114" to-port="2" /> <edge from-layer="8106" from-port="2" to-layer="8107" to-port="2" /> <edge from-layer="8107" from-port="4" to-layer="8109" to-port="2" /> <edge from-layer="8108" from-port="0" to-layer="8109" to-port="3" /> <edge from-layer="8109" from-port="4" to-layer="8116" to-port="0" /> <edge from-layer="8110" from-port="0" to-layer="8116" to-port="1" /> <edge from-layer="8111" from-port="0" to-layer="8114" to-port="0" /> <edge from-layer="8112" from-port="0" to-layer="8114" to-port="1" /> <edge from-layer="8113" from-port="0" to-layer="8114" to-port="3" /> <edge from-layer="8114" from-port="4" to-layer="8116" to-port="2" /> <edge from-layer="8115" from-port="0" to-layer="8116" to-port="3" /> <edge from-layer="8116" from-port="4" to-layer="8119" to-port="0" /> <edge from-layer="8117" from-port="0" to-layer="8119" to-port="1" /> <edge from-layer="8118" from-port="0" to-layer="8119" to-port="2" /> <edge from-layer="8119" from-port="3" to-layer="8120" to-port="0" /> <edge from-layer="8120" from-port="2" to-layer="8134" to-port="0" /> <edge from-layer="8120" from-port="2" to-layer="8121" to-port="0" /> <edge from-layer="8121" from-port="1" to-layer="8133" to-port="0" /> <edge from-layer="8122" from-port="0" to-layer="8124" to-port="1" /> <edge from-layer="8123" from-port="0" to-layer="8124" to-port="2" /> <edge from-layer="8124" from-port="3" to-layer="8128" to-port="0" /> <edge from-layer="8125" from-port="0" to-layer="8128" to-port="1" /> <edge from-layer="8126" from-port="0" to-layer="8128" to-port="2" /> <edge from-layer="8127" from-port="0" to-layer="8128" to-port="3" /> <edge from-layer="8128" from-port="4" to-layer="8132" to-port="2" /> <edge from-layer="8128" from-port="4" to-layer="8130" to-port="0" /> <edge from-layer="8129" from-port="0" to-layer="8130" to-port="1" /> <edge from-layer="8130" from-port="2" to-layer="8132" to-port="0" /> <edge from-layer="8131" from-port="0" to-layer="8132" to-port="1" /> <edge from-layer="8132" from-port="3" to-layer="8133" to-port="1" /> <edge from-layer="8133" from-port="2" to-layer="8134" to-port="1" /> <edge from-layer="8134" from-port="2" to-layer="8135" to-port="3" /> <edge from-layer="8135" from-port="4" to-layer="8137" to-port="0" /> <edge from-layer="8136" from-port="0" to-layer="8137" to-port="1" /> <edge from-layer="8137" from-port="2" to-layer="8139" to-port="0" /> <edge from-layer="8138" from-port="0" to-layer="8139" to-port="1" /> <edge from-layer="8139" from-port="2" to-layer="8147" to-port="0" /> <edge from-layer="8140" from-port="0" to-layer="8141" to-port="0" /> <edge from-layer="8141" from-port="1" to-layer="8144" to-port="0" /> <edge from-layer="8142" from-port="0" to-layer="8143" to-port="0" /> <edge from-layer="8143" from-port="1" to-layer="8144" to-port="1" /> <edge from-layer="8144" from-port="2" to-layer="8146" to-port="0" /> <edge from-layer="8145" from-port="0" to-layer="8146" to-port="1" /> <edge from-layer="8146" from-port="2" to-layer="8147" to-port="1" /> <edge from-layer="8147" from-port="2" to-layer="8149" to-port="0" /> <edge from-layer="8148" from-port="0" to-layer="8149" to-port="1" /> <edge from-layer="8149" from-port="2" to-layer="8150" to-port="1" /> <edge from-layer="8150" from-port="2" to-layer="8151" to-port="0" /> <edge from-layer="8151" from-port="2" to-layer="8153" to-port="0" /> <edge from-layer="8151" from-port="2" to-layer="8180" to-port="0" /> <edge from-layer="8151" from-port="2" to-layer="8382" to-port="1" /> <edge from-layer="8152" from-port="0" to-layer="8153" to-port="1" /> <edge from-layer="8153" from-port="2" to-layer="8155" to-port="0" /> <edge from-layer="8154" from-port="0" to-layer="8155" to-port="1" /> <edge from-layer="8155" from-port="2" to-layer="8157" to-port="0" /> <edge from-layer="8156" from-port="0" to-layer="8157" to-port="1" /> <edge from-layer="8157" from-port="2" to-layer="8165" to-port="0" /> <edge from-layer="8158" from-port="0" to-layer="8159" to-port="0" /> <edge from-layer="8159" from-port="1" to-layer="8162" to-port="0" /> <edge from-layer="8160" from-port="0" to-layer="8161" to-port="0" /> <edge from-layer="8161" from-port="1" to-layer="8162" to-port="1" /> <edge from-layer="8162" from-port="2" to-layer="8164" to-port="0" /> <edge from-layer="8163" from-port="0" to-layer="8164" to-port="1" /> <edge from-layer="8164" from-port="2" to-layer="8165" to-port="1" /> <edge from-layer="8165" from-port="2" to-layer="8167" to-port="0" /> <edge from-layer="8166" from-port="0" to-layer="8167" to-port="1" /> <edge from-layer="8167" from-port="2" to-layer="8168" to-port="0" /> <edge from-layer="8168" from-port="1" to-layer="8176" to-port="0" /> <edge from-layer="8169" from-port="0" to-layer="8170" to-port="0" /> <edge from-layer="8170" from-port="1" to-layer="8173" to-port="0" /> <edge from-layer="8171" from-port="0" to-layer="8172" to-port="0" /> <edge from-layer="8172" from-port="1" to-layer="8173" to-port="1" /> <edge from-layer="8173" from-port="2" to-layer="8175" to-port="0" /> <edge from-layer="8174" from-port="0" to-layer="8175" to-port="1" /> <edge from-layer="8175" from-port="2" to-layer="8176" to-port="1" /> <edge from-layer="8176" from-port="2" to-layer="8178" to-port="0" /> <edge from-layer="8177" from-port="0" to-layer="8178" to-port="1" /> <edge from-layer="8178" from-port="2" to-layer="8381" to-port="0" /> <edge from-layer="8179" from-port="0" to-layer="8180" to-port="1" /> <edge from-layer="8180" from-port="2" to-layer="8182" to-port="0" /> <edge from-layer="8181" from-port="0" to-layer="8182" to-port="1" /> <edge from-layer="8182" from-port="2" to-layer="8184" to-port="0" /> <edge from-layer="8183" from-port="0" to-layer="8184" to-port="1" /> <edge from-layer="8184" from-port="2" to-layer="8192" to-port="0" /> <edge from-layer="8185" from-port="0" to-layer="8186" to-port="0" /> <edge from-layer="8186" from-port="1" to-layer="8189" to-port="0" /> <edge from-layer="8187" from-port="0" to-layer="8188" to-port="0" /> <edge from-layer="8188" from-port="1" to-layer="8189" to-port="1" /> <edge from-layer="8189" from-port="2" to-layer="8191" to-port="0" /> <edge from-layer="8190" from-port="0" to-layer="8191" to-port="1" /> <edge from-layer="8191" from-port="2" to-layer="8192" to-port="1" /> <edge from-layer="8192" from-port="2" to-layer="8194" to-port="0" /> <edge from-layer="8193" from-port="0" to-layer="8194" to-port="1" /> <edge from-layer="8194" from-port="2" to-layer="8196" to-port="0" /> <edge from-layer="8195" from-port="0" to-layer="8196" to-port="1" /> <edge from-layer="8196" from-port="2" to-layer="8212" to-port="0" /> <edge from-layer="8196" from-port="2" to-layer="8318" to-port="0" /> <edge from-layer="8196" from-port="2" to-layer="8200" to-port="0" /> <edge from-layer="8197" from-port="0" to-layer="8200" to-port="1" /> <edge from-layer="8198" from-port="0" to-layer="8200" to-port="2" /> <edge from-layer="8199" from-port="0" to-layer="8200" to-port="3" /> <edge from-layer="8200" from-port="4" to-layer="8202" to-port="0" /> <edge from-layer="8201" from-port="0" to-layer="8202" to-port="1" /> <edge from-layer="8202" from-port="2" to-layer="8206" to-port="0" /> <edge from-layer="8202" from-port="2" to-layer="8271" to-port="0" /> <edge from-layer="8203" from-port="0" to-layer="8206" to-port="1" /> <edge from-layer="8204" from-port="0" to-layer="8206" to-port="2" /> <edge from-layer="8205" from-port="0" to-layer="8206" to-port="3" /> <edge from-layer="8206" from-port="4" to-layer="8257" to-port="0" /> <edge from-layer="8206" from-port="4" to-layer="8252" to-port="0" /> <edge from-layer="8206" from-port="4" to-layer="8238" to-port="0" /> <edge from-layer="8206" from-port="4" to-layer="8241" to-port="0" /> <edge from-layer="8207" from-port="0" to-layer="8234" to-port="0" /> <edge from-layer="8208" from-port="0" to-layer="8234" to-port="1" /> <edge from-layer="8209" from-port="0" to-layer="8212" to-port="1" /> <edge from-layer="8210" from-port="0" to-layer="8212" to-port="2" /> <edge from-layer="8211" from-port="0" to-layer="8212" to-port="3" /> <edge from-layer="8212" from-port="4" to-layer="8214" to-port="0" /> <edge from-layer="8213" from-port="0" to-layer="8214" to-port="1" /> <edge from-layer="8214" from-port="2" to-layer="8303" to-port="0" /> <edge from-layer="8214" from-port="2" to-layer="8215" to-port="0" /> <edge from-layer="8214" from-port="2" to-layer="8276" to-port="0" /> <edge from-layer="8215" from-port="1" to-layer="8217" to-port="0" /> <edge from-layer="8216" from-port="0" to-layer="8217" to-port="2" /> <edge from-layer="8217" from-port="3" to-layer="8230" to-port="0" /> <edge from-layer="8218" from-port="0" to-layer="8223" to-port="0" /> <edge from-layer="8219" from-port="0" to-layer="8222" to-port="1" /> <edge from-layer="8220" from-port="0" to-layer="8222" to-port="2" /> <edge from-layer="8221" from-port="0" to-layer="8222" to-port="3" /> <edge from-layer="8222" from-port="4" to-layer="8223" to-port="1" /> <edge from-layer="8223" from-port="2" to-layer="8224" to-port="0" /> <edge from-layer="8224" from-port="1" to-layer="8226" to-port="0" /> <edge from-layer="8225" from-port="0" to-layer="8226" to-port="2" /> <edge from-layer="8226" from-port="3" to-layer="8227" to-port="0" /> <edge from-layer="8226" from-port="3" to-layer="8305" to-port="0" /> <edge from-layer="8227" from-port="1" to-layer="8229" to-port="0" /> <edge from-layer="8228" from-port="0" to-layer="8229" to-port="2" /> <edge from-layer="8229" from-port="3" to-layer="8230" to-port="1" /> <edge from-layer="8230" from-port="2" to-layer="8232" to-port="0" /> <edge from-layer="8231" from-port="0" to-layer="8232" to-port="1" /> <edge from-layer="8232" from-port="2" to-layer="8262" to-port="2" /> <edge from-layer="8232" from-port="2" to-layer="8234" to-port="2" /> <edge from-layer="8233" from-port="0" to-layer="8234" to-port="3" /> <edge from-layer="8234" from-port="4" to-layer="8236" to-port="0" /> <edge from-layer="8235" from-port="0" to-layer="8236" to-port="2" /> <edge from-layer="8236" from-port="3" to-layer="8237" to-port="0" /> <edge from-layer="8237" from-port="2" to-layer="8238" to-port="1" /> <edge from-layer="8237" from-port="2" to-layer="8277" to-port="1" /> <edge from-layer="8238" from-port="2" to-layer="8267" to-port="0" /> <edge from-layer="8239" from-port="0" to-layer="8249" to-port="0" /> <edge from-layer="8240" from-port="0" to-layer="8249" to-port="1" /> <edge from-layer="8241" from-port="1" to-layer="8243" to-port="0" /> <edge from-layer="8242" from-port="0" to-layer="8243" to-port="2" /> <edge from-layer="8243" from-port="3" to-layer="8244" to-port="0" /> <edge from-layer="8244" from-port="2" to-layer="8245" to-port="0" /> <edge from-layer="8245" from-port="1" to-layer="8247" to-port="0" /> <edge from-layer="8246" from-port="0" to-layer="8247" to-port="1" /> <edge from-layer="8247" from-port="2" to-layer="8249" to-port="2" /> <edge from-layer="8248" from-port="0" to-layer="8249" to-port="3" /> <edge from-layer="8249" from-port="4" to-layer="8252" to-port="1" /> <edge from-layer="8249" from-port="4" to-layer="8257" to-port="2" /> <edge from-layer="8250" from-port="0" to-layer="8252" to-port="2" /> <edge from-layer="8251" from-port="0" to-layer="8252" to-port="3" /> <edge from-layer="8252" from-port="4" to-layer="8254" to-port="0" /> <edge from-layer="8253" from-port="0" to-layer="8254" to-port="1" /> <edge from-layer="8254" from-port="2" to-layer="8258" to-port="0" /> <edge from-layer="8255" from-port="0" to-layer="8257" to-port="1" /> <edge from-layer="8256" from-port="0" to-layer="8257" to-port="3" /> <edge from-layer="8257" from-port="4" to-layer="8258" to-port="1" /> <edge from-layer="8258" from-port="2" to-layer="8266" to-port="0" /> <edge from-layer="8259" from-port="0" to-layer="8262" to-port="0" /> <edge from-layer="8260" from-port="0" to-layer="8262" to-port="1" /> <edge from-layer="8261" from-port="0" to-layer="8262" to-port="3" /> <edge from-layer="8262" from-port="4" to-layer="8264" to-port="0" /> <edge from-layer="8263" from-port="0" to-layer="8264" to-port="2" /> <edge from-layer="8264" from-port="3" to-layer="8265" to-port="0" /> <edge from-layer="8265" from-port="2" to-layer="8298" to-port="1" /> <edge from-layer="8265" from-port="2" to-layer="8266" to-port="1" /> <edge from-layer="8266" from-port="2" to-layer="8267" to-port="1" /> <edge from-layer="8267" from-port="2" to-layer="8272" to-port="0" /> <edge from-layer="8268" from-port="0" to-layer="8271" to-port="1" /> <edge from-layer="8269" from-port="0" to-layer="8271" to-port="2" /> <edge from-layer="8270" from-port="0" to-layer="8271" to-port="3" /> <edge from-layer="8271" from-port="4" to-layer="8272" to-port="1" /> <edge from-layer="8272" from-port="2" to-layer="8328" to-port="0" /> <edge from-layer="8272" from-port="2" to-layer="8366" to-port="0" /> <edge from-layer="8273" from-port="0" to-layer="8276" to-port="1" /> <edge from-layer="8274" from-port="0" to-layer="8276" to-port="2" /> <edge from-layer="8275" from-port="0" to-layer="8276" to-port="3" /> <edge from-layer="8276" from-port="4" to-layer="8277" to-port="0" /> <edge from-layer="8276" from-port="4" to-layer="8296" to-port="0" /> <edge from-layer="8276" from-port="4" to-layer="8280" to-port="0" /> <edge from-layer="8276" from-port="4" to-layer="8291" to-port="0" /> <edge from-layer="8277" from-port="2" to-layer="8299" to-port="0" /> <edge from-layer="8278" from-port="0" to-layer="8288" to-port="0" /> <edge from-layer="8279" from-port="0" to-layer="8288" to-port="1" /> <edge from-layer="8280" from-port="1" to-layer="8282" to-port="0" /> <edge from-layer="8281" from-port="0" to-layer="8282" to-port="2" /> <edge from-layer="8282" from-port="3" to-layer="8283" to-port="0" /> <edge from-layer="8283" from-port="2" to-layer="8284" to-port="0" /> <edge from-layer="8284" from-port="1" to-layer="8286" to-port="0" /> <edge from-layer="8285" from-port="0" to-layer="8286" to-port="1" /> <edge from-layer="8286" from-port="2" to-layer="8288" to-port="2" /> <edge from-layer="8287" from-port="0" to-layer="8288" to-port="3" /> <edge from-layer="8288" from-port="4" to-layer="8296" to-port="2" /> <edge from-layer="8288" from-port="4" to-layer="8291" to-port="1" /> <edge from-layer="8289" from-port="0" to-layer="8291" to-port="2" /> <edge from-layer="8290" from-port="0" to-layer="8291" to-port="3" /> <edge from-layer="8291" from-port="4" to-layer="8293" to-port="0" /> <edge from-layer="8292" from-port="0" to-layer="8293" to-port="1" /> <edge from-layer="8293" from-port="2" to-layer="8297" to-port="0" /> <edge from-layer="8294" from-port="0" to-layer="8296" to-port="1" /> <edge from-layer="8295" from-port="0" to-layer="8296" to-port="3" /> <edge from-layer="8296" from-port="4" to-layer="8297" to-port="1" /> <edge from-layer="8297" from-port="2" to-layer="8298" to-port="0" /> <edge from-layer="8298" from-port="2" to-layer="8299" to-port="1" /> <edge from-layer="8299" from-port="2" to-layer="8304" to-port="0" /> <edge from-layer="8300" from-port="0" to-layer="8303" to-port="1" /> <edge from-layer="8301" from-port="0" to-layer="8303" to-port="2" /> <edge from-layer="8302" from-port="0" to-layer="8303" to-port="3" /> <edge from-layer="8303" from-port="4" to-layer="8304" to-port="1" /> <edge from-layer="8304" from-port="2" to-layer="8305" to-port="1" /> <edge from-layer="8305" from-port="2" to-layer="8366" to-port="1" /> <edge from-layer="8305" from-port="2" to-layer="10115" to-port="0" /> <edge from-layer="8305" from-port="2" to-layer="8325" to-port="0" /> <edge from-layer="8306" from-port="0" to-layer="8311" to-port="0" /> <edge from-layer="8307" from-port="0" to-layer="8310" to-port="1" /> <edge from-layer="8308" from-port="0" to-layer="8310" to-port="2" /> <edge from-layer="8309" from-port="0" to-layer="8310" to-port="3" /> <edge from-layer="8310" from-port="4" to-layer="8311" to-port="1" /> <edge from-layer="8311" from-port="2" to-layer="8312" to-port="0" /> <edge from-layer="8312" from-port="1" to-layer="8314" to-port="0" /> <edge from-layer="8313" from-port="0" to-layer="8314" to-port="2" /> <edge from-layer="8314" from-port="3" to-layer="8321" to-port="0" /> <edge from-layer="8315" from-port="0" to-layer="8318" to-port="1" /> <edge from-layer="8316" from-port="0" to-layer="8318" to-port="2" /> <edge from-layer="8317" from-port="0" to-layer="8318" to-port="3" /> <edge from-layer="8318" from-port="4" to-layer="8320" to-port="0" /> <edge from-layer="8319" from-port="0" to-layer="8320" to-port="1" /> <edge from-layer="8320" from-port="2" to-layer="8321" to-port="1" /> <edge from-layer="8321" from-port="2" to-layer="10114" to-port="0" /> <edge from-layer="8321" from-port="2" to-layer="8366" to-port="2" /> <edge from-layer="8322" from-port="0" to-layer="8340" to-port="0" /> <edge from-layer="8323" from-port="0" to-layer="8335" to-port="0" /> <edge from-layer="8324" from-port="0" to-layer="8338" to-port="1" /> <edge from-layer="8324" from-port="0" to-layer="8335" to-port="1" /> <edge from-layer="8325" from-port="1" to-layer="8327" to-port="0" /> <edge from-layer="8326" from-port="0" to-layer="8327" to-port="2" /> <edge from-layer="8327" from-port="3" to-layer="8331" to-port="0" /> <edge from-layer="8327" from-port="3" to-layer="8337" to-port="0" /> <edge from-layer="8328" from-port="1" to-layer="8330" to-port="0" /> <edge from-layer="8328" from-port="1" to-layer="8355" to-port="0" /> <edge from-layer="8329" from-port="0" to-layer="8330" to-port="2" /> <edge from-layer="8330" from-port="3" to-layer="8331" to-port="1" /> <edge from-layer="8331" from-port="2" to-layer="8333" to-port="0" /> <edge from-layer="8332" from-port="0" to-layer="8337" to-port="1" /> <edge from-layer="8332" from-port="0" to-layer="8333" to-port="1" /> <edge from-layer="8333" from-port="2" to-layer="8335" to-port="2" /> <edge from-layer="8334" from-port="0" to-layer="8335" to-port="3" /> <edge from-layer="8334" from-port="0" to-layer="8338" to-port="3" /> <edge from-layer="8335" from-port="4" to-layer="8340" to-port="1" /> <edge from-layer="8336" from-port="0" to-layer="8338" to-port="0" /> <edge from-layer="8337" from-port="2" to-layer="8338" to-port="2" /> <edge from-layer="8337" from-port="2" to-layer="8345" to-port="2" /> <edge from-layer="8338" from-port="4" to-layer="8340" to-port="2" /> <edge from-layer="8339" from-port="0" to-layer="8340" to-port="3" /> <edge from-layer="8340" from-port="4" to-layer="8347" to-port="0" /> <edge from-layer="8341" from-port="0" to-layer="8347" to-port="1" /> <edge from-layer="8342" from-port="0" to-layer="8345" to-port="0" /> <edge from-layer="8343" from-port="0" to-layer="8345" to-port="1" /> <edge from-layer="8344" from-port="0" to-layer="8345" to-port="3" /> <edge from-layer="8345" from-port="4" to-layer="8347" to-port="2" /> <edge from-layer="8346" from-port="0" to-layer="8347" to-port="3" /> <edge from-layer="8347" from-port="4" to-layer="8350" to-port="0" /> <edge from-layer="8348" from-port="0" to-layer="8350" to-port="1" /> <edge from-layer="8349" from-port="0" to-layer="8350" to-port="2" /> <edge from-layer="8350" from-port="3" to-layer="8351" to-port="0" /> <edge from-layer="8351" from-port="2" to-layer="8352" to-port="0" /> <edge from-layer="8351" from-port="2" to-layer="8365" to-port="0" /> <edge from-layer="8352" from-port="1" to-layer="8364" to-port="0" /> <edge from-layer="8353" from-port="0" to-layer="8355" to-port="1" /> <edge from-layer="8354" from-port="0" to-layer="8355" to-port="2" /> <edge from-layer="8355" from-port="3" to-layer="8359" to-port="0" /> <edge from-layer="8356" from-port="0" to-layer="8359" to-port="1" /> <edge from-layer="8357" from-port="0" to-layer="8359" to-port="2" /> <edge from-layer="8358" from-port="0" to-layer="8359" to-port="3" /> <edge from-layer="8359" from-port="4" to-layer="8363" to-port="2" /> <edge from-layer="8359" from-port="4" to-layer="8361" to-port="0" /> <edge from-layer="8360" from-port="0" to-layer="8361" to-port="1" /> <edge from-layer="8361" from-port="2" to-layer="8363" to-port="0" /> <edge from-layer="8362" from-port="0" to-layer="8363" to-port="1" /> <edge from-layer="8363" from-port="3" to-layer="8364" to-port="1" /> <edge from-layer="8364" from-port="2" to-layer="8365" to-port="1" /> <edge from-layer="8365" from-port="2" to-layer="8366" to-port="3" /> <edge from-layer="8366" from-port="4" to-layer="8368" to-port="0" /> <edge from-layer="8367" from-port="0" to-layer="8368" to-port="1" /> <edge from-layer="8368" from-port="2" to-layer="8370" to-port="0" /> <edge from-layer="8369" from-port="0" to-layer="8370" to-port="1" /> <edge from-layer="8370" from-port="2" to-layer="8378" to-port="0" /> <edge from-layer="8371" from-port="0" to-layer="8372" to-port="0" /> <edge from-layer="8372" from-port="1" to-layer="8375" to-port="0" /> <edge from-layer="8373" from-port="0" to-layer="8374" to-port="0" /> <edge from-layer="8374" from-port="1" to-layer="8375" to-port="1" /> <edge from-layer="8375" from-port="2" to-layer="8377" to-port="0" /> <edge from-layer="8376" from-port="0" to-layer="8377" to-port="1" /> <edge from-layer="8377" from-port="2" to-layer="8378" to-port="1" /> <edge from-layer="8378" from-port="2" to-layer="8380" to-port="0" /> <edge from-layer="8379" from-port="0" to-layer="8380" to-port="1" /> <edge from-layer="8380" from-port="2" to-layer="8381" to-port="1" /> <edge from-layer="8381" from-port="2" to-layer="8382" to-port="0" /> <edge from-layer="8382" from-port="2" to-layer="8613" to-port="1" /> <edge from-layer="8382" from-port="2" to-layer="8411" to-port="0" /> <edge from-layer="8382" from-port="2" to-layer="8384" to-port="0" /> <edge from-layer="8383" from-port="0" to-layer="8384" to-port="1" /> <edge from-layer="8384" from-port="2" to-layer="8386" to-port="0" /> <edge from-layer="8385" from-port="0" to-layer="8386" to-port="1" /> <edge from-layer="8386" from-port="2" to-layer="8388" to-port="0" /> <edge from-layer="8387" from-port="0" to-layer="8388" to-port="1" /> <edge from-layer="8388" from-port="2" to-layer="8396" to-port="0" /> <edge from-layer="8389" from-port="0" to-layer="8390" to-port="0" /> <edge from-layer="8390" from-port="1" to-layer="8393" to-port="0" /> <edge from-layer="8391" from-port="0" to-layer="8392" to-port="0" /> <edge from-layer="8392" from-port="1" to-layer="8393" to-port="1" /> <edge from-layer="8393" from-port="2" to-layer="8395" to-port="0" /> <edge from-layer="8394" from-port="0" to-layer="8395" to-port="1" /> <edge from-layer="8395" from-port="2" to-layer="8396" to-port="1" /> <edge from-layer="8396" from-port="2" to-layer="8398" to-port="0" /> <edge from-layer="8397" from-port="0" to-layer="8398" to-port="1" /> <edge from-layer="8398" from-port="2" to-layer="8399" to-port="0" /> <edge from-layer="8399" from-port="1" to-layer="8407" to-port="0" /> <edge from-layer="8400" from-port="0" to-layer="8401" to-port="0" /> <edge from-layer="8401" from-port="1" to-layer="8404" to-port="0" /> <edge from-layer="8402" from-port="0" to-layer="8403" to-port="0" /> <edge from-layer="8403" from-port="1" to-layer="8404" to-port="1" /> <edge from-layer="8404" from-port="2" to-layer="8406" to-port="0" /> <edge from-layer="8405" from-port="0" to-layer="8406" to-port="1" /> <edge from-layer="8406" from-port="2" to-layer="8407" to-port="1" /> <edge from-layer="8407" from-port="2" to-layer="8409" to-port="0" /> <edge from-layer="8408" from-port="0" to-layer="8409" to-port="1" /> <edge from-layer="8409" from-port="2" to-layer="8612" to-port="0" /> <edge from-layer="8410" from-port="0" to-layer="8411" to-port="1" /> <edge from-layer="8411" from-port="2" to-layer="8413" to-port="0" /> <edge from-layer="8412" from-port="0" to-layer="8413" to-port="1" /> <edge from-layer="8413" from-port="2" to-layer="8415" to-port="0" /> <edge from-layer="8414" from-port="0" to-layer="8415" to-port="1" /> <edge from-layer="8415" from-port="2" to-layer="8423" to-port="0" /> <edge from-layer="8416" from-port="0" to-layer="8417" to-port="0" /> <edge from-layer="8417" from-port="1" to-layer="8420" to-port="0" /> <edge from-layer="8418" from-port="0" to-layer="8419" to-port="0" /> <edge from-layer="8419" from-port="1" to-layer="8420" to-port="1" /> <edge from-layer="8420" from-port="2" to-layer="8422" to-port="0" /> <edge from-layer="8421" from-port="0" to-layer="8422" to-port="1" /> <edge from-layer="8422" from-port="2" to-layer="8423" to-port="1" /> <edge from-layer="8423" from-port="2" to-layer="8425" to-port="0" /> <edge from-layer="8424" from-port="0" to-layer="8425" to-port="1" /> <edge from-layer="8425" from-port="2" to-layer="8427" to-port="0" /> <edge from-layer="8426" from-port="0" to-layer="8427" to-port="1" /> <edge from-layer="8427" from-port="2" to-layer="8443" to-port="0" /> <edge from-layer="8427" from-port="2" to-layer="8431" to-port="0" /> <edge from-layer="8427" from-port="2" to-layer="8549" to-port="0" /> <edge from-layer="8428" from-port="0" to-layer="8431" to-port="1" /> <edge from-layer="8429" from-port="0" to-layer="8431" to-port="2" /> <edge from-layer="8430" from-port="0" to-layer="8431" to-port="3" /> <edge from-layer="8431" from-port="4" to-layer="8433" to-port="0" /> <edge from-layer="8432" from-port="0" to-layer="8433" to-port="1" /> <edge from-layer="8433" from-port="2" to-layer="8437" to-port="0" /> <edge from-layer="8433" from-port="2" to-layer="8502" to-port="0" /> <edge from-layer="8434" from-port="0" to-layer="8437" to-port="1" /> <edge from-layer="8435" from-port="0" to-layer="8437" to-port="2" /> <edge from-layer="8436" from-port="0" to-layer="8437" to-port="3" /> <edge from-layer="8437" from-port="4" to-layer="8488" to-port="0" /> <edge from-layer="8437" from-port="4" to-layer="8483" to-port="0" /> <edge from-layer="8437" from-port="4" to-layer="8472" to-port="0" /> <edge from-layer="8437" from-port="4" to-layer="8469" to-port="0" /> <edge from-layer="8438" from-port="0" to-layer="8465" to-port="0" /> <edge from-layer="8439" from-port="0" to-layer="8465" to-port="1" /> <edge from-layer="8440" from-port="0" to-layer="8443" to-port="1" /> <edge from-layer="8441" from-port="0" to-layer="8443" to-port="2" /> <edge from-layer="8442" from-port="0" to-layer="8443" to-port="3" /> <edge from-layer="8443" from-port="4" to-layer="8445" to-port="0" /> <edge from-layer="8444" from-port="0" to-layer="8445" to-port="1" /> <edge from-layer="8445" from-port="2" to-layer="8507" to-port="0" /> <edge from-layer="8445" from-port="2" to-layer="8534" to-port="0" /> <edge from-layer="8445" from-port="2" to-layer="8446" to-port="0" /> <edge from-layer="8446" from-port="1" to-layer="8448" to-port="0" /> <edge from-layer="8447" from-port="0" to-layer="8448" to-port="2" /> <edge from-layer="8448" from-port="3" to-layer="8461" to-port="0" /> <edge from-layer="8449" from-port="0" to-layer="8454" to-port="0" /> <edge from-layer="8450" from-port="0" to-layer="8453" to-port="1" /> <edge from-layer="8451" from-port="0" to-layer="8453" to-port="2" /> <edge from-layer="8452" from-port="0" to-layer="8453" to-port="3" /> <edge from-layer="8453" from-port="4" to-layer="8454" to-port="1" /> <edge from-layer="8454" from-port="2" to-layer="8455" to-port="0" /> <edge from-layer="8455" from-port="1" to-layer="8457" to-port="0" /> <edge from-layer="8456" from-port="0" to-layer="8457" to-port="2" /> <edge from-layer="8457" from-port="3" to-layer="8458" to-port="0" /> <edge from-layer="8457" from-port="3" to-layer="8536" to-port="0" /> <edge from-layer="8458" from-port="1" to-layer="8460" to-port="0" /> <edge from-layer="8459" from-port="0" to-layer="8460" to-port="2" /> <edge from-layer="8460" from-port="3" to-layer="8461" to-port="1" /> <edge from-layer="8461" from-port="2" to-layer="8463" to-port="0" /> <edge from-layer="8462" from-port="0" to-layer="8463" to-port="1" /> <edge from-layer="8463" from-port="2" to-layer="8493" to-port="2" /> <edge from-layer="8463" from-port="2" to-layer="8465" to-port="2" /> <edge from-layer="8464" from-port="0" to-layer="8465" to-port="3" /> <edge from-layer="8465" from-port="4" to-layer="8467" to-port="0" /> <edge from-layer="8466" from-port="0" to-layer="8467" to-port="2" /> <edge from-layer="8467" from-port="3" to-layer="8468" to-port="0" /> <edge from-layer="8468" from-port="2" to-layer="8469" to-port="1" /> <edge from-layer="8468" from-port="2" to-layer="8508" to-port="1" /> <edge from-layer="8469" from-port="2" to-layer="8498" to-port="0" /> <edge from-layer="8470" from-port="0" to-layer="8480" to-port="0" /> <edge from-layer="8471" from-port="0" to-layer="8480" to-port="1" /> <edge from-layer="8472" from-port="1" to-layer="8474" to-port="0" /> <edge from-layer="8473" from-port="0" to-layer="8474" to-port="2" /> <edge from-layer="8474" from-port="3" to-layer="8475" to-port="0" /> <edge from-layer="8475" from-port="2" to-layer="8476" to-port="0" /> <edge from-layer="8476" from-port="1" to-layer="8478" to-port="0" /> <edge from-layer="8477" from-port="0" to-layer="8478" to-port="1" /> <edge from-layer="8478" from-port="2" to-layer="8480" to-port="2" /> <edge from-layer="8479" from-port="0" to-layer="8480" to-port="3" /> <edge from-layer="8480" from-port="4" to-layer="8488" to-port="2" /> <edge from-layer="8480" from-port="4" to-layer="8483" to-port="1" /> <edge from-layer="8481" from-port="0" to-layer="8483" to-port="2" /> <edge from-layer="8482" from-port="0" to-layer="8483" to-port="3" /> <edge from-layer="8483" from-port="4" to-layer="8485" to-port="0" /> <edge from-layer="8484" from-port="0" to-layer="8485" to-port="1" /> <edge from-layer="8485" from-port="2" to-layer="8489" to-port="0" /> <edge from-layer="8486" from-port="0" to-layer="8488" to-port="1" /> <edge from-layer="8487" from-port="0" to-layer="8488" to-port="3" /> <edge from-layer="8488" from-port="4" to-layer="8489" to-port="1" /> <edge from-layer="8489" from-port="2" to-layer="8497" to-port="0" /> <edge from-layer="8490" from-port="0" to-layer="8493" to-port="0" /> <edge from-layer="8491" from-port="0" to-layer="8493" to-port="1" /> <edge from-layer="8492" from-port="0" to-layer="8493" to-port="3" /> <edge from-layer="8493" from-port="4" to-layer="8495" to-port="0" /> <edge from-layer="8494" from-port="0" to-layer="8495" to-port="2" /> <edge from-layer="8495" from-port="3" to-layer="8496" to-port="0" /> <edge from-layer="8496" from-port="2" to-layer="8529" to-port="1" /> <edge from-layer="8496" from-port="2" to-layer="8497" to-port="1" /> <edge from-layer="8497" from-port="2" to-layer="8498" to-port="1" /> <edge from-layer="8498" from-port="2" to-layer="8503" to-port="0" /> <edge from-layer="8499" from-port="0" to-layer="8502" to-port="1" /> <edge from-layer="8500" from-port="0" to-layer="8502" to-port="2" /> <edge from-layer="8501" from-port="0" to-layer="8502" to-port="3" /> <edge from-layer="8502" from-port="4" to-layer="8503" to-port="1" /> <edge from-layer="8503" from-port="2" to-layer="8559" to-port="0" /> <edge from-layer="8503" from-port="2" to-layer="8597" to-port="0" /> <edge from-layer="8504" from-port="0" to-layer="8507" to-port="1" /> <edge from-layer="8505" from-port="0" to-layer="8507" to-port="2" /> <edge from-layer="8506" from-port="0" to-layer="8507" to-port="3" /> <edge from-layer="8507" from-port="4" to-layer="8527" to-port="0" /> <edge from-layer="8507" from-port="4" to-layer="8522" to-port="0" /> <edge from-layer="8507" from-port="4" to-layer="8508" to-port="0" /> <edge from-layer="8507" from-port="4" to-layer="8511" to-port="0" /> <edge from-layer="8508" from-port="2" to-layer="8530" to-port="0" /> <edge from-layer="8509" from-port="0" to-layer="8519" to-port="0" /> <edge from-layer="8510" from-port="0" to-layer="8519" to-port="1" /> <edge from-layer="8511" from-port="1" to-layer="8513" to-port="0" /> <edge from-layer="8512" from-port="0" to-layer="8513" to-port="2" /> <edge from-layer="8513" from-port="3" to-layer="8514" to-port="0" /> <edge from-layer="8514" from-port="2" to-layer="8515" to-port="0" /> <edge from-layer="8515" from-port="1" to-layer="8517" to-port="0" /> <edge from-layer="8516" from-port="0" to-layer="8517" to-port="1" /> <edge from-layer="8517" from-port="2" to-layer="8519" to-port="2" /> <edge from-layer="8518" from-port="0" to-layer="8519" to-port="3" /> <edge from-layer="8519" from-port="4" to-layer="8527" to-port="2" /> <edge from-layer="8519" from-port="4" to-layer="8522" to-port="1" /> <edge from-layer="8520" from-port="0" to-layer="8522" to-port="2" /> <edge from-layer="8521" from-port="0" to-layer="8522" to-port="3" /> <edge from-layer="8522" from-port="4" to-layer="8524" to-port="0" /> <edge from-layer="8523" from-port="0" to-layer="8524" to-port="1" /> <edge from-layer="8524" from-port="2" to-layer="8528" to-port="0" /> <edge from-layer="8525" from-port="0" to-layer="8527" to-port="1" /> <edge from-layer="8526" from-port="0" to-layer="8527" to-port="3" /> <edge from-layer="8527" from-port="4" to-layer="8528" to-port="1" /> <edge from-layer="8528" from-port="2" to-layer="8529" to-port="0" /> <edge from-layer="8529" from-port="2" to-layer="8530" to-port="1" /> <edge from-layer="8530" from-port="2" to-layer="8535" to-port="0" /> <edge from-layer="8531" from-port="0" to-layer="8534" to-port="1" /> <edge from-layer="8532" from-port="0" to-layer="8534" to-port="2" /> <edge from-layer="8533" from-port="0" to-layer="8534" to-port="3" /> <edge from-layer="8534" from-port="4" to-layer="8535" to-port="1" /> <edge from-layer="8535" from-port="2" to-layer="8536" to-port="1" /> <edge from-layer="8536" from-port="2" to-layer="8597" to-port="1" /> <edge from-layer="8536" from-port="2" to-layer="10113" to-port="0" /> <edge from-layer="8536" from-port="2" to-layer="8556" to-port="0" /> <edge from-layer="8537" from-port="0" to-layer="8542" to-port="0" /> <edge from-layer="8538" from-port="0" to-layer="8541" to-port="1" /> <edge from-layer="8539" from-port="0" to-layer="8541" to-port="2" /> <edge from-layer="8540" from-port="0" to-layer="8541" to-port="3" /> <edge from-layer="8541" from-port="4" to-layer="8542" to-port="1" /> <edge from-layer="8542" from-port="2" to-layer="8543" to-port="0" /> <edge from-layer="8543" from-port="1" to-layer="8545" to-port="0" /> <edge from-layer="8544" from-port="0" to-layer="8545" to-port="2" /> <edge from-layer="8545" from-port="3" to-layer="8552" to-port="0" /> <edge from-layer="8546" from-port="0" to-layer="8549" to-port="1" /> <edge from-layer="8547" from-port="0" to-layer="8549" to-port="2" /> <edge from-layer="8548" from-port="0" to-layer="8549" to-port="3" /> <edge from-layer="8549" from-port="4" to-layer="8551" to-port="0" /> <edge from-layer="8550" from-port="0" to-layer="8551" to-port="1" /> <edge from-layer="8551" from-port="2" to-layer="8552" to-port="1" /> <edge from-layer="8552" from-port="2" to-layer="10112" to-port="0" /> <edge from-layer="8552" from-port="2" to-layer="8597" to-port="2" /> <edge from-layer="8553" from-port="0" to-layer="8571" to-port="0" /> <edge from-layer="8554" from-port="0" to-layer="8566" to-port="0" /> <edge from-layer="8555" from-port="0" to-layer="8569" to-port="1" /> <edge from-layer="8555" from-port="0" to-layer="8566" to-port="1" /> <edge from-layer="8556" from-port="1" to-layer="8558" to-port="0" /> <edge from-layer="8557" from-port="0" to-layer="8558" to-port="2" /> <edge from-layer="8558" from-port="3" to-layer="8562" to-port="0" /> <edge from-layer="8558" from-port="3" to-layer="8568" to-port="0" /> <edge from-layer="8559" from-port="1" to-layer="8561" to-port="0" /> <edge from-layer="8559" from-port="1" to-layer="8586" to-port="0" /> <edge from-layer="8560" from-port="0" to-layer="8561" to-port="2" /> <edge from-layer="8561" from-port="3" to-layer="8562" to-port="1" /> <edge from-layer="8562" from-port="2" to-layer="8564" to-port="0" /> <edge from-layer="8563" from-port="0" to-layer="8568" to-port="1" /> <edge from-layer="8563" from-port="0" to-layer="8564" to-port="1" /> <edge from-layer="8564" from-port="2" to-layer="8566" to-port="2" /> <edge from-layer="8565" from-port="0" to-layer="8569" to-port="3" /> <edge from-layer="8565" from-port="0" to-layer="8566" to-port="3" /> <edge from-layer="8566" from-port="4" to-layer="8571" to-port="1" /> <edge from-layer="8567" from-port="0" to-layer="8569" to-port="0" /> <edge from-layer="8568" from-port="2" to-layer="8576" to-port="2" /> <edge from-layer="8568" from-port="2" to-layer="8569" to-port="2" /> <edge from-layer="8569" from-port="4" to-layer="8571" to-port="2" /> <edge from-layer="8570" from-port="0" to-layer="8571" to-port="3" /> <edge from-layer="8571" from-port="4" to-layer="8578" to-port="0" /> <edge from-layer="8572" from-port="0" to-layer="8578" to-port="1" /> <edge from-layer="8573" from-port="0" to-layer="8576" to-port="0" /> <edge from-layer="8574" from-port="0" to-layer="8576" to-port="1" /> <edge from-layer="8575" from-port="0" to-layer="8576" to-port="3" /> <edge from-layer="8576" from-port="4" to-layer="8578" to-port="2" /> <edge from-layer="8577" from-port="0" to-layer="8578" to-port="3" /> <edge from-layer="8578" from-port="4" to-layer="8581" to-port="0" /> <edge from-layer="8579" from-port="0" to-layer="8581" to-port="1" /> <edge from-layer="8580" from-port="0" to-layer="8581" to-port="2" /> <edge from-layer="8581" from-port="3" to-layer="8582" to-port="0" /> <edge from-layer="8582" from-port="2" to-layer="8583" to-port="0" /> <edge from-layer="8582" from-port="2" to-layer="8596" to-port="0" /> <edge from-layer="8583" from-port="1" to-layer="8595" to-port="0" /> <edge from-layer="8584" from-port="0" to-layer="8586" to-port="1" /> <edge from-layer="8585" from-port="0" to-layer="8586" to-port="2" /> <edge from-layer="8586" from-port="3" to-layer="8590" to-port="0" /> <edge from-layer="8587" from-port="0" to-layer="8590" to-port="1" /> <edge from-layer="8588" from-port="0" to-layer="8590" to-port="2" /> <edge from-layer="8589" from-port="0" to-layer="8590" to-port="3" /> <edge from-layer="8590" from-port="4" to-layer="8594" to-port="2" /> <edge from-layer="8590" from-port="4" to-layer="8592" to-port="0" /> <edge from-layer="8591" from-port="0" to-layer="8592" to-port="1" /> <edge from-layer="8592" from-port="2" to-layer="8594" to-port="0" /> <edge from-layer="8593" from-port="0" to-layer="8594" to-port="1" /> <edge from-layer="8594" from-port="3" to-layer="8595" to-port="1" /> <edge from-layer="8595" from-port="2" to-layer="8596" to-port="1" /> <edge from-layer="8596" from-port="2" to-layer="8597" to-port="3" /> <edge from-layer="8597" from-port="4" to-layer="8599" to-port="0" /> <edge from-layer="8598" from-port="0" to-layer="8599" to-port="1" /> <edge from-layer="8599" from-port="2" to-layer="8601" to-port="0" /> <edge from-layer="8600" from-port="0" to-layer="8601" to-port="1" /> <edge from-layer="8601" from-port="2" to-layer="8609" to-port="0" /> <edge from-layer="8602" from-port="0" to-layer="8603" to-port="0" /> <edge from-layer="8603" from-port="1" to-layer="8606" to-port="0" /> <edge from-layer="8604" from-port="0" to-layer="8605" to-port="0" /> <edge from-layer="8605" from-port="1" to-layer="8606" to-port="1" /> <edge from-layer="8606" from-port="2" to-layer="8608" to-port="0" /> <edge from-layer="8607" from-port="0" to-layer="8608" to-port="1" /> <edge from-layer="8608" from-port="2" to-layer="8609" to-port="1" /> <edge from-layer="8609" from-port="2" to-layer="8611" to-port="0" /> <edge from-layer="8610" from-port="0" to-layer="8611" to-port="1" /> <edge from-layer="8611" from-port="2" to-layer="8612" to-port="1" /> <edge from-layer="8612" from-port="2" to-layer="8613" to-port="0" /> <edge from-layer="8613" from-port="2" to-layer="8844" to-port="1" /> <edge from-layer="8613" from-port="2" to-layer="8642" to-port="0" /> <edge from-layer="8613" from-port="2" to-layer="8615" to-port="0" /> <edge from-layer="8614" from-port="0" to-layer="8615" to-port="1" /> <edge from-layer="8615" from-port="2" to-layer="8617" to-port="0" /> <edge from-layer="8616" from-port="0" to-layer="8617" to-port="1" /> <edge from-layer="8617" from-port="2" to-layer="8619" to-port="0" /> <edge from-layer="8618" from-port="0" to-layer="8619" to-port="1" /> <edge from-layer="8619" from-port="2" to-layer="8627" to-port="0" /> <edge from-layer="8620" from-port="0" to-layer="8621" to-port="0" /> <edge from-layer="8621" from-port="1" to-layer="8624" to-port="0" /> <edge from-layer="8622" from-port="0" to-layer="8623" to-port="0" /> <edge from-layer="8623" from-port="1" to-layer="8624" to-port="1" /> <edge from-layer="8624" from-port="2" to-layer="8626" to-port="0" /> <edge from-layer="8625" from-port="0" to-layer="8626" to-port="1" /> <edge from-layer="8626" from-port="2" to-layer="8627" to-port="1" /> <edge from-layer="8627" from-port="2" to-layer="8629" to-port="0" /> <edge from-layer="8628" from-port="0" to-layer="8629" to-port="1" /> <edge from-layer="8629" from-port="2" to-layer="8630" to-port="0" /> <edge from-layer="8630" from-port="1" to-layer="8638" to-port="0" /> <edge from-layer="8631" from-port="0" to-layer="8632" to-port="0" /> <edge from-layer="8632" from-port="1" to-layer="8635" to-port="0" /> <edge from-layer="8633" from-port="0" to-layer="8634" to-port="0" /> <edge from-layer="8634" from-port="1" to-layer="8635" to-port="1" /> <edge from-layer="8635" from-port="2" to-layer="8637" to-port="0" /> <edge from-layer="8636" from-port="0" to-layer="8637" to-port="1" /> <edge from-layer="8637" from-port="2" to-layer="8638" to-port="1" /> <edge from-layer="8638" from-port="2" to-layer="8640" to-port="0" /> <edge from-layer="8639" from-port="0" to-layer="8640" to-port="1" /> <edge from-layer="8640" from-port="2" to-layer="8843" to-port="0" /> <edge from-layer="8641" from-port="0" to-layer="8642" to-port="1" /> <edge from-layer="8642" from-port="2" to-layer="8644" to-port="0" /> <edge from-layer="8643" from-port="0" to-layer="8644" to-port="1" /> <edge from-layer="8644" from-port="2" to-layer="8646" to-port="0" /> <edge from-layer="8645" from-port="0" to-layer="8646" to-port="1" /> <edge from-layer="8646" from-port="2" to-layer="8654" to-port="0" /> <edge from-layer="8647" from-port="0" to-layer="8648" to-port="0" /> <edge from-layer="8648" from-port="1" to-layer="8651" to-port="0" /> <edge from-layer="8649" from-port="0" to-layer="8650" to-port="0" /> <edge from-layer="8650" from-port="1" to-layer="8651" to-port="1" /> <edge from-layer="8651" from-port="2" to-layer="8653" to-port="0" /> <edge from-layer="8652" from-port="0" to-layer="8653" to-port="1" /> <edge from-layer="8653" from-port="2" to-layer="8654" to-port="1" /> <edge from-layer="8654" from-port="2" to-layer="8656" to-port="0" /> <edge from-layer="8655" from-port="0" to-layer="8656" to-port="1" /> <edge from-layer="8656" from-port="2" to-layer="8658" to-port="0" /> <edge from-layer="8657" from-port="0" to-layer="8658" to-port="1" /> <edge from-layer="8658" from-port="2" to-layer="8662" to-port="0" /> <edge from-layer="8658" from-port="2" to-layer="8780" to-port="0" /> <edge from-layer="8658" from-port="2" to-layer="8674" to-port="0" /> <edge from-layer="8659" from-port="0" to-layer="8662" to-port="1" /> <edge from-layer="8660" from-port="0" to-layer="8662" to-port="2" /> <edge from-layer="8661" from-port="0" to-layer="8662" to-port="3" /> <edge from-layer="8662" from-port="4" to-layer="8664" to-port="0" /> <edge from-layer="8663" from-port="0" to-layer="8664" to-port="1" /> <edge from-layer="8664" from-port="2" to-layer="8733" to-port="0" /> <edge from-layer="8664" from-port="2" to-layer="8668" to-port="0" /> <edge from-layer="8665" from-port="0" to-layer="8668" to-port="1" /> <edge from-layer="8666" from-port="0" to-layer="8668" to-port="2" /> <edge from-layer="8667" from-port="0" to-layer="8668" to-port="3" /> <edge from-layer="8668" from-port="4" to-layer="8714" to-port="0" /> <edge from-layer="8668" from-port="4" to-layer="8703" to-port="0" /> <edge from-layer="8668" from-port="4" to-layer="8700" to-port="0" /> <edge from-layer="8668" from-port="4" to-layer="8719" to-port="0" /> <edge from-layer="8669" from-port="0" to-layer="8696" to-port="0" /> <edge from-layer="8670" from-port="0" to-layer="8696" to-port="1" /> <edge from-layer="8671" from-port="0" to-layer="8674" to-port="1" /> <edge from-layer="8672" from-port="0" to-layer="8674" to-port="2" /> <edge from-layer="8673" from-port="0" to-layer="8674" to-port="3" /> <edge from-layer="8674" from-port="4" to-layer="8676" to-port="0" /> <edge from-layer="8675" from-port="0" to-layer="8676" to-port="1" /> <edge from-layer="8676" from-port="2" to-layer="8677" to-port="0" /> <edge from-layer="8676" from-port="2" to-layer="8765" to-port="0" /> <edge from-layer="8676" from-port="2" to-layer="8738" to-port="0" /> <edge from-layer="8677" from-port="1" to-layer="8679" to-port="0" /> <edge from-layer="8678" from-port="0" to-layer="8679" to-port="2" /> <edge from-layer="8679" from-port="3" to-layer="8692" to-port="0" /> <edge from-layer="8680" from-port="0" to-layer="8685" to-port="0" /> <edge from-layer="8681" from-port="0" to-layer="8684" to-port="1" /> <edge from-layer="8682" from-port="0" to-layer="8684" to-port="2" /> <edge from-layer="8683" from-port="0" to-layer="8684" to-port="3" /> <edge from-layer="8684" from-port="4" to-layer="8685" to-port="1" /> <edge from-layer="8685" from-port="2" to-layer="8686" to-port="0" /> <edge from-layer="8686" from-port="1" to-layer="8688" to-port="0" /> <edge from-layer="8687" from-port="0" to-layer="8688" to-port="2" /> <edge from-layer="8688" from-port="3" to-layer="8767" to-port="0" /> <edge from-layer="8688" from-port="3" to-layer="8689" to-port="0" /> <edge from-layer="8689" from-port="1" to-layer="8691" to-port="0" /> <edge from-layer="8690" from-port="0" to-layer="8691" to-port="2" /> <edge from-layer="8691" from-port="3" to-layer="8692" to-port="1" /> <edge from-layer="8692" from-port="2" to-layer="8694" to-port="0" /> <edge from-layer="8693" from-port="0" to-layer="8694" to-port="1" /> <edge from-layer="8694" from-port="2" to-layer="8696" to-port="2" /> <edge from-layer="8694" from-port="2" to-layer="8724" to-port="2" /> <edge from-layer="8695" from-port="0" to-layer="8696" to-port="3" /> <edge from-layer="8696" from-port="4" to-layer="8698" to-port="0" /> <edge from-layer="8697" from-port="0" to-layer="8698" to-port="2" /> <edge from-layer="8698" from-port="3" to-layer="8699" to-port="0" /> <edge from-layer="8699" from-port="2" to-layer="8700" to-port="1" /> <edge from-layer="8699" from-port="2" to-layer="8739" to-port="1" /> <edge from-layer="8700" from-port="2" to-layer="8729" to-port="0" /> <edge from-layer="8701" from-port="0" to-layer="8711" to-port="0" /> <edge from-layer="8702" from-port="0" to-layer="8711" to-port="1" /> <edge from-layer="8703" from-port="1" to-layer="8705" to-port="0" /> <edge from-layer="8704" from-port="0" to-layer="8705" to-port="2" /> <edge from-layer="8705" from-port="3" to-layer="8706" to-port="0" /> <edge from-layer="8706" from-port="2" to-layer="8707" to-port="0" /> <edge from-layer="8707" from-port="1" to-layer="8709" to-port="0" /> <edge from-layer="8708" from-port="0" to-layer="8709" to-port="1" /> <edge from-layer="8709" from-port="2" to-layer="8711" to-port="2" /> <edge from-layer="8710" from-port="0" to-layer="8711" to-port="3" /> <edge from-layer="8711" from-port="4" to-layer="8719" to-port="2" /> <edge from-layer="8711" from-port="4" to-layer="8714" to-port="1" /> <edge from-layer="8712" from-port="0" to-layer="8714" to-port="2" /> <edge from-layer="8713" from-port="0" to-layer="8714" to-port="3" /> <edge from-layer="8714" from-port="4" to-layer="8716" to-port="0" /> <edge from-layer="8715" from-port="0" to-layer="8716" to-port="1" /> <edge from-layer="8716" from-port="2" to-layer="8720" to-port="0" /> <edge from-layer="8717" from-port="0" to-layer="8719" to-port="1" /> <edge from-layer="8718" from-port="0" to-layer="8719" to-port="3" /> <edge from-layer="8719" from-port="4" to-layer="8720" to-port="1" /> <edge from-layer="8720" from-port="2" to-layer="8728" to-port="0" /> <edge from-layer="8721" from-port="0" to-layer="8724" to-port="0" /> <edge from-layer="8722" from-port="0" to-layer="8724" to-port="1" /> <edge from-layer="8723" from-port="0" to-layer="8724" to-port="3" /> <edge from-layer="8724" from-port="4" to-layer="8726" to-port="0" /> <edge from-layer="8725" from-port="0" to-layer="8726" to-port="2" /> <edge from-layer="8726" from-port="3" to-layer="8727" to-port="0" /> <edge from-layer="8727" from-port="2" to-layer="8760" to-port="1" /> <edge from-layer="8727" from-port="2" to-layer="8728" to-port="1" /> <edge from-layer="8728" from-port="2" to-layer="8729" to-port="1" /> <edge from-layer="8729" from-port="2" to-layer="8734" to-port="0" /> <edge from-layer="8730" from-port="0" to-layer="8733" to-port="1" /> <edge from-layer="8731" from-port="0" to-layer="8733" to-port="2" /> <edge from-layer="8732" from-port="0" to-layer="8733" to-port="3" /> <edge from-layer="8733" from-port="4" to-layer="8734" to-port="1" /> <edge from-layer="8734" from-port="2" to-layer="8790" to-port="0" /> <edge from-layer="8734" from-port="2" to-layer="8828" to-port="0" /> <edge from-layer="8735" from-port="0" to-layer="8738" to-port="1" /> <edge from-layer="8736" from-port="0" to-layer="8738" to-port="2" /> <edge from-layer="8737" from-port="0" to-layer="8738" to-port="3" /> <edge from-layer="8738" from-port="4" to-layer="8739" to-port="0" /> <edge from-layer="8738" from-port="4" to-layer="8742" to-port="0" /> <edge from-layer="8738" from-port="4" to-layer="8753" to-port="0" /> <edge from-layer="8738" from-port="4" to-layer="8758" to-port="0" /> <edge from-layer="8739" from-port="2" to-layer="8761" to-port="0" /> <edge from-layer="8740" from-port="0" to-layer="8750" to-port="0" /> <edge from-layer="8741" from-port="0" to-layer="8750" to-port="1" /> <edge from-layer="8742" from-port="1" to-layer="8744" to-port="0" /> <edge from-layer="8743" from-port="0" to-layer="8744" to-port="2" /> <edge from-layer="8744" from-port="3" to-layer="8745" to-port="0" /> <edge from-layer="8745" from-port="2" to-layer="8746" to-port="0" /> <edge from-layer="8746" from-port="1" to-layer="8748" to-port="0" /> <edge from-layer="8747" from-port="0" to-layer="8748" to-port="1" /> <edge from-layer="8748" from-port="2" to-layer="8750" to-port="2" /> <edge from-layer="8749" from-port="0" to-layer="8750" to-port="3" /> <edge from-layer="8750" from-port="4" to-layer="8753" to-port="1" /> <edge from-layer="8750" from-port="4" to-layer="8758" to-port="2" /> <edge from-layer="8751" from-port="0" to-layer="8753" to-port="2" /> <edge from-layer="8752" from-port="0" to-layer="8753" to-port="3" /> <edge from-layer="8753" from-port="4" to-layer="8755" to-port="0" /> <edge from-layer="8754" from-port="0" to-layer="8755" to-port="1" /> <edge from-layer="8755" from-port="2" to-layer="8759" to-port="0" /> <edge from-layer="8756" from-port="0" to-layer="8758" to-port="1" /> <edge from-layer="8757" from-port="0" to-layer="8758" to-port="3" /> <edge from-layer="8758" from-port="4" to-layer="8759" to-port="1" /> <edge from-layer="8759" from-port="2" to-layer="8760" to-port="0" /> <edge from-layer="8760" from-port="2" to-layer="8761" to-port="1" /> <edge from-layer="8761" from-port="2" to-layer="8766" to-port="0" /> <edge from-layer="8762" from-port="0" to-layer="8765" to-port="1" /> <edge from-layer="8763" from-port="0" to-layer="8765" to-port="2" /> <edge from-layer="8764" from-port="0" to-layer="8765" to-port="3" /> <edge from-layer="8765" from-port="4" to-layer="8766" to-port="1" /> <edge from-layer="8766" from-port="2" to-layer="8767" to-port="1" /> <edge from-layer="8767" from-port="2" to-layer="8787" to-port="0" /> <edge from-layer="8767" from-port="2" to-layer="10111" to-port="0" /> <edge from-layer="8767" from-port="2" to-layer="8828" to-port="1" /> <edge from-layer="8768" from-port="0" to-layer="8773" to-port="0" /> <edge from-layer="8769" from-port="0" to-layer="8772" to-port="1" /> <edge from-layer="8770" from-port="0" to-layer="8772" to-port="2" /> <edge from-layer="8771" from-port="0" to-layer="8772" to-port="3" /> <edge from-layer="8772" from-port="4" to-layer="8773" to-port="1" /> <edge from-layer="8773" from-port="2" to-layer="8774" to-port="0" /> <edge from-layer="8774" from-port="1" to-layer="8776" to-port="0" /> <edge from-layer="8775" from-port="0" to-layer="8776" to-port="2" /> <edge from-layer="8776" from-port="3" to-layer="8783" to-port="0" /> <edge from-layer="8777" from-port="0" to-layer="8780" to-port="1" /> <edge from-layer="8778" from-port="0" to-layer="8780" to-port="2" /> <edge from-layer="8779" from-port="0" to-layer="8780" to-port="3" /> <edge from-layer="8780" from-port="4" to-layer="8782" to-port="0" /> <edge from-layer="8781" from-port="0" to-layer="8782" to-port="1" /> <edge from-layer="8782" from-port="2" to-layer="8783" to-port="1" /> <edge from-layer="8783" from-port="2" to-layer="8828" to-port="2" /> <edge from-layer="8783" from-port="2" to-layer="10110" to-port="0" /> <edge from-layer="8784" from-port="0" to-layer="8802" to-port="0" /> <edge from-layer="8785" from-port="0" to-layer="8797" to-port="0" /> <edge from-layer="8786" from-port="0" to-layer="8797" to-port="1" /> <edge from-layer="8786" from-port="0" to-layer="8800" to-port="1" /> <edge from-layer="8787" from-port="1" to-layer="8789" to-port="0" /> <edge from-layer="8788" from-port="0" to-layer="8789" to-port="2" /> <edge from-layer="8789" from-port="3" to-layer="8793" to-port="0" /> <edge from-layer="8789" from-port="3" to-layer="8799" to-port="0" /> <edge from-layer="8790" from-port="1" to-layer="8792" to-port="0" /> <edge from-layer="8790" from-port="1" to-layer="8817" to-port="0" /> <edge from-layer="8791" from-port="0" to-layer="8792" to-port="2" /> <edge from-layer="8792" from-port="3" to-layer="8793" to-port="1" /> <edge from-layer="8793" from-port="2" to-layer="8795" to-port="0" /> <edge from-layer="8794" from-port="0" to-layer="8799" to-port="1" /> <edge from-layer="8794" from-port="0" to-layer="8795" to-port="1" /> <edge from-layer="8795" from-port="2" to-layer="8797" to-port="2" /> <edge from-layer="8796" from-port="0" to-layer="8800" to-port="3" /> <edge from-layer="8796" from-port="0" to-layer="8797" to-port="3" /> <edge from-layer="8797" from-port="4" to-layer="8802" to-port="1" /> <edge from-layer="8798" from-port="0" to-layer="8800" to-port="0" /> <edge from-layer="8799" from-port="2" to-layer="8800" to-port="2" /> <edge from-layer="8799" from-port="2" to-layer="8807" to-port="2" /> <edge from-layer="8800" from-port="4" to-layer="8802" to-port="2" /> <edge from-layer="8801" from-port="0" to-layer="8802" to-port="3" /> <edge from-layer="8802" from-port="4" to-layer="8809" to-port="0" /> <edge from-layer="8803" from-port="0" to-layer="8809" to-port="1" /> <edge from-layer="8804" from-port="0" to-layer="8807" to-port="0" /> <edge from-layer="8805" from-port="0" to-layer="8807" to-port="1" /> <edge from-layer="8806" from-port="0" to-layer="8807" to-port="3" /> <edge from-layer="8807" from-port="4" to-layer="8809" to-port="2" /> <edge from-layer="8808" from-port="0" to-layer="8809" to-port="3" /> <edge from-layer="8809" from-port="4" to-layer="8812" to-port="0" /> <edge from-layer="8810" from-port="0" to-layer="8812" to-port="1" /> <edge from-layer="8811" from-port="0" to-layer="8812" to-port="2" /> <edge from-layer="8812" from-port="3" to-layer="8813" to-port="0" /> <edge from-layer="8813" from-port="2" to-layer="8814" to-port="0" /> <edge from-layer="8813" from-port="2" to-layer="8827" to-port="0" /> <edge from-layer="8814" from-port="1" to-layer="8826" to-port="0" /> <edge from-layer="8815" from-port="0" to-layer="8817" to-port="1" /> <edge from-layer="8816" from-port="0" to-layer="8817" to-port="2" /> <edge from-layer="8817" from-port="3" to-layer="8821" to-port="0" /> <edge from-layer="8818" from-port="0" to-layer="8821" to-port="1" /> <edge from-layer="8819" from-port="0" to-layer="8821" to-port="2" /> <edge from-layer="8820" from-port="0" to-layer="8821" to-port="3" /> <edge from-layer="8821" from-port="4" to-layer="8825" to-port="2" /> <edge from-layer="8821" from-port="4" to-layer="8823" to-port="0" /> <edge from-layer="8822" from-port="0" to-layer="8823" to-port="1" /> <edge from-layer="8823" from-port="2" to-layer="8825" to-port="0" /> <edge from-layer="8824" from-port="0" to-layer="8825" to-port="1" /> <edge from-layer="8825" from-port="3" to-layer="8826" to-port="1" /> <edge from-layer="8826" from-port="2" to-layer="8827" to-port="1" /> <edge from-layer="8827" from-port="2" to-layer="8828" to-port="3" /> <edge from-layer="8828" from-port="4" to-layer="8830" to-port="0" /> <edge from-layer="8829" from-port="0" to-layer="8830" to-port="1" /> <edge from-layer="8830" from-port="2" to-layer="8832" to-port="0" /> <edge from-layer="8831" from-port="0" to-layer="8832" to-port="1" /> <edge from-layer="8832" from-port="2" to-layer="8840" to-port="0" /> <edge from-layer="8833" from-port="0" to-layer="8834" to-port="0" /> <edge from-layer="8834" from-port="1" to-layer="8837" to-port="0" /> <edge from-layer="8835" from-port="0" to-layer="8836" to-port="0" /> <edge from-layer="8836" from-port="1" to-layer="8837" to-port="1" /> <edge from-layer="8837" from-port="2" to-layer="8839" to-port="0" /> <edge from-layer="8838" from-port="0" to-layer="8839" to-port="1" /> <edge from-layer="8839" from-port="2" to-layer="8840" to-port="1" /> <edge from-layer="8840" from-port="2" to-layer="8842" to-port="0" /> <edge from-layer="8841" from-port="0" to-layer="8842" to-port="1" /> <edge from-layer="8842" from-port="2" to-layer="8843" to-port="1" /> <edge from-layer="8843" from-port="2" to-layer="8844" to-port="0" /> <edge from-layer="8844" from-port="2" to-layer="8873" to-port="0" /> <edge from-layer="8844" from-port="2" to-layer="8846" to-port="0" /> <edge from-layer="8844" from-port="2" to-layer="9075" to-port="1" /> <edge from-layer="8845" from-port="0" to-layer="8846" to-port="1" /> <edge from-layer="8846" from-port="2" to-layer="8848" to-port="0" /> <edge from-layer="8847" from-port="0" to-layer="8848" to-port="1" /> <edge from-layer="8848" from-port="2" to-layer="8850" to-port="0" /> <edge from-layer="8849" from-port="0" to-layer="8850" to-port="1" /> <edge from-layer="8850" from-port="2" to-layer="8858" to-port="0" /> <edge from-layer="8851" from-port="0" to-layer="8852" to-port="0" /> <edge from-layer="8852" from-port="1" to-layer="8855" to-port="0" /> <edge from-layer="8853" from-port="0" to-layer="8854" to-port="0" /> <edge from-layer="8854" from-port="1" to-layer="8855" to-port="1" /> <edge from-layer="8855" from-port="2" to-layer="8857" to-port="0" /> <edge from-layer="8856" from-port="0" to-layer="8857" to-port="1" /> <edge from-layer="8857" from-port="2" to-layer="8858" to-port="1" /> <edge from-layer="8858" from-port="2" to-layer="8860" to-port="0" /> <edge from-layer="8859" from-port="0" to-layer="8860" to-port="1" /> <edge from-layer="8860" from-port="2" to-layer="8861" to-port="0" /> <edge from-layer="8861" from-port="1" to-layer="8869" to-port="0" /> <edge from-layer="8862" from-port="0" to-layer="8863" to-port="0" /> <edge from-layer="8863" from-port="1" to-layer="8866" to-port="0" /> <edge from-layer="8864" from-port="0" to-layer="8865" to-port="0" /> <edge from-layer="8865" from-port="1" to-layer="8866" to-port="1" /> <edge from-layer="8866" from-port="2" to-layer="8868" to-port="0" /> <edge from-layer="8867" from-port="0" to-layer="8868" to-port="1" /> <edge from-layer="8868" from-port="2" to-layer="8869" to-port="1" /> <edge from-layer="8869" from-port="2" to-layer="8871" to-port="0" /> <edge from-layer="8870" from-port="0" to-layer="8871" to-port="1" /> <edge from-layer="8871" from-port="2" to-layer="9074" to-port="0" /> <edge from-layer="8872" from-port="0" to-layer="8873" to-port="1" /> <edge from-layer="8873" from-port="2" to-layer="8875" to-port="0" /> <edge from-layer="8874" from-port="0" to-layer="8875" to-port="1" /> <edge from-layer="8875" from-port="2" to-layer="8877" to-port="0" /> <edge from-layer="8876" from-port="0" to-layer="8877" to-port="1" /> <edge from-layer="8877" from-port="2" to-layer="8885" to-port="0" /> <edge from-layer="8878" from-port="0" to-layer="8879" to-port="0" /> <edge from-layer="8879" from-port="1" to-layer="8882" to-port="0" /> <edge from-layer="8880" from-port="0" to-layer="8881" to-port="0" /> <edge from-layer="8881" from-port="1" to-layer="8882" to-port="1" /> <edge from-layer="8882" from-port="2" to-layer="8884" to-port="0" /> <edge from-layer="8883" from-port="0" to-layer="8884" to-port="1" /> <edge from-layer="8884" from-port="2" to-layer="8885" to-port="1" /> <edge from-layer="8885" from-port="2" to-layer="8887" to-port="0" /> <edge from-layer="8886" from-port="0" to-layer="8887" to-port="1" /> <edge from-layer="8887" from-port="2" to-layer="8889" to-port="0" /> <edge from-layer="8888" from-port="0" to-layer="8889" to-port="1" /> <edge from-layer="8889" from-port="2" to-layer="8893" to-port="0" /> <edge from-layer="8889" from-port="2" to-layer="8905" to-port="0" /> <edge from-layer="8889" from-port="2" to-layer="9011" to-port="0" /> <edge from-layer="8890" from-port="0" to-layer="8893" to-port="1" /> <edge from-layer="8891" from-port="0" to-layer="8893" to-port="2" /> <edge from-layer="8892" from-port="0" to-layer="8893" to-port="3" /> <edge from-layer="8893" from-port="4" to-layer="8895" to-port="0" /> <edge from-layer="8894" from-port="0" to-layer="8895" to-port="1" /> <edge from-layer="8895" from-port="2" to-layer="8899" to-port="0" /> <edge from-layer="8895" from-port="2" to-layer="8964" to-port="0" /> <edge from-layer="8896" from-port="0" to-layer="8899" to-port="1" /> <edge from-layer="8897" from-port="0" to-layer="8899" to-port="2" /> <edge from-layer="8898" from-port="0" to-layer="8899" to-port="3" /> <edge from-layer="8899" from-port="4" to-layer="8931" to-port="0" /> <edge from-layer="8899" from-port="4" to-layer="8934" to-port="0" /> <edge from-layer="8899" from-port="4" to-layer="8945" to-port="0" /> <edge from-layer="8899" from-port="4" to-layer="8950" to-port="0" /> <edge from-layer="8900" from-port="0" to-layer="8927" to-port="0" /> <edge from-layer="8901" from-port="0" to-layer="8927" to-port="1" /> <edge from-layer="8902" from-port="0" to-layer="8905" to-port="1" /> <edge from-layer="8903" from-port="0" to-layer="8905" to-port="2" /> <edge from-layer="8904" from-port="0" to-layer="8905" to-port="3" /> <edge from-layer="8905" from-port="4" to-layer="8907" to-port="0" /> <edge from-layer="8906" from-port="0" to-layer="8907" to-port="1" /> <edge from-layer="8907" from-port="2" to-layer="8996" to-port="0" /> <edge from-layer="8907" from-port="2" to-layer="8969" to-port="0" /> <edge from-layer="8907" from-port="2" to-layer="8908" to-port="0" /> <edge from-layer="8908" from-port="1" to-layer="8910" to-port="0" /> <edge from-layer="8909" from-port="0" to-layer="8910" to-port="2" /> <edge from-layer="8910" from-port="3" to-layer="8923" to-port="0" /> <edge from-layer="8911" from-port="0" to-layer="8916" to-port="0" /> <edge from-layer="8912" from-port="0" to-layer="8915" to-port="1" /> <edge from-layer="8913" from-port="0" to-layer="8915" to-port="2" /> <edge from-layer="8914" from-port="0" to-layer="8915" to-port="3" /> <edge from-layer="8915" from-port="4" to-layer="8916" to-port="1" /> <edge from-layer="8916" from-port="2" to-layer="8917" to-port="0" /> <edge from-layer="8917" from-port="1" to-layer="8919" to-port="0" /> <edge from-layer="8918" from-port="0" to-layer="8919" to-port="2" /> <edge from-layer="8919" from-port="3" to-layer="8920" to-port="0" /> <edge from-layer="8919" from-port="3" to-layer="8998" to-port="0" /> <edge from-layer="8920" from-port="1" to-layer="8922" to-port="0" /> <edge from-layer="8921" from-port="0" to-layer="8922" to-port="2" /> <edge from-layer="8922" from-port="3" to-layer="8923" to-port="1" /> <edge from-layer="8923" from-port="2" to-layer="8925" to-port="0" /> <edge from-layer="8924" from-port="0" to-layer="8925" to-port="1" /> <edge from-layer="8925" from-port="2" to-layer="8927" to-port="2" /> <edge from-layer="8925" from-port="2" to-layer="8955" to-port="2" /> <edge from-layer="8926" from-port="0" to-layer="8927" to-port="3" /> <edge from-layer="8927" from-port="4" to-layer="8929" to-port="0" /> <edge from-layer="8928" from-port="0" to-layer="8929" to-port="2" /> <edge from-layer="8929" from-port="3" to-layer="8930" to-port="0" /> <edge from-layer="8930" from-port="2" to-layer="8931" to-port="1" /> <edge from-layer="8930" from-port="2" to-layer="8970" to-port="1" /> <edge from-layer="8931" from-port="2" to-layer="8960" to-port="0" /> <edge from-layer="8932" from-port="0" to-layer="8942" to-port="0" /> <edge from-layer="8933" from-port="0" to-layer="8942" to-port="1" /> <edge from-layer="8934" from-port="1" to-layer="8936" to-port="0" /> <edge from-layer="8935" from-port="0" to-layer="8936" to-port="2" /> <edge from-layer="8936" from-port="3" to-layer="8937" to-port="0" /> <edge from-layer="8937" from-port="2" to-layer="8938" to-port="0" /> <edge from-layer="8938" from-port="1" to-layer="8940" to-port="0" /> <edge from-layer="8939" from-port="0" to-layer="8940" to-port="1" /> <edge from-layer="8940" from-port="2" to-layer="8942" to-port="2" /> <edge from-layer="8941" from-port="0" to-layer="8942" to-port="3" /> <edge from-layer="8942" from-port="4" to-layer="8945" to-port="1" /> <edge from-layer="8942" from-port="4" to-layer="8950" to-port="2" /> <edge from-layer="8943" from-port="0" to-layer="8945" to-port="2" /> <edge from-layer="8944" from-port="0" to-layer="8945" to-port="3" /> <edge from-layer="8945" from-port="4" to-layer="8947" to-port="0" /> <edge from-layer="8946" from-port="0" to-layer="8947" to-port="1" /> <edge from-layer="8947" from-port="2" to-layer="8951" to-port="0" /> <edge from-layer="8948" from-port="0" to-layer="8950" to-port="1" /> <edge from-layer="8949" from-port="0" to-layer="8950" to-port="3" /> <edge from-layer="8950" from-port="4" to-layer="8951" to-port="1" /> <edge from-layer="8951" from-port="2" to-layer="8959" to-port="0" /> <edge from-layer="8952" from-port="0" to-layer="8955" to-port="0" /> <edge from-layer="8953" from-port="0" to-layer="8955" to-port="1" /> <edge from-layer="8954" from-port="0" to-layer="8955" to-port="3" /> <edge from-layer="8955" from-port="4" to-layer="8957" to-port="0" /> <edge from-layer="8956" from-port="0" to-layer="8957" to-port="2" /> <edge from-layer="8957" from-port="3" to-layer="8958" to-port="0" /> <edge from-layer="8958" from-port="2" to-layer="8991" to-port="1" /> <edge from-layer="8958" from-port="2" to-layer="8959" to-port="1" /> <edge from-layer="8959" from-port="2" to-layer="8960" to-port="1" /> <edge from-layer="8960" from-port="2" to-layer="8965" to-port="0" /> <edge from-layer="8961" from-port="0" to-layer="8964" to-port="1" /> <edge from-layer="8962" from-port="0" to-layer="8964" to-port="2" /> <edge from-layer="8963" from-port="0" to-layer="8964" to-port="3" /> <edge from-layer="8964" from-port="4" to-layer="8965" to-port="1" /> <edge from-layer="8965" from-port="2" to-layer="9059" to-port="0" /> <edge from-layer="8965" from-port="2" to-layer="9021" to-port="0" /> <edge from-layer="8966" from-port="0" to-layer="8969" to-port="1" /> <edge from-layer="8967" from-port="0" to-layer="8969" to-port="2" /> <edge from-layer="8968" from-port="0" to-layer="8969" to-port="3" /> <edge from-layer="8969" from-port="4" to-layer="8970" to-port="0" /> <edge from-layer="8969" from-port="4" to-layer="8973" to-port="0" /> <edge from-layer="8969" from-port="4" to-layer="8989" to-port="0" /> <edge from-layer="8969" from-port="4" to-layer="8984" to-port="0" /> <edge from-layer="8970" from-port="2" to-layer="8992" to-port="0" /> <edge from-layer="8971" from-port="0" to-layer="8981" to-port="0" /> <edge from-layer="8972" from-port="0" to-layer="8981" to-port="1" /> <edge from-layer="8973" from-port="1" to-layer="8975" to-port="0" /> <edge from-layer="8974" from-port="0" to-layer="8975" to-port="2" /> <edge from-layer="8975" from-port="3" to-layer="8976" to-port="0" /> <edge from-layer="8976" from-port="2" to-layer="8977" to-port="0" /> <edge from-layer="8977" from-port="1" to-layer="8979" to-port="0" /> <edge from-layer="8978" from-port="0" to-layer="8979" to-port="1" /> <edge from-layer="8979" from-port="2" to-layer="8981" to-port="2" /> <edge from-layer="8980" from-port="0" to-layer="8981" to-port="3" /> <edge from-layer="8981" from-port="4" to-layer="8984" to-port="1" /> <edge from-layer="8981" from-port="4" to-layer="8989" to-port="2" /> <edge from-layer="8982" from-port="0" to-layer="8984" to-port="2" /> <edge from-layer="8983" from-port="0" to-layer="8984" to-port="3" /> <edge from-layer="8984" from-port="4" to-layer="8986" to-port="0" /> <edge from-layer="8985" from-port="0" to-layer="8986" to-port="1" /> <edge from-layer="8986" from-port="2" to-layer="8990" to-port="0" /> <edge from-layer="8987" from-port="0" to-layer="8989" to-port="1" /> <edge from-layer="8988" from-port="0" to-layer="8989" to-port="3" /> <edge from-layer="8989" from-port="4" to-layer="8990" to-port="1" /> <edge from-layer="8990" from-port="2" to-layer="8991" to-port="0" /> <edge from-layer="8991" from-port="2" to-layer="8992" to-port="1" /> <edge from-layer="8992" from-port="2" to-layer="8997" to-port="0" /> <edge from-layer="8993" from-port="0" to-layer="8996" to-port="1" /> <edge from-layer="8994" from-port="0" to-layer="8996" to-port="2" /> <edge from-layer="8995" from-port="0" to-layer="8996" to-port="3" /> <edge from-layer="8996" from-port="4" to-layer="8997" to-port="1" /> <edge from-layer="8997" from-port="2" to-layer="8998" to-port="1" /> <edge from-layer="8998" from-port="2" to-layer="10109" to-port="0" /> <edge from-layer="8998" from-port="2" to-layer="9059" to-port="1" /> <edge from-layer="8998" from-port="2" to-layer="9018" to-port="0" /> <edge from-layer="8999" from-port="0" to-layer="9004" to-port="0" /> <edge from-layer="9000" from-port="0" to-layer="9003" to-port="1" /> <edge from-layer="9001" from-port="0" to-layer="9003" to-port="2" /> <edge from-layer="9002" from-port="0" to-layer="9003" to-port="3" /> <edge from-layer="9003" from-port="4" to-layer="9004" to-port="1" /> <edge from-layer="9004" from-port="2" to-layer="9005" to-port="0" /> <edge from-layer="9005" from-port="1" to-layer="9007" to-port="0" /> <edge from-layer="9006" from-port="0" to-layer="9007" to-port="2" /> <edge from-layer="9007" from-port="3" to-layer="9014" to-port="0" /> <edge from-layer="9008" from-port="0" to-layer="9011" to-port="1" /> <edge from-layer="9009" from-port="0" to-layer="9011" to-port="2" /> <edge from-layer="9010" from-port="0" to-layer="9011" to-port="3" /> <edge from-layer="9011" from-port="4" to-layer="9013" to-port="0" /> <edge from-layer="9012" from-port="0" to-layer="9013" to-port="1" /> <edge from-layer="9013" from-port="2" to-layer="9014" to-port="1" /> <edge from-layer="9014" from-port="2" to-layer="9059" to-port="2" /> <edge from-layer="9014" from-port="2" to-layer="10108" to-port="0" /> <edge from-layer="9015" from-port="0" to-layer="9033" to-port="0" /> <edge from-layer="9016" from-port="0" to-layer="9028" to-port="0" /> <edge from-layer="9017" from-port="0" to-layer="9028" to-port="1" /> <edge from-layer="9017" from-port="0" to-layer="9031" to-port="1" /> <edge from-layer="9018" from-port="1" to-layer="9020" to-port="0" /> <edge from-layer="9019" from-port="0" to-layer="9020" to-port="2" /> <edge from-layer="9020" from-port="3" to-layer="9024" to-port="0" /> <edge from-layer="9020" from-port="3" to-layer="9030" to-port="0" /> <edge from-layer="9021" from-port="1" to-layer="9048" to-port="0" /> <edge from-layer="9021" from-port="1" to-layer="9023" to-port="0" /> <edge from-layer="9022" from-port="0" to-layer="9023" to-port="2" /> <edge from-layer="9023" from-port="3" to-layer="9024" to-port="1" /> <edge from-layer="9024" from-port="2" to-layer="9026" to-port="0" /> <edge from-layer="9025" from-port="0" to-layer="9030" to-port="1" /> <edge from-layer="9025" from-port="0" to-layer="9026" to-port="1" /> <edge from-layer="9026" from-port="2" to-layer="9028" to-port="2" /> <edge from-layer="9027" from-port="0" to-layer="9028" to-port="3" /> <edge from-layer="9027" from-port="0" to-layer="9031" to-port="3" /> <edge from-layer="9028" from-port="4" to-layer="9033" to-port="1" /> <edge from-layer="9029" from-port="0" to-layer="9031" to-port="0" /> <edge from-layer="9030" from-port="2" to-layer="9031" to-port="2" /> <edge from-layer="9030" from-port="2" to-layer="9038" to-port="2" /> <edge from-layer="9031" from-port="4" to-layer="9033" to-port="2" /> <edge from-layer="9032" from-port="0" to-layer="9033" to-port="3" /> <edge from-layer="9033" from-port="4" to-layer="9040" to-port="0" /> <edge from-layer="9034" from-port="0" to-layer="9040" to-port="1" /> <edge from-layer="9035" from-port="0" to-layer="9038" to-port="0" /> <edge from-layer="9036" from-port="0" to-layer="9038" to-port="1" /> <edge from-layer="9037" from-port="0" to-layer="9038" to-port="3" /> <edge from-layer="9038" from-port="4" to-layer="9040" to-port="2" /> <edge from-layer="9039" from-port="0" to-layer="9040" to-port="3" /> <edge from-layer="9040" from-port="4" to-layer="9043" to-port="0" /> <edge from-layer="9041" from-port="0" to-layer="9043" to-port="1" /> <edge from-layer="9042" from-port="0" to-layer="9043" to-port="2" /> <edge from-layer="9043" from-port="3" to-layer="9044" to-port="0" /> <edge from-layer="9044" from-port="2" to-layer="9058" to-port="0" /> <edge from-layer="9044" from-port="2" to-layer="9045" to-port="0" /> <edge from-layer="9045" from-port="1" to-layer="9057" to-port="0" /> <edge from-layer="9046" from-port="0" to-layer="9048" to-port="1" /> <edge from-layer="9047" from-port="0" to-layer="9048" to-port="2" /> <edge from-layer="9048" from-port="3" to-layer="9052" to-port="0" /> <edge from-layer="9049" from-port="0" to-layer="9052" to-port="1" /> <edge from-layer="9050" from-port="0" to-layer="9052" to-port="2" /> <edge from-layer="9051" from-port="0" to-layer="9052" to-port="3" /> <edge from-layer="9052" from-port="4" to-layer="9056" to-port="2" /> <edge from-layer="9052" from-port="4" to-layer="9054" to-port="0" /> <edge from-layer="9053" from-port="0" to-layer="9054" to-port="1" /> <edge from-layer="9054" from-port="2" to-layer="9056" to-port="0" /> <edge from-layer="9055" from-port="0" to-layer="9056" to-port="1" /> <edge from-layer="9056" from-port="3" to-layer="9057" to-port="1" /> <edge from-layer="9057" from-port="2" to-layer="9058" to-port="1" /> <edge from-layer="9058" from-port="2" to-layer="9059" to-port="3" /> <edge from-layer="9059" from-port="4" to-layer="9061" to-port="0" /> <edge from-layer="9060" from-port="0" to-layer="9061" to-port="1" /> <edge from-layer="9061" from-port="2" to-layer="9063" to-port="0" /> <edge from-layer="9062" from-port="0" to-layer="9063" to-port="1" /> <edge from-layer="9063" from-port="2" to-layer="9071" to-port="0" /> <edge from-layer="9064" from-port="0" to-layer="9065" to-port="0" /> <edge from-layer="9065" from-port="1" to-layer="9068" to-port="0" /> <edge from-layer="9066" from-port="0" to-layer="9067" to-port="0" /> <edge from-layer="9067" from-port="1" to-layer="9068" to-port="1" /> <edge from-layer="9068" from-port="2" to-layer="9070" to-port="0" /> <edge from-layer="9069" from-port="0" to-layer="9070" to-port="1" /> <edge from-layer="9070" from-port="2" to-layer="9071" to-port="1" /> <edge from-layer="9071" from-port="2" to-layer="9073" to-port="0" /> <edge from-layer="9072" from-port="0" to-layer="9073" to-port="1" /> <edge from-layer="9073" from-port="2" to-layer="9074" to-port="1" /> <edge from-layer="9074" from-port="2" to-layer="9075" to-port="0" /> <edge from-layer="9075" from-port="2" to-layer="9306" to-port="1" /> <edge from-layer="9075" from-port="2" to-layer="9077" to-port="0" /> <edge from-layer="9075" from-port="2" to-layer="9104" to-port="0" /> <edge from-layer="9076" from-port="0" to-layer="9077" to-port="1" /> <edge from-layer="9077" from-port="2" to-layer="9079" to-port="0" /> <edge from-layer="9078" from-port="0" to-layer="9079" to-port="1" /> <edge from-layer="9079" from-port="2" to-layer="9081" to-port="0" /> <edge from-layer="9080" from-port="0" to-layer="9081" to-port="1" /> <edge from-layer="9081" from-port="2" to-layer="9089" to-port="0" /> <edge from-layer="9082" from-port="0" to-layer="9083" to-port="0" /> <edge from-layer="9083" from-port="1" to-layer="9086" to-port="0" /> <edge from-layer="9084" from-port="0" to-layer="9085" to-port="0" /> <edge from-layer="9085" from-port="1" to-layer="9086" to-port="1" /> <edge from-layer="9086" from-port="2" to-layer="9088" to-port="0" /> <edge from-layer="9087" from-port="0" to-layer="9088" to-port="1" /> <edge from-layer="9088" from-port="2" to-layer="9089" to-port="1" /> <edge from-layer="9089" from-port="2" to-layer="9091" to-port="0" /> <edge from-layer="9090" from-port="0" to-layer="9091" to-port="1" /> <edge from-layer="9091" from-port="2" to-layer="9092" to-port="0" /> <edge from-layer="9092" from-port="1" to-layer="9100" to-port="0" /> <edge from-layer="9093" from-port="0" to-layer="9094" to-port="0" /> <edge from-layer="9094" from-port="1" to-layer="9097" to-port="0" /> <edge from-layer="9095" from-port="0" to-layer="9096" to-port="0" /> <edge from-layer="9096" from-port="1" to-layer="9097" to-port="1" /> <edge from-layer="9097" from-port="2" to-layer="9099" to-port="0" /> <edge from-layer="9098" from-port="0" to-layer="9099" to-port="1" /> <edge from-layer="9099" from-port="2" to-layer="9100" to-port="1" /> <edge from-layer="9100" from-port="2" to-layer="9102" to-port="0" /> <edge from-layer="9101" from-port="0" to-layer="9102" to-port="1" /> <edge from-layer="9102" from-port="2" to-layer="9305" to-port="0" /> <edge from-layer="9103" from-port="0" to-layer="9104" to-port="1" /> <edge from-layer="9104" from-port="2" to-layer="9106" to-port="0" /> <edge from-layer="9105" from-port="0" to-layer="9106" to-port="1" /> <edge from-layer="9106" from-port="2" to-layer="9108" to-port="0" /> <edge from-layer="9107" from-port="0" to-layer="9108" to-port="1" /> <edge from-layer="9108" from-port="2" to-layer="9116" to-port="0" /> <edge from-layer="9109" from-port="0" to-layer="9110" to-port="0" /> <edge from-layer="9110" from-port="1" to-layer="9113" to-port="0" /> <edge from-layer="9111" from-port="0" to-layer="9112" to-port="0" /> <edge from-layer="9112" from-port="1" to-layer="9113" to-port="1" /> <edge from-layer="9113" from-port="2" to-layer="9115" to-port="0" /> <edge from-layer="9114" from-port="0" to-layer="9115" to-port="1" /> <edge from-layer="9115" from-port="2" to-layer="9116" to-port="1" /> <edge from-layer="9116" from-port="2" to-layer="9118" to-port="0" /> <edge from-layer="9117" from-port="0" to-layer="9118" to-port="1" /> <edge from-layer="9118" from-port="2" to-layer="9120" to-port="0" /> <edge from-layer="9119" from-port="0" to-layer="9120" to-port="1" /> <edge from-layer="9120" from-port="2" to-layer="9242" to-port="0" /> <edge from-layer="9120" from-port="2" to-layer="9124" to-port="0" /> <edge from-layer="9120" from-port="2" to-layer="9136" to-port="0" /> <edge from-layer="9121" from-port="0" to-layer="9124" to-port="1" /> <edge from-layer="9122" from-port="0" to-layer="9124" to-port="2" /> <edge from-layer="9123" from-port="0" to-layer="9124" to-port="3" /> <edge from-layer="9124" from-port="4" to-layer="9126" to-port="0" /> <edge from-layer="9125" from-port="0" to-layer="9126" to-port="1" /> <edge from-layer="9126" from-port="2" to-layer="9195" to-port="0" /> <edge from-layer="9126" from-port="2" to-layer="9130" to-port="0" /> <edge from-layer="9127" from-port="0" to-layer="9130" to-port="1" /> <edge from-layer="9128" from-port="0" to-layer="9130" to-port="2" /> <edge from-layer="9129" from-port="0" to-layer="9130" to-port="3" /> <edge from-layer="9130" from-port="4" to-layer="9181" to-port="0" /> <edge from-layer="9130" from-port="4" to-layer="9176" to-port="0" /> <edge from-layer="9130" from-port="4" to-layer="9165" to-port="0" /> <edge from-layer="9130" from-port="4" to-layer="9162" to-port="0" /> <edge from-layer="9131" from-port="0" to-layer="9158" to-port="0" /> <edge from-layer="9132" from-port="0" to-layer="9158" to-port="1" /> <edge from-layer="9133" from-port="0" to-layer="9136" to-port="1" /> <edge from-layer="9134" from-port="0" to-layer="9136" to-port="2" /> <edge from-layer="9135" from-port="0" to-layer="9136" to-port="3" /> <edge from-layer="9136" from-port="4" to-layer="9138" to-port="0" /> <edge from-layer="9137" from-port="0" to-layer="9138" to-port="1" /> <edge from-layer="9138" from-port="2" to-layer="9139" to-port="0" /> <edge from-layer="9138" from-port="2" to-layer="9200" to-port="0" /> <edge from-layer="9138" from-port="2" to-layer="9227" to-port="0" /> <edge from-layer="9139" from-port="1" to-layer="9141" to-port="0" /> <edge from-layer="9140" from-port="0" to-layer="9141" to-port="2" /> <edge from-layer="9141" from-port="3" to-layer="9154" to-port="0" /> <edge from-layer="9142" from-port="0" to-layer="9147" to-port="0" /> <edge from-layer="9143" from-port="0" to-layer="9146" to-port="1" /> <edge from-layer="9144" from-port="0" to-layer="9146" to-port="2" /> <edge from-layer="9145" from-port="0" to-layer="9146" to-port="3" /> <edge from-layer="9146" from-port="4" to-layer="9147" to-port="1" /> <edge from-layer="9147" from-port="2" to-layer="9148" to-port="0" /> <edge from-layer="9148" from-port="1" to-layer="9150" to-port="0" /> <edge from-layer="9149" from-port="0" to-layer="9150" to-port="2" /> <edge from-layer="9150" from-port="3" to-layer="9229" to-port="0" /> <edge from-layer="9150" from-port="3" to-layer="9151" to-port="0" /> <edge from-layer="9151" from-port="1" to-layer="9153" to-port="0" /> <edge from-layer="9152" from-port="0" to-layer="9153" to-port="2" /> <edge from-layer="9153" from-port="3" to-layer="9154" to-port="1" /> <edge from-layer="9154" from-port="2" to-layer="9156" to-port="0" /> <edge from-layer="9155" from-port="0" to-layer="9156" to-port="1" /> <edge from-layer="9156" from-port="2" to-layer="9158" to-port="2" /> <edge from-layer="9156" from-port="2" to-layer="9186" to-port="2" /> <edge from-layer="9157" from-port="0" to-layer="9158" to-port="3" /> <edge from-layer="9158" from-port="4" to-layer="9160" to-port="0" /> <edge from-layer="9159" from-port="0" to-layer="9160" to-port="2" /> <edge from-layer="9160" from-port="3" to-layer="9161" to-port="0" /> <edge from-layer="9161" from-port="2" to-layer="9162" to-port="1" /> <edge from-layer="9161" from-port="2" to-layer="9201" to-port="1" /> <edge from-layer="9162" from-port="2" to-layer="9191" to-port="0" /> <edge from-layer="9163" from-port="0" to-layer="9173" to-port="0" /> <edge from-layer="9164" from-port="0" to-layer="9173" to-port="1" /> <edge from-layer="9165" from-port="1" to-layer="9167" to-port="0" /> <edge from-layer="9166" from-port="0" to-layer="9167" to-port="2" /> <edge from-layer="9167" from-port="3" to-layer="9168" to-port="0" /> <edge from-layer="9168" from-port="2" to-layer="9169" to-port="0" /> <edge from-layer="9169" from-port="1" to-layer="9171" to-port="0" /> <edge from-layer="9170" from-port="0" to-layer="9171" to-port="1" /> <edge from-layer="9171" from-port="2" to-layer="9173" to-port="2" /> <edge from-layer="9172" from-port="0" to-layer="9173" to-port="3" /> <edge from-layer="9173" from-port="4" to-layer="9181" to-port="2" /> <edge from-layer="9173" from-port="4" to-layer="9176" to-port="1" /> <edge from-layer="9174" from-port="0" to-layer="9176" to-port="2" /> <edge from-layer="9175" from-port="0" to-layer="9176" to-port="3" /> <edge from-layer="9176" from-port="4" to-layer="9178" to-port="0" /> <edge from-layer="9177" from-port="0" to-layer="9178" to-port="1" /> <edge from-layer="9178" from-port="2" to-layer="9182" to-port="0" /> <edge from-layer="9179" from-port="0" to-layer="9181" to-port="1" /> <edge from-layer="9180" from-port="0" to-layer="9181" to-port="3" /> <edge from-layer="9181" from-port="4" to-layer="9182" to-port="1" /> <edge from-layer="9182" from-port="2" to-layer="9190" to-port="0" /> <edge from-layer="9183" from-port="0" to-layer="9186" to-port="0" /> <edge from-layer="9184" from-port="0" to-layer="9186" to-port="1" /> <edge from-layer="9185" from-port="0" to-layer="9186" to-port="3" /> <edge from-layer="9186" from-port="4" to-layer="9188" to-port="0" /> <edge from-layer="9187" from-port="0" to-layer="9188" to-port="2" /> <edge from-layer="9188" from-port="3" to-layer="9189" to-port="0" /> <edge from-layer="9189" from-port="2" to-layer="9222" to-port="1" /> <edge from-layer="9189" from-port="2" to-layer="9190" to-port="1" /> <edge from-layer="9190" from-port="2" to-layer="9191" to-port="1" /> <edge from-layer="9191" from-port="2" to-layer="9196" to-port="0" /> <edge from-layer="9192" from-port="0" to-layer="9195" to-port="1" /> <edge from-layer="9193" from-port="0" to-layer="9195" to-port="2" /> <edge from-layer="9194" from-port="0" to-layer="9195" to-port="3" /> <edge from-layer="9195" from-port="4" to-layer="9196" to-port="1" /> <edge from-layer="9196" from-port="2" to-layer="9252" to-port="0" /> <edge from-layer="9196" from-port="2" to-layer="9290" to-port="0" /> <edge from-layer="9197" from-port="0" to-layer="9200" to-port="1" /> <edge from-layer="9198" from-port="0" to-layer="9200" to-port="2" /> <edge from-layer="9199" from-port="0" to-layer="9200" to-port="3" /> <edge from-layer="9200" from-port="4" to-layer="9215" to-port="0" /> <edge from-layer="9200" from-port="4" to-layer="9220" to-port="0" /> <edge from-layer="9200" from-port="4" to-layer="9204" to-port="0" /> <edge from-layer="9200" from-port="4" to-layer="9201" to-port="0" /> <edge from-layer="9201" from-port="2" to-layer="9223" to-port="0" /> <edge from-layer="9202" from-port="0" to-layer="9212" to-port="0" /> <edge from-layer="9203" from-port="0" to-layer="9212" to-port="1" /> <edge from-layer="9204" from-port="1" to-layer="9206" to-port="0" /> <edge from-layer="9205" from-port="0" to-layer="9206" to-port="2" /> <edge from-layer="9206" from-port="3" to-layer="9207" to-port="0" /> <edge from-layer="9207" from-port="2" to-layer="9208" to-port="0" /> <edge from-layer="9208" from-port="1" to-layer="9210" to-port="0" /> <edge from-layer="9209" from-port="0" to-layer="9210" to-port="1" /> <edge from-layer="9210" from-port="2" to-layer="9212" to-port="2" /> <edge from-layer="9211" from-port="0" to-layer="9212" to-port="3" /> <edge from-layer="9212" from-port="4" to-layer="9215" to-port="1" /> <edge from-layer="9212" from-port="4" to-layer="9220" to-port="2" /> <edge from-layer="9213" from-port="0" to-layer="9215" to-port="2" /> <edge from-layer="9214" from-port="0" to-layer="9215" to-port="3" /> <edge from-layer="9215" from-port="4" to-layer="9217" to-port="0" /> <edge from-layer="9216" from-port="0" to-layer="9217" to-port="1" /> <edge from-layer="9217" from-port="2" to-layer="9221" to-port="0" /> <edge from-layer="9218" from-port="0" to-layer="9220" to-port="1" /> <edge from-layer="9219" from-port="0" to-layer="9220" to-port="3" /> <edge from-layer="9220" from-port="4" to-layer="9221" to-port="1" /> <edge from-layer="9221" from-port="2" to-layer="9222" to-port="0" /> <edge from-layer="9222" from-port="2" to-layer="9223" to-port="1" /> <edge from-layer="9223" from-port="2" to-layer="9228" to-port="0" /> <edge from-layer="9224" from-port="0" to-layer="9227" to-port="1" /> <edge from-layer="9225" from-port="0" to-layer="9227" to-port="2" /> <edge from-layer="9226" from-port="0" to-layer="9227" to-port="3" /> <edge from-layer="9227" from-port="4" to-layer="9228" to-port="1" /> <edge from-layer="9228" from-port="2" to-layer="9229" to-port="1" /> <edge from-layer="9229" from-port="2" to-layer="10107" to-port="0" /> <edge from-layer="9229" from-port="2" to-layer="9290" to-port="1" /> <edge from-layer="9229" from-port="2" to-layer="9249" to-port="0" /> <edge from-layer="9230" from-port="0" to-layer="9235" to-port="0" /> <edge from-layer="9231" from-port="0" to-layer="9234" to-port="1" /> <edge from-layer="9232" from-port="0" to-layer="9234" to-port="2" /> <edge from-layer="9233" from-port="0" to-layer="9234" to-port="3" /> <edge from-layer="9234" from-port="4" to-layer="9235" to-port="1" /> <edge from-layer="9235" from-port="2" to-layer="9236" to-port="0" /> <edge from-layer="9236" from-port="1" to-layer="9238" to-port="0" /> <edge from-layer="9237" from-port="0" to-layer="9238" to-port="2" /> <edge from-layer="9238" from-port="3" to-layer="9245" to-port="0" /> <edge from-layer="9239" from-port="0" to-layer="9242" to-port="1" /> <edge from-layer="9240" from-port="0" to-layer="9242" to-port="2" /> <edge from-layer="9241" from-port="0" to-layer="9242" to-port="3" /> <edge from-layer="9242" from-port="4" to-layer="9244" to-port="0" /> <edge from-layer="9243" from-port="0" to-layer="9244" to-port="1" /> <edge from-layer="9244" from-port="2" to-layer="9245" to-port="1" /> <edge from-layer="9245" from-port="2" to-layer="9290" to-port="2" /> <edge from-layer="9245" from-port="2" to-layer="10106" to-port="0" /> <edge from-layer="9246" from-port="0" to-layer="9264" to-port="0" /> <edge from-layer="9247" from-port="0" to-layer="9259" to-port="0" /> <edge from-layer="9248" from-port="0" to-layer="9259" to-port="1" /> <edge from-layer="9248" from-port="0" to-layer="9262" to-port="1" /> <edge from-layer="9249" from-port="1" to-layer="9251" to-port="0" /> <edge from-layer="9250" from-port="0" to-layer="9251" to-port="2" /> <edge from-layer="9251" from-port="3" to-layer="9255" to-port="0" /> <edge from-layer="9251" from-port="3" to-layer="9261" to-port="0" /> <edge from-layer="9252" from-port="1" to-layer="9254" to-port="0" /> <edge from-layer="9252" from-port="1" to-layer="9279" to-port="0" /> <edge from-layer="9253" from-port="0" to-layer="9254" to-port="2" /> <edge from-layer="9254" from-port="3" to-layer="9255" to-port="1" /> <edge from-layer="9255" from-port="2" to-layer="9257" to-port="0" /> <edge from-layer="9256" from-port="0" to-layer="9257" to-port="1" /> <edge from-layer="9256" from-port="0" to-layer="9261" to-port="1" /> <edge from-layer="9257" from-port="2" to-layer="9259" to-port="2" /> <edge from-layer="9258" from-port="0" to-layer="9259" to-port="3" /> <edge from-layer="9258" from-port="0" to-layer="9262" to-port="3" /> <edge from-layer="9259" from-port="4" to-layer="9264" to-port="1" /> <edge from-layer="9260" from-port="0" to-layer="9262" to-port="0" /> <edge from-layer="9261" from-port="2" to-layer="9262" to-port="2" /> <edge from-layer="9261" from-port="2" to-layer="9269" to-port="2" /> <edge from-layer="9262" from-port="4" to-layer="9264" to-port="2" /> <edge from-layer="9263" from-port="0" to-layer="9264" to-port="3" /> <edge from-layer="9264" from-port="4" to-layer="9271" to-port="0" /> <edge from-layer="9265" from-port="0" to-layer="9271" to-port="1" /> <edge from-layer="9266" from-port="0" to-layer="9269" to-port="0" /> <edge from-layer="9267" from-port="0" to-layer="9269" to-port="1" /> <edge from-layer="9268" from-port="0" to-layer="9269" to-port="3" /> <edge from-layer="9269" from-port="4" to-layer="9271" to-port="2" /> <edge from-layer="9270" from-port="0" to-layer="9271" to-port="3" /> <edge from-layer="9271" from-port="4" to-layer="9274" to-port="0" /> <edge from-layer="9272" from-port="0" to-layer="9274" to-port="1" /> <edge from-layer="9273" from-port="0" to-layer="9274" to-port="2" /> <edge from-layer="9274" from-port="3" to-layer="9275" to-port="0" /> <edge from-layer="9275" from-port="2" to-layer="9289" to-port="0" /> <edge from-layer="9275" from-port="2" to-layer="9276" to-port="0" /> <edge from-layer="9276" from-port="1" to-layer="9288" to-port="0" /> <edge from-layer="9277" from-port="0" to-layer="9279" to-port="1" /> <edge from-layer="9278" from-port="0" to-layer="9279" to-port="2" /> <edge from-layer="9279" from-port="3" to-layer="9283" to-port="0" /> <edge from-layer="9280" from-port="0" to-layer="9283" to-port="1" /> <edge from-layer="9281" from-port="0" to-layer="9283" to-port="2" /> <edge from-layer="9282" from-port="0" to-layer="9283" to-port="3" /> <edge from-layer="9283" from-port="4" to-layer="9285" to-port="0" /> <edge from-layer="9283" from-port="4" to-layer="9287" to-port="2" /> <edge from-layer="9284" from-port="0" to-layer="9285" to-port="1" /> <edge from-layer="9285" from-port="2" to-layer="9287" to-port="0" /> <edge from-layer="9286" from-port="0" to-layer="9287" to-port="1" /> <edge from-layer="9287" from-port="3" to-layer="9288" to-port="1" /> <edge from-layer="9288" from-port="2" to-layer="9289" to-port="1" /> <edge from-layer="9289" from-port="2" to-layer="9290" to-port="3" /> <edge from-layer="9290" from-port="4" to-layer="9292" to-port="0" /> <edge from-layer="9291" from-port="0" to-layer="9292" to-port="1" /> <edge from-layer="9292" from-port="2" to-layer="9294" to-port="0" /> <edge from-layer="9293" from-port="0" to-layer="9294" to-port="1" /> <edge from-layer="9294" from-port="2" to-layer="9302" to-port="0" /> <edge from-layer="9295" from-port="0" to-layer="9296" to-port="0" /> <edge from-layer="9296" from-port="1" to-layer="9299" to-port="0" /> <edge from-layer="9297" from-port="0" to-layer="9298" to-port="0" /> <edge from-layer="9298" from-port="1" to-layer="9299" to-port="1" /> <edge from-layer="9299" from-port="2" to-layer="9301" to-port="0" /> <edge from-layer="9300" from-port="0" to-layer="9301" to-port="1" /> <edge from-layer="9301" from-port="2" to-layer="9302" to-port="1" /> <edge from-layer="9302" from-port="2" to-layer="9304" to-port="0" /> <edge from-layer="9303" from-port="0" to-layer="9304" to-port="1" /> <edge from-layer="9304" from-port="2" to-layer="9305" to-port="1" /> <edge from-layer="9305" from-port="2" to-layer="9306" to-port="0" /> <edge from-layer="9306" from-port="2" to-layer="9335" to-port="0" /> <edge from-layer="9306" from-port="2" to-layer="9537" to-port="1" /> <edge from-layer="9306" from-port="2" to-layer="9308" to-port="0" /> <edge from-layer="9307" from-port="0" to-layer="9308" to-port="1" /> <edge from-layer="9308" from-port="2" to-layer="9310" to-port="0" /> <edge from-layer="9309" from-port="0" to-layer="9310" to-port="1" /> <edge from-layer="9310" from-port="2" to-layer="9312" to-port="0" /> <edge from-layer="9311" from-port="0" to-layer="9312" to-port="1" /> <edge from-layer="9312" from-port="2" to-layer="9320" to-port="0" /> <edge from-layer="9313" from-port="0" to-layer="9314" to-port="0" /> <edge from-layer="9314" from-port="1" to-layer="9317" to-port="0" /> <edge from-layer="9315" from-port="0" to-layer="9316" to-port="0" /> <edge from-layer="9316" from-port="1" to-layer="9317" to-port="1" /> <edge from-layer="9317" from-port="2" to-layer="9319" to-port="0" /> <edge from-layer="9318" from-port="0" to-layer="9319" to-port="1" /> <edge from-layer="9319" from-port="2" to-layer="9320" to-port="1" /> <edge from-layer="9320" from-port="2" to-layer="9322" to-port="0" /> <edge from-layer="9321" from-port="0" to-layer="9322" to-port="1" /> <edge from-layer="9322" from-port="2" to-layer="9323" to-port="0" /> <edge from-layer="9323" from-port="1" to-layer="9331" to-port="0" /> <edge from-layer="9324" from-port="0" to-layer="9325" to-port="0" /> <edge from-layer="9325" from-port="1" to-layer="9328" to-port="0" /> <edge from-layer="9326" from-port="0" to-layer="9327" to-port="0" /> <edge from-layer="9327" from-port="1" to-layer="9328" to-port="1" /> <edge from-layer="9328" from-port="2" to-layer="9330" to-port="0" /> <edge from-layer="9329" from-port="0" to-layer="9330" to-port="1" /> <edge from-layer="9330" from-port="2" to-layer="9331" to-port="1" /> <edge from-layer="9331" from-port="2" to-layer="9333" to-port="0" /> <edge from-layer="9332" from-port="0" to-layer="9333" to-port="1" /> <edge from-layer="9333" from-port="2" to-layer="9536" to-port="0" /> <edge from-layer="9334" from-port="0" to-layer="9335" to-port="1" /> <edge from-layer="9335" from-port="2" to-layer="9337" to-port="0" /> <edge from-layer="9336" from-port="0" to-layer="9337" to-port="1" /> <edge from-layer="9337" from-port="2" to-layer="9339" to-port="0" /> <edge from-layer="9338" from-port="0" to-layer="9339" to-port="1" /> <edge from-layer="9339" from-port="2" to-layer="9347" to-port="0" /> <edge from-layer="9340" from-port="0" to-layer="9341" to-port="0" /> <edge from-layer="9341" from-port="1" to-layer="9344" to-port="0" /> <edge from-layer="9342" from-port="0" to-layer="9343" to-port="0" /> <edge from-layer="9343" from-port="1" to-layer="9344" to-port="1" /> <edge from-layer="9344" from-port="2" to-layer="9346" to-port="0" /> <edge from-layer="9345" from-port="0" to-layer="9346" to-port="1" /> <edge from-layer="9346" from-port="2" to-layer="9347" to-port="1" /> <edge from-layer="9347" from-port="2" to-layer="9349" to-port="0" /> <edge from-layer="9348" from-port="0" to-layer="9349" to-port="1" /> <edge from-layer="9349" from-port="2" to-layer="9351" to-port="0" /> <edge from-layer="9350" from-port="0" to-layer="9351" to-port="1" /> <edge from-layer="9351" from-port="2" to-layer="9355" to-port="0" /> <edge from-layer="9351" from-port="2" to-layer="9473" to-port="0" /> <edge from-layer="9351" from-port="2" to-layer="9367" to-port="0" /> <edge from-layer="9352" from-port="0" to-layer="9355" to-port="1" /> <edge from-layer="9353" from-port="0" to-layer="9355" to-port="2" /> <edge from-layer="9354" from-port="0" to-layer="9355" to-port="3" /> <edge from-layer="9355" from-port="4" to-layer="9357" to-port="0" /> <edge from-layer="9356" from-port="0" to-layer="9357" to-port="1" /> <edge from-layer="9357" from-port="2" to-layer="9426" to-port="0" /> <edge from-layer="9357" from-port="2" to-layer="9361" to-port="0" /> <edge from-layer="9358" from-port="0" to-layer="9361" to-port="1" /> <edge from-layer="9359" from-port="0" to-layer="9361" to-port="2" /> <edge from-layer="9360" from-port="0" to-layer="9361" to-port="3" /> <edge from-layer="9361" from-port="4" to-layer="9407" to-port="0" /> <edge from-layer="9361" from-port="4" to-layer="9396" to-port="0" /> <edge from-layer="9361" from-port="4" to-layer="9393" to-port="0" /> <edge from-layer="9361" from-port="4" to-layer="9412" to-port="0" /> <edge from-layer="9362" from-port="0" to-layer="9389" to-port="0" /> <edge from-layer="9363" from-port="0" to-layer="9389" to-port="1" /> <edge from-layer="9364" from-port="0" to-layer="9367" to-port="1" /> <edge from-layer="9365" from-port="0" to-layer="9367" to-port="2" /> <edge from-layer="9366" from-port="0" to-layer="9367" to-port="3" /> <edge from-layer="9367" from-port="4" to-layer="9369" to-port="0" /> <edge from-layer="9368" from-port="0" to-layer="9369" to-port="1" /> <edge from-layer="9369" from-port="2" to-layer="9431" to-port="0" /> <edge from-layer="9369" from-port="2" to-layer="9458" to-port="0" /> <edge from-layer="9369" from-port="2" to-layer="9370" to-port="0" /> <edge from-layer="9370" from-port="1" to-layer="9372" to-port="0" /> <edge from-layer="9371" from-port="0" to-layer="9372" to-port="2" /> <edge from-layer="9372" from-port="3" to-layer="9385" to-port="0" /> <edge from-layer="9373" from-port="0" to-layer="9378" to-port="0" /> <edge from-layer="9374" from-port="0" to-layer="9377" to-port="1" /> <edge from-layer="9375" from-port="0" to-layer="9377" to-port="2" /> <edge from-layer="9376" from-port="0" to-layer="9377" to-port="3" /> <edge from-layer="9377" from-port="4" to-layer="9378" to-port="1" /> <edge from-layer="9378" from-port="2" to-layer="9379" to-port="0" /> <edge from-layer="9379" from-port="1" to-layer="9381" to-port="0" /> <edge from-layer="9380" from-port="0" to-layer="9381" to-port="2" /> <edge from-layer="9381" from-port="3" to-layer="9382" to-port="0" /> <edge from-layer="9381" from-port="3" to-layer="9460" to-port="0" /> <edge from-layer="9382" from-port="1" to-layer="9384" to-port="0" /> <edge from-layer="9383" from-port="0" to-layer="9384" to-port="2" /> <edge from-layer="9384" from-port="3" to-layer="9385" to-port="1" /> <edge from-layer="9385" from-port="2" to-layer="9387" to-port="0" /> <edge from-layer="9386" from-port="0" to-layer="9387" to-port="1" /> <edge from-layer="9387" from-port="2" to-layer="9417" to-port="2" /> <edge from-layer="9387" from-port="2" to-layer="9389" to-port="2" /> <edge from-layer="9388" from-port="0" to-layer="9389" to-port="3" /> <edge from-layer="9389" from-port="4" to-layer="9391" to-port="0" /> <edge from-layer="9390" from-port="0" to-layer="9391" to-port="2" /> <edge from-layer="9391" from-port="3" to-layer="9392" to-port="0" /> <edge from-layer="9392" from-port="2" to-layer="9432" to-port="1" /> <edge from-layer="9392" from-port="2" to-layer="9393" to-port="1" /> <edge from-layer="9393" from-port="2" to-layer="9422" to-port="0" /> <edge from-layer="9394" from-port="0" to-layer="9404" to-port="0" /> <edge from-layer="9395" from-port="0" to-layer="9404" to-port="1" /> <edge from-layer="9396" from-port="1" to-layer="9398" to-port="0" /> <edge from-layer="9397" from-port="0" to-layer="9398" to-port="2" /> <edge from-layer="9398" from-port="3" to-layer="9399" to-port="0" /> <edge from-layer="9399" from-port="2" to-layer="9400" to-port="0" /> <edge from-layer="9400" from-port="1" to-layer="9402" to-port="0" /> <edge from-layer="9401" from-port="0" to-layer="9402" to-port="1" /> <edge from-layer="9402" from-port="2" to-layer="9404" to-port="2" /> <edge from-layer="9403" from-port="0" to-layer="9404" to-port="3" /> <edge from-layer="9404" from-port="4" to-layer="9412" to-port="2" /> <edge from-layer="9404" from-port="4" to-layer="9407" to-port="1" /> <edge from-layer="9405" from-port="0" to-layer="9407" to-port="2" /> <edge from-layer="9406" from-port="0" to-layer="9407" to-port="3" /> <edge from-layer="9407" from-port="4" to-layer="9409" to-port="0" /> <edge from-layer="9408" from-port="0" to-layer="9409" to-port="1" /> <edge from-layer="9409" from-port="2" to-layer="9413" to-port="0" /> <edge from-layer="9410" from-port="0" to-layer="9412" to-port="1" /> <edge from-layer="9411" from-port="0" to-layer="9412" to-port="3" /> <edge from-layer="9412" from-port="4" to-layer="9413" to-port="1" /> <edge from-layer="9413" from-port="2" to-layer="9421" to-port="0" /> <edge from-layer="9414" from-port="0" to-layer="9417" to-port="0" /> <edge from-layer="9415" from-port="0" to-layer="9417" to-port="1" /> <edge from-layer="9416" from-port="0" to-layer="9417" to-port="3" /> <edge from-layer="9417" from-port="4" to-layer="9419" to-port="0" /> <edge from-layer="9418" from-port="0" to-layer="9419" to-port="2" /> <edge from-layer="9419" from-port="3" to-layer="9420" to-port="0" /> <edge from-layer="9420" from-port="2" to-layer="9453" to-port="1" /> <edge from-layer="9420" from-port="2" to-layer="9421" to-port="1" /> <edge from-layer="9421" from-port="2" to-layer="9422" to-port="1" /> <edge from-layer="9422" from-port="2" to-layer="9427" to-port="0" /> <edge from-layer="9423" from-port="0" to-layer="9426" to-port="1" /> <edge from-layer="9424" from-port="0" to-layer="9426" to-port="2" /> <edge from-layer="9425" from-port="0" to-layer="9426" to-port="3" /> <edge from-layer="9426" from-port="4" to-layer="9427" to-port="1" /> <edge from-layer="9427" from-port="2" to-layer="9483" to-port="0" /> <edge from-layer="9427" from-port="2" to-layer="9521" to-port="0" /> <edge from-layer="9428" from-port="0" to-layer="9431" to-port="1" /> <edge from-layer="9429" from-port="0" to-layer="9431" to-port="2" /> <edge from-layer="9430" from-port="0" to-layer="9431" to-port="3" /> <edge from-layer="9431" from-port="4" to-layer="9451" to-port="0" /> <edge from-layer="9431" from-port="4" to-layer="9446" to-port="0" /> <edge from-layer="9431" from-port="4" to-layer="9432" to-port="0" /> <edge from-layer="9431" from-port="4" to-layer="9435" to-port="0" /> <edge from-layer="9432" from-port="2" to-layer="9454" to-port="0" /> <edge from-layer="9433" from-port="0" to-layer="9443" to-port="0" /> <edge from-layer="9434" from-port="0" to-layer="9443" to-port="1" /> <edge from-layer="9435" from-port="1" to-layer="9437" to-port="0" /> <edge from-layer="9436" from-port="0" to-layer="9437" to-port="2" /> <edge from-layer="9437" from-port="3" to-layer="9438" to-port="0" /> <edge from-layer="9438" from-port="2" to-layer="9439" to-port="0" /> <edge from-layer="9439" from-port="1" to-layer="9441" to-port="0" /> <edge from-layer="9440" from-port="0" to-layer="9441" to-port="1" /> <edge from-layer="9441" from-port="2" to-layer="9443" to-port="2" /> <edge from-layer="9442" from-port="0" to-layer="9443" to-port="3" /> <edge from-layer="9443" from-port="4" to-layer="9451" to-port="2" /> <edge from-layer="9443" from-port="4" to-layer="9446" to-port="1" /> <edge from-layer="9444" from-port="0" to-layer="9446" to-port="2" /> <edge from-layer="9445" from-port="0" to-layer="9446" to-port="3" /> <edge from-layer="9446" from-port="4" to-layer="9448" to-port="0" /> <edge from-layer="9447" from-port="0" to-layer="9448" to-port="1" /> <edge from-layer="9448" from-port="2" to-layer="9452" to-port="0" /> <edge from-layer="9449" from-port="0" to-layer="9451" to-port="1" /> <edge from-layer="9450" from-port="0" to-layer="9451" to-port="3" /> <edge from-layer="9451" from-port="4" to-layer="9452" to-port="1" /> <edge from-layer="9452" from-port="2" to-layer="9453" to-port="0" /> <edge from-layer="9453" from-port="2" to-layer="9454" to-port="1" /> <edge from-layer="9454" from-port="2" to-layer="9459" to-port="0" /> <edge from-layer="9455" from-port="0" to-layer="9458" to-port="1" /> <edge from-layer="9456" from-port="0" to-layer="9458" to-port="2" /> <edge from-layer="9457" from-port="0" to-layer="9458" to-port="3" /> <edge from-layer="9458" from-port="4" to-layer="9459" to-port="1" /> <edge from-layer="9459" from-port="2" to-layer="9460" to-port="1" /> <edge from-layer="9460" from-port="2" to-layer="9480" to-port="0" /> <edge from-layer="9460" from-port="2" to-layer="9521" to-port="1" /> <edge from-layer="9460" from-port="2" to-layer="10103" to-port="0" /> <edge from-layer="9461" from-port="0" to-layer="9466" to-port="0" /> <edge from-layer="9462" from-port="0" to-layer="9465" to-port="1" /> <edge from-layer="9463" from-port="0" to-layer="9465" to-port="2" /> <edge from-layer="9464" from-port="0" to-layer="9465" to-port="3" /> <edge from-layer="9465" from-port="4" to-layer="9466" to-port="1" /> <edge from-layer="9466" from-port="2" to-layer="9467" to-port="0" /> <edge from-layer="9467" from-port="1" to-layer="9469" to-port="0" /> <edge from-layer="9468" from-port="0" to-layer="9469" to-port="2" /> <edge from-layer="9469" from-port="3" to-layer="9476" to-port="0" /> <edge from-layer="9470" from-port="0" to-layer="9473" to-port="1" /> <edge from-layer="9471" from-port="0" to-layer="9473" to-port="2" /> <edge from-layer="9472" from-port="0" to-layer="9473" to-port="3" /> <edge from-layer="9473" from-port="4" to-layer="9475" to-port="0" /> <edge from-layer="9474" from-port="0" to-layer="9475" to-port="1" /> <edge from-layer="9475" from-port="2" to-layer="9476" to-port="1" /> <edge from-layer="9476" from-port="2" to-layer="9521" to-port="2" /> <edge from-layer="9476" from-port="2" to-layer="10102" to-port="0" /> <edge from-layer="9477" from-port="0" to-layer="9495" to-port="0" /> <edge from-layer="9478" from-port="0" to-layer="9490" to-port="0" /> <edge from-layer="9479" from-port="0" to-layer="9490" to-port="1" /> <edge from-layer="9479" from-port="0" to-layer="9493" to-port="1" /> <edge from-layer="9480" from-port="1" to-layer="9482" to-port="0" /> <edge from-layer="9481" from-port="0" to-layer="9482" to-port="2" /> <edge from-layer="9482" from-port="3" to-layer="9486" to-port="0" /> <edge from-layer="9482" from-port="3" to-layer="9492" to-port="0" /> <edge from-layer="9483" from-port="1" to-layer="9485" to-port="0" /> <edge from-layer="9483" from-port="1" to-layer="9510" to-port="0" /> <edge from-layer="9484" from-port="0" to-layer="9485" to-port="2" /> <edge from-layer="9485" from-port="3" to-layer="9486" to-port="1" /> <edge from-layer="9486" from-port="2" to-layer="9488" to-port="0" /> <edge from-layer="9487" from-port="0" to-layer="9488" to-port="1" /> <edge from-layer="9487" from-port="0" to-layer="9492" to-port="1" /> <edge from-layer="9488" from-port="2" to-layer="9490" to-port="2" /> <edge from-layer="9489" from-port="0" to-layer="9493" to-port="3" /> <edge from-layer="9489" from-port="0" to-layer="9490" to-port="3" /> <edge from-layer="9490" from-port="4" to-layer="9495" to-port="1" /> <edge from-layer="9491" from-port="0" to-layer="9493" to-port="0" /> <edge from-layer="9492" from-port="2" to-layer="9493" to-port="2" /> <edge from-layer="9492" from-port="2" to-layer="9500" to-port="2" /> <edge from-layer="9493" from-port="4" to-layer="9495" to-port="2" /> <edge from-layer="9494" from-port="0" to-layer="9495" to-port="3" /> <edge from-layer="9495" from-port="4" to-layer="9502" to-port="0" /> <edge from-layer="9496" from-port="0" to-layer="9502" to-port="1" /> <edge from-layer="9497" from-port="0" to-layer="9500" to-port="0" /> <edge from-layer="9498" from-port="0" to-layer="9500" to-port="1" /> <edge from-layer="9499" from-port="0" to-layer="9500" to-port="3" /> <edge from-layer="9500" from-port="4" to-layer="9502" to-port="2" /> <edge from-layer="9501" from-port="0" to-layer="9502" to-port="3" /> <edge from-layer="9502" from-port="4" to-layer="9505" to-port="0" /> <edge from-layer="9503" from-port="0" to-layer="9505" to-port="1" /> <edge from-layer="9504" from-port="0" to-layer="9505" to-port="2" /> <edge from-layer="9505" from-port="3" to-layer="9506" to-port="0" /> <edge from-layer="9506" from-port="2" to-layer="9507" to-port="0" /> <edge from-layer="9506" from-port="2" to-layer="9520" to-port="0" /> <edge from-layer="9507" from-port="1" to-layer="9519" to-port="0" /> <edge from-layer="9508" from-port="0" to-layer="9510" to-port="1" /> <edge from-layer="9509" from-port="0" to-layer="9510" to-port="2" /> <edge from-layer="9510" from-port="3" to-layer="9514" to-port="0" /> <edge from-layer="9511" from-port="0" to-layer="9514" to-port="1" /> <edge from-layer="9512" from-port="0" to-layer="9514" to-port="2" /> <edge from-layer="9513" from-port="0" to-layer="9514" to-port="3" /> <edge from-layer="9514" from-port="4" to-layer="9516" to-port="0" /> <edge from-layer="9514" from-port="4" to-layer="9518" to-port="2" /> <edge from-layer="9515" from-port="0" to-layer="9516" to-port="1" /> <edge from-layer="9516" from-port="2" to-layer="9518" to-port="0" /> <edge from-layer="9517" from-port="0" to-layer="9518" to-port="1" /> <edge from-layer="9518" from-port="3" to-layer="9519" to-port="1" /> <edge from-layer="9519" from-port="2" to-layer="9520" to-port="1" /> <edge from-layer="9520" from-port="2" to-layer="9521" to-port="3" /> <edge from-layer="9521" from-port="4" to-layer="9523" to-port="0" /> <edge from-layer="9522" from-port="0" to-layer="9523" to-port="1" /> <edge from-layer="9523" from-port="2" to-layer="9525" to-port="0" /> <edge from-layer="9524" from-port="0" to-layer="9525" to-port="1" /> <edge from-layer="9525" from-port="2" to-layer="9533" to-port="0" /> <edge from-layer="9526" from-port="0" to-layer="9527" to-port="0" /> <edge from-layer="9527" from-port="1" to-layer="9530" to-port="0" /> <edge from-layer="9528" from-port="0" to-layer="9529" to-port="0" /> <edge from-layer="9529" from-port="1" to-layer="9530" to-port="1" /> <edge from-layer="9530" from-port="2" to-layer="9532" to-port="0" /> <edge from-layer="9531" from-port="0" to-layer="9532" to-port="1" /> <edge from-layer="9532" from-port="2" to-layer="9533" to-port="1" /> <edge from-layer="9533" from-port="2" to-layer="9535" to-port="0" /> <edge from-layer="9534" from-port="0" to-layer="9535" to-port="1" /> <edge from-layer="9535" from-port="2" to-layer="9536" to-port="1" /> <edge from-layer="9536" from-port="2" to-layer="9537" to-port="0" /> <edge from-layer="9537" from-port="2" to-layer="9566" to-port="0" /> <edge from-layer="9537" from-port="2" to-layer="9768" to-port="1" /> <edge from-layer="9537" from-port="2" to-layer="9539" to-port="0" /> <edge from-layer="9538" from-port="0" to-layer="9539" to-port="1" /> <edge from-layer="9539" from-port="2" to-layer="9541" to-port="0" /> <edge from-layer="9540" from-port="0" to-layer="9541" to-port="1" /> <edge from-layer="9541" from-port="2" to-layer="9543" to-port="0" /> <edge from-layer="9542" from-port="0" to-layer="9543" to-port="1" /> <edge from-layer="9543" from-port="2" to-layer="9551" to-port="0" /> <edge from-layer="9544" from-port="0" to-layer="9545" to-port="0" /> <edge from-layer="9545" from-port="1" to-layer="9548" to-port="0" /> <edge from-layer="9546" from-port="0" to-layer="9547" to-port="0" /> <edge from-layer="9547" from-port="1" to-layer="9548" to-port="1" /> <edge from-layer="9548" from-port="2" to-layer="9550" to-port="0" /> <edge from-layer="9549" from-port="0" to-layer="9550" to-port="1" /> <edge from-layer="9550" from-port="2" to-layer="9551" to-port="1" /> <edge from-layer="9551" from-port="2" to-layer="9553" to-port="0" /> <edge from-layer="9552" from-port="0" to-layer="9553" to-port="1" /> <edge from-layer="9553" from-port="2" to-layer="9554" to-port="0" /> <edge from-layer="9554" from-port="1" to-layer="9562" to-port="0" /> <edge from-layer="9555" from-port="0" to-layer="9556" to-port="0" /> <edge from-layer="9556" from-port="1" to-layer="9559" to-port="0" /> <edge from-layer="9557" from-port="0" to-layer="9558" to-port="0" /> <edge from-layer="9558" from-port="1" to-layer="9559" to-port="1" /> <edge from-layer="9559" from-port="2" to-layer="9561" to-port="0" /> <edge from-layer="9560" from-port="0" to-layer="9561" to-port="1" /> <edge from-layer="9561" from-port="2" to-layer="9562" to-port="1" /> <edge from-layer="9562" from-port="2" to-layer="9564" to-port="0" /> <edge from-layer="9563" from-port="0" to-layer="9564" to-port="1" /> <edge from-layer="9564" from-port="2" to-layer="9767" to-port="0" /> <edge from-layer="9565" from-port="0" to-layer="9566" to-port="1" /> <edge from-layer="9566" from-port="2" to-layer="9568" to-port="0" /> <edge from-layer="9567" from-port="0" to-layer="9568" to-port="1" /> <edge from-layer="9568" from-port="2" to-layer="9570" to-port="0" /> <edge from-layer="9569" from-port="0" to-layer="9570" to-port="1" /> <edge from-layer="9570" from-port="2" to-layer="9578" to-port="0" /> <edge from-layer="9571" from-port="0" to-layer="9572" to-port="0" /> <edge from-layer="9572" from-port="1" to-layer="9575" to-port="0" /> <edge from-layer="9573" from-port="0" to-layer="9574" to-port="0" /> <edge from-layer="9574" from-port="1" to-layer="9575" to-port="1" /> <edge from-layer="9575" from-port="2" to-layer="9577" to-port="0" /> <edge from-layer="9576" from-port="0" to-layer="9577" to-port="1" /> <edge from-layer="9577" from-port="2" to-layer="9578" to-port="1" /> <edge from-layer="9578" from-port="2" to-layer="9580" to-port="0" /> <edge from-layer="9579" from-port="0" to-layer="9580" to-port="1" /> <edge from-layer="9580" from-port="2" to-layer="9582" to-port="0" /> <edge from-layer="9581" from-port="0" to-layer="9582" to-port="1" /> <edge from-layer="9582" from-port="2" to-layer="9704" to-port="0" /> <edge from-layer="9582" from-port="2" to-layer="9598" to-port="0" /> <edge from-layer="9582" from-port="2" to-layer="9586" to-port="0" /> <edge from-layer="9583" from-port="0" to-layer="9586" to-port="1" /> <edge from-layer="9584" from-port="0" to-layer="9586" to-port="2" /> <edge from-layer="9585" from-port="0" to-layer="9586" to-port="3" /> <edge from-layer="9586" from-port="4" to-layer="9588" to-port="0" /> <edge from-layer="9587" from-port="0" to-layer="9588" to-port="1" /> <edge from-layer="9588" from-port="2" to-layer="9657" to-port="0" /> <edge from-layer="9588" from-port="2" to-layer="9592" to-port="0" /> <edge from-layer="9589" from-port="0" to-layer="9592" to-port="1" /> <edge from-layer="9590" from-port="0" to-layer="9592" to-port="2" /> <edge from-layer="9591" from-port="0" to-layer="9592" to-port="3" /> <edge from-layer="9592" from-port="4" to-layer="9638" to-port="0" /> <edge from-layer="9592" from-port="4" to-layer="9624" to-port="0" /> <edge from-layer="9592" from-port="4" to-layer="9627" to-port="0" /> <edge from-layer="9592" from-port="4" to-layer="9643" to-port="0" /> <edge from-layer="9593" from-port="0" to-layer="9620" to-port="0" /> <edge from-layer="9594" from-port="0" to-layer="9620" to-port="1" /> <edge from-layer="9595" from-port="0" to-layer="9598" to-port="1" /> <edge from-layer="9596" from-port="0" to-layer="9598" to-port="2" /> <edge from-layer="9597" from-port="0" to-layer="9598" to-port="3" /> <edge from-layer="9598" from-port="4" to-layer="9600" to-port="0" /> <edge from-layer="9599" from-port="0" to-layer="9600" to-port="1" /> <edge from-layer="9600" from-port="2" to-layer="9601" to-port="0" /> <edge from-layer="9600" from-port="2" to-layer="9662" to-port="0" /> <edge from-layer="9600" from-port="2" to-layer="9689" to-port="0" /> <edge from-layer="9601" from-port="1" to-layer="9603" to-port="0" /> <edge from-layer="9602" from-port="0" to-layer="9603" to-port="2" /> <edge from-layer="9603" from-port="3" to-layer="9616" to-port="0" /> <edge from-layer="9604" from-port="0" to-layer="9609" to-port="0" /> <edge from-layer="9605" from-port="0" to-layer="9608" to-port="1" /> <edge from-layer="9606" from-port="0" to-layer="9608" to-port="2" /> <edge from-layer="9607" from-port="0" to-layer="9608" to-port="3" /> <edge from-layer="9608" from-port="4" to-layer="9609" to-port="1" /> <edge from-layer="9609" from-port="2" to-layer="9610" to-port="0" /> <edge from-layer="9610" from-port="1" to-layer="9612" to-port="0" /> <edge from-layer="9611" from-port="0" to-layer="9612" to-port="2" /> <edge from-layer="9612" from-port="3" to-layer="9691" to-port="0" /> <edge from-layer="9612" from-port="3" to-layer="9613" to-port="0" /> <edge from-layer="9613" from-port="1" to-layer="9615" to-port="0" /> <edge from-layer="9614" from-port="0" to-layer="9615" to-port="2" /> <edge from-layer="9615" from-port="3" to-layer="9616" to-port="1" /> <edge from-layer="9616" from-port="2" to-layer="9618" to-port="0" /> <edge from-layer="9617" from-port="0" to-layer="9618" to-port="1" /> <edge from-layer="9618" from-port="2" to-layer="9648" to-port="2" /> <edge from-layer="9618" from-port="2" to-layer="9620" to-port="2" /> <edge from-layer="9619" from-port="0" to-layer="9620" to-port="3" /> <edge from-layer="9620" from-port="4" to-layer="9622" to-port="0" /> <edge from-layer="9621" from-port="0" to-layer="9622" to-port="2" /> <edge from-layer="9622" from-port="3" to-layer="9623" to-port="0" /> <edge from-layer="9623" from-port="2" to-layer="9663" to-port="1" /> <edge from-layer="9623" from-port="2" to-layer="9624" to-port="1" /> <edge from-layer="9624" from-port="2" to-layer="9653" to-port="0" /> <edge from-layer="9625" from-port="0" to-layer="9635" to-port="0" /> <edge from-layer="9626" from-port="0" to-layer="9635" to-port="1" /> <edge from-layer="9627" from-port="1" to-layer="9629" to-port="0" /> <edge from-layer="9628" from-port="0" to-layer="9629" to-port="2" /> <edge from-layer="9629" from-port="3" to-layer="9630" to-port="0" /> <edge from-layer="9630" from-port="2" to-layer="9631" to-port="0" /> <edge from-layer="9631" from-port="1" to-layer="9633" to-port="0" /> <edge from-layer="9632" from-port="0" to-layer="9633" to-port="1" /> <edge from-layer="9633" from-port="2" to-layer="9635" to-port="2" /> <edge from-layer="9634" from-port="0" to-layer="9635" to-port="3" /> <edge from-layer="9635" from-port="4" to-layer="9643" to-port="2" /> <edge from-layer="9635" from-port="4" to-layer="9638" to-port="1" /> <edge from-layer="9636" from-port="0" to-layer="9638" to-port="2" /> <edge from-layer="9637" from-port="0" to-layer="9638" to-port="3" /> <edge from-layer="9638" from-port="4" to-layer="9640" to-port="0" /> <edge from-layer="9639" from-port="0" to-layer="9640" to-port="1" /> <edge from-layer="9640" from-port="2" to-layer="9644" to-port="0" /> <edge from-layer="9641" from-port="0" to-layer="9643" to-port="1" /> <edge from-layer="9642" from-port="0" to-layer="9643" to-port="3" /> <edge from-layer="9643" from-port="4" to-layer="9644" to-port="1" /> <edge from-layer="9644" from-port="2" to-layer="9652" to-port="0" /> <edge from-layer="9645" from-port="0" to-layer="9648" to-port="0" /> <edge from-layer="9646" from-port="0" to-layer="9648" to-port="1" /> <edge from-layer="9647" from-port="0" to-layer="9648" to-port="3" /> <edge from-layer="9648" from-port="4" to-layer="9650" to-port="0" /> <edge from-layer="9649" from-port="0" to-layer="9650" to-port="2" /> <edge from-layer="9650" from-port="3" to-layer="9651" to-port="0" /> <edge from-layer="9651" from-port="2" to-layer="9652" to-port="1" /> <edge from-layer="9651" from-port="2" to-layer="9684" to-port="1" /> <edge from-layer="9652" from-port="2" to-layer="9653" to-port="1" /> <edge from-layer="9653" from-port="2" to-layer="9658" to-port="0" /> <edge from-layer="9654" from-port="0" to-layer="9657" to-port="1" /> <edge from-layer="9655" from-port="0" to-layer="9657" to-port="2" /> <edge from-layer="9656" from-port="0" to-layer="9657" to-port="3" /> <edge from-layer="9657" from-port="4" to-layer="9658" to-port="1" /> <edge from-layer="9658" from-port="2" to-layer="9714" to-port="0" /> <edge from-layer="9658" from-port="2" to-layer="9752" to-port="0" /> <edge from-layer="9659" from-port="0" to-layer="9662" to-port="1" /> <edge from-layer="9660" from-port="0" to-layer="9662" to-port="2" /> <edge from-layer="9661" from-port="0" to-layer="9662" to-port="3" /> <edge from-layer="9662" from-port="4" to-layer="9663" to-port="0" /> <edge from-layer="9662" from-port="4" to-layer="9677" to-port="0" /> <edge from-layer="9662" from-port="4" to-layer="9682" to-port="0" /> <edge from-layer="9662" from-port="4" to-layer="9666" to-port="0" /> <edge from-layer="9663" from-port="2" to-layer="9685" to-port="0" /> <edge from-layer="9664" from-port="0" to-layer="9674" to-port="0" /> <edge from-layer="9665" from-port="0" to-layer="9674" to-port="1" /> <edge from-layer="9666" from-port="1" to-layer="9668" to-port="0" /> <edge from-layer="9667" from-port="0" to-layer="9668" to-port="2" /> <edge from-layer="9668" from-port="3" to-layer="9669" to-port="0" /> <edge from-layer="9669" from-port="2" to-layer="9670" to-port="0" /> <edge from-layer="9670" from-port="1" to-layer="9672" to-port="0" /> <edge from-layer="9671" from-port="0" to-layer="9672" to-port="1" /> <edge from-layer="9672" from-port="2" to-layer="9674" to-port="2" /> <edge from-layer="9673" from-port="0" to-layer="9674" to-port="3" /> <edge from-layer="9674" from-port="4" to-layer="9677" to-port="1" /> <edge from-layer="9674" from-port="4" to-layer="9682" to-port="2" /> <edge from-layer="9675" from-port="0" to-layer="9677" to-port="2" /> <edge from-layer="9676" from-port="0" to-layer="9677" to-port="3" /> <edge from-layer="9677" from-port="4" to-layer="9679" to-port="0" /> <edge from-layer="9678" from-port="0" to-layer="9679" to-port="1" /> <edge from-layer="9679" from-port="2" to-layer="9683" to-port="0" /> <edge from-layer="9680" from-port="0" to-layer="9682" to-port="1" /> <edge from-layer="9681" from-port="0" to-layer="9682" to-port="3" /> <edge from-layer="9682" from-port="4" to-layer="9683" to-port="1" /> <edge from-layer="9683" from-port="2" to-layer="9684" to-port="0" /> <edge from-layer="9684" from-port="2" to-layer="9685" to-port="1" /> <edge from-layer="9685" from-port="2" to-layer="9690" to-port="0" /> <edge from-layer="9686" from-port="0" to-layer="9689" to-port="1" /> <edge from-layer="9687" from-port="0" to-layer="9689" to-port="2" /> <edge from-layer="9688" from-port="0" to-layer="9689" to-port="3" /> <edge from-layer="9689" from-port="4" to-layer="9690" to-port="1" /> <edge from-layer="9690" from-port="2" to-layer="9691" to-port="1" /> <edge from-layer="9691" from-port="2" to-layer="9711" to-port="0" /> <edge from-layer="9691" from-port="2" to-layer="10101" to-port="0" /> <edge from-layer="9691" from-port="2" to-layer="9752" to-port="1" /> <edge from-layer="9692" from-port="0" to-layer="9697" to-port="0" /> <edge from-layer="9693" from-port="0" to-layer="9696" to-port="1" /> <edge from-layer="9694" from-port="0" to-layer="9696" to-port="2" /> <edge from-layer="9695" from-port="0" to-layer="9696" to-port="3" /> <edge from-layer="9696" from-port="4" to-layer="9697" to-port="1" /> <edge from-layer="9697" from-port="2" to-layer="9698" to-port="0" /> <edge from-layer="9698" from-port="1" to-layer="9700" to-port="0" /> <edge from-layer="9699" from-port="0" to-layer="9700" to-port="2" /> <edge from-layer="9700" from-port="3" to-layer="9707" to-port="0" /> <edge from-layer="9701" from-port="0" to-layer="9704" to-port="1" /> <edge from-layer="9702" from-port="0" to-layer="9704" to-port="2" /> <edge from-layer="9703" from-port="0" to-layer="9704" to-port="3" /> <edge from-layer="9704" from-port="4" to-layer="9706" to-port="0" /> <edge from-layer="9705" from-port="0" to-layer="9706" to-port="1" /> <edge from-layer="9706" from-port="2" to-layer="9707" to-port="1" /> <edge from-layer="9707" from-port="2" to-layer="10100" to-port="0" /> <edge from-layer="9707" from-port="2" to-layer="9752" to-port="2" /> <edge from-layer="9708" from-port="0" to-layer="9726" to-port="0" /> <edge from-layer="9709" from-port="0" to-layer="9721" to-port="0" /> <edge from-layer="9710" from-port="0" to-layer="9721" to-port="1" /> <edge from-layer="9710" from-port="0" to-layer="9724" to-port="1" /> <edge from-layer="9711" from-port="1" to-layer="9713" to-port="0" /> <edge from-layer="9712" from-port="0" to-layer="9713" to-port="2" /> <edge from-layer="9713" from-port="3" to-layer="9717" to-port="0" /> <edge from-layer="9713" from-port="3" to-layer="9723" to-port="0" /> <edge from-layer="9714" from-port="1" to-layer="9741" to-port="0" /> <edge from-layer="9714" from-port="1" to-layer="9716" to-port="0" /> <edge from-layer="9715" from-port="0" to-layer="9716" to-port="2" /> <edge from-layer="9716" from-port="3" to-layer="9717" to-port="1" /> <edge from-layer="9717" from-port="2" to-layer="9719" to-port="0" /> <edge from-layer="9718" from-port="0" to-layer="9719" to-port="1" /> <edge from-layer="9718" from-port="0" to-layer="9723" to-port="1" /> <edge from-layer="9719" from-port="2" to-layer="9721" to-port="2" /> <edge from-layer="9720" from-port="0" to-layer="9724" to-port="3" /> <edge from-layer="9720" from-port="0" to-layer="9721" to-port="3" /> <edge from-layer="9721" from-port="4" to-layer="9726" to-port="1" /> <edge from-layer="9722" from-port="0" to-layer="9724" to-port="0" /> <edge from-layer="9723" from-port="2" to-layer="9724" to-port="2" /> <edge from-layer="9723" from-port="2" to-layer="9731" to-port="2" /> <edge from-layer="9724" from-port="4" to-layer="9726" to-port="2" /> <edge from-layer="9725" from-port="0" to-layer="9726" to-port="3" /> <edge from-layer="9726" from-port="4" to-layer="9733" to-port="0" /> <edge from-layer="9727" from-port="0" to-layer="9733" to-port="1" /> <edge from-layer="9728" from-port="0" to-layer="9731" to-port="0" /> <edge from-layer="9729" from-port="0" to-layer="9731" to-port="1" /> <edge from-layer="9730" from-port="0" to-layer="9731" to-port="3" /> <edge from-layer="9731" from-port="4" to-layer="9733" to-port="2" /> <edge from-layer="9732" from-port="0" to-layer="9733" to-port="3" /> <edge from-layer="9733" from-port="4" to-layer="9736" to-port="0" /> <edge from-layer="9734" from-port="0" to-layer="9736" to-port="1" /> <edge from-layer="9735" from-port="0" to-layer="9736" to-port="2" /> <edge from-layer="9736" from-port="3" to-layer="9737" to-port="0" /> <edge from-layer="9737" from-port="2" to-layer="9751" to-port="0" /> <edge from-layer="9737" from-port="2" to-layer="9738" to-port="0" /> <edge from-layer="9738" from-port="1" to-layer="9750" to-port="0" /> <edge from-layer="9739" from-port="0" to-layer="9741" to-port="1" /> <edge from-layer="9740" from-port="0" to-layer="9741" to-port="2" /> <edge from-layer="9741" from-port="3" to-layer="9745" to-port="0" /> <edge from-layer="9742" from-port="0" to-layer="9745" to-port="1" /> <edge from-layer="9743" from-port="0" to-layer="9745" to-port="2" /> <edge from-layer="9744" from-port="0" to-layer="9745" to-port="3" /> <edge from-layer="9745" from-port="4" to-layer="9749" to-port="2" /> <edge from-layer="9745" from-port="4" to-layer="9747" to-port="0" /> <edge from-layer="9746" from-port="0" to-layer="9747" to-port="1" /> <edge from-layer="9747" from-port="2" to-layer="9749" to-port="0" /> <edge from-layer="9748" from-port="0" to-layer="9749" to-port="1" /> <edge from-layer="9749" from-port="3" to-layer="9750" to-port="1" /> <edge from-layer="9750" from-port="2" to-layer="9751" to-port="1" /> <edge from-layer="9751" from-port="2" to-layer="9752" to-port="3" /> <edge from-layer="9752" from-port="4" to-layer="9754" to-port="0" /> <edge from-layer="9753" from-port="0" to-layer="9754" to-port="1" /> <edge from-layer="9754" from-port="2" to-layer="9756" to-port="0" /> <edge from-layer="9755" from-port="0" to-layer="9756" to-port="1" /> <edge from-layer="9756" from-port="2" to-layer="9764" to-port="0" /> <edge from-layer="9757" from-port="0" to-layer="9758" to-port="0" /> <edge from-layer="9758" from-port="1" to-layer="9761" to-port="0" /> <edge from-layer="9759" from-port="0" to-layer="9760" to-port="0" /> <edge from-layer="9760" from-port="1" to-layer="9761" to-port="1" /> <edge from-layer="9761" from-port="2" to-layer="9763" to-port="0" /> <edge from-layer="9762" from-port="0" to-layer="9763" to-port="1" /> <edge from-layer="9763" from-port="2" to-layer="9764" to-port="1" /> <edge from-layer="9764" from-port="2" to-layer="9766" to-port="0" /> <edge from-layer="9765" from-port="0" to-layer="9766" to-port="1" /> <edge from-layer="9766" from-port="2" to-layer="9767" to-port="1" /> <edge from-layer="9767" from-port="2" to-layer="9768" to-port="0" /> <edge from-layer="9768" from-port="2" to-layer="9770" to-port="0" /> <edge from-layer="9768" from-port="2" to-layer="9999" to-port="1" /> <edge from-layer="9768" from-port="2" to-layer="9797" to-port="0" /> <edge from-layer="9769" from-port="0" to-layer="9770" to-port="1" /> <edge from-layer="9770" from-port="2" to-layer="9772" to-port="0" /> <edge from-layer="9771" from-port="0" to-layer="9772" to-port="1" /> <edge from-layer="9772" from-port="2" to-layer="9774" to-port="0" /> <edge from-layer="9773" from-port="0" to-layer="9774" to-port="1" /> <edge from-layer="9774" from-port="2" to-layer="9782" to-port="0" /> <edge from-layer="9775" from-port="0" to-layer="9776" to-port="0" /> <edge from-layer="9776" from-port="1" to-layer="9779" to-port="0" /> <edge from-layer="9777" from-port="0" to-layer="9778" to-port="0" /> <edge from-layer="9778" from-port="1" to-layer="9779" to-port="1" /> <edge from-layer="9779" from-port="2" to-layer="9781" to-port="0" /> <edge from-layer="9780" from-port="0" to-layer="9781" to-port="1" /> <edge from-layer="9781" from-port="2" to-layer="9782" to-port="1" /> <edge from-layer="9782" from-port="2" to-layer="9784" to-port="0" /> <edge from-layer="9783" from-port="0" to-layer="9784" to-port="1" /> <edge from-layer="9784" from-port="2" to-layer="9785" to-port="0" /> <edge from-layer="9785" from-port="1" to-layer="9793" to-port="0" /> <edge from-layer="9786" from-port="0" to-layer="9787" to-port="0" /> <edge from-layer="9787" from-port="1" to-layer="9790" to-port="0" /> <edge from-layer="9788" from-port="0" to-layer="9789" to-port="0" /> <edge from-layer="9789" from-port="1" to-layer="9790" to-port="1" /> <edge from-layer="9790" from-port="2" to-layer="9792" to-port="0" /> <edge from-layer="9791" from-port="0" to-layer="9792" to-port="1" /> <edge from-layer="9792" from-port="2" to-layer="9793" to-port="1" /> <edge from-layer="9793" from-port="2" to-layer="9795" to-port="0" /> <edge from-layer="9794" from-port="0" to-layer="9795" to-port="1" /> <edge from-layer="9795" from-port="2" to-layer="9998" to-port="0" /> <edge from-layer="9796" from-port="0" to-layer="9797" to-port="1" /> <edge from-layer="9797" from-port="2" to-layer="9799" to-port="0" /> <edge from-layer="9798" from-port="0" to-layer="9799" to-port="1" /> <edge from-layer="9799" from-port="2" to-layer="9801" to-port="0" /> <edge from-layer="9800" from-port="0" to-layer="9801" to-port="1" /> <edge from-layer="9801" from-port="2" to-layer="9809" to-port="0" /> <edge from-layer="9802" from-port="0" to-layer="9803" to-port="0" /> <edge from-layer="9803" from-port="1" to-layer="9806" to-port="0" /> <edge from-layer="9804" from-port="0" to-layer="9805" to-port="0" /> <edge from-layer="9805" from-port="1" to-layer="9806" to-port="1" /> <edge from-layer="9806" from-port="2" to-layer="9808" to-port="0" /> <edge from-layer="9807" from-port="0" to-layer="9808" to-port="1" /> <edge from-layer="9808" from-port="2" to-layer="9809" to-port="1" /> <edge from-layer="9809" from-port="2" to-layer="9811" to-port="0" /> <edge from-layer="9810" from-port="0" to-layer="9811" to-port="1" /> <edge from-layer="9811" from-port="2" to-layer="9813" to-port="0" /> <edge from-layer="9812" from-port="0" to-layer="9813" to-port="1" /> <edge from-layer="9813" from-port="2" to-layer="9935" to-port="0" /> <edge from-layer="9813" from-port="2" to-layer="9817" to-port="0" /> <edge from-layer="9813" from-port="2" to-layer="9829" to-port="0" /> <edge from-layer="9814" from-port="0" to-layer="9817" to-port="1" /> <edge from-layer="9815" from-port="0" to-layer="9817" to-port="2" /> <edge from-layer="9816" from-port="0" to-layer="9817" to-port="3" /> <edge from-layer="9817" from-port="4" to-layer="9819" to-port="0" /> <edge from-layer="9818" from-port="0" to-layer="9819" to-port="1" /> <edge from-layer="9819" from-port="2" to-layer="9888" to-port="0" /> <edge from-layer="9819" from-port="2" to-layer="9823" to-port="0" /> <edge from-layer="9820" from-port="0" to-layer="9823" to-port="1" /> <edge from-layer="9821" from-port="0" to-layer="9823" to-port="2" /> <edge from-layer="9822" from-port="0" to-layer="9823" to-port="3" /> <edge from-layer="9823" from-port="4" to-layer="9874" to-port="0" /> <edge from-layer="9823" from-port="4" to-layer="9869" to-port="0" /> <edge from-layer="9823" from-port="4" to-layer="9858" to-port="0" /> <edge from-layer="9823" from-port="4" to-layer="9855" to-port="0" /> <edge from-layer="9824" from-port="0" to-layer="9851" to-port="0" /> <edge from-layer="9825" from-port="0" to-layer="9851" to-port="1" /> <edge from-layer="9826" from-port="0" to-layer="9829" to-port="1" /> <edge from-layer="9827" from-port="0" to-layer="9829" to-port="2" /> <edge from-layer="9828" from-port="0" to-layer="9829" to-port="3" /> <edge from-layer="9829" from-port="4" to-layer="9831" to-port="0" /> <edge from-layer="9830" from-port="0" to-layer="9831" to-port="1" /> <edge from-layer="9831" from-port="2" to-layer="9893" to-port="0" /> <edge from-layer="9831" from-port="2" to-layer="9832" to-port="0" /> <edge from-layer="9831" from-port="2" to-layer="9920" to-port="0" /> <edge from-layer="9832" from-port="1" to-layer="9834" to-port="0" /> <edge from-layer="9833" from-port="0" to-layer="9834" to-port="2" /> <edge from-layer="9834" from-port="3" to-layer="9847" to-port="0" /> <edge from-layer="9835" from-port="0" to-layer="9840" to-port="0" /> <edge from-layer="9836" from-port="0" to-layer="9839" to-port="1" /> <edge from-layer="9837" from-port="0" to-layer="9839" to-port="2" /> <edge from-layer="9838" from-port="0" to-layer="9839" to-port="3" /> <edge from-layer="9839" from-port="4" to-layer="9840" to-port="1" /> <edge from-layer="9840" from-port="2" to-layer="9841" to-port="0" /> <edge from-layer="9841" from-port="1" to-layer="9843" to-port="0" /> <edge from-layer="9842" from-port="0" to-layer="9843" to-port="2" /> <edge from-layer="9843" from-port="3" to-layer="9922" to-port="0" /> <edge from-layer="9843" from-port="3" to-layer="9844" to-port="0" /> <edge from-layer="9844" from-port="1" to-layer="9846" to-port="0" /> <edge from-layer="9845" from-port="0" to-layer="9846" to-port="2" /> <edge from-layer="9846" from-port="3" to-layer="9847" to-port="1" /> <edge from-layer="9847" from-port="2" to-layer="9849" to-port="0" /> <edge from-layer="9848" from-port="0" to-layer="9849" to-port="1" /> <edge from-layer="9849" from-port="2" to-layer="9851" to-port="2" /> <edge from-layer="9849" from-port="2" to-layer="9879" to-port="2" /> <edge from-layer="9850" from-port="0" to-layer="9851" to-port="3" /> <edge from-layer="9851" from-port="4" to-layer="9853" to-port="0" /> <edge from-layer="9852" from-port="0" to-layer="9853" to-port="2" /> <edge from-layer="9853" from-port="3" to-layer="9854" to-port="0" /> <edge from-layer="9854" from-port="2" to-layer="9855" to-port="1" /> <edge from-layer="9854" from-port="2" to-layer="9894" to-port="1" /> <edge from-layer="9855" from-port="2" to-layer="9884" to-port="0" /> <edge from-layer="9856" from-port="0" to-layer="9866" to-port="0" /> <edge from-layer="9857" from-port="0" to-layer="9866" to-port="1" /> <edge from-layer="9858" from-port="1" to-layer="9860" to-port="0" /> <edge from-layer="9859" from-port="0" to-layer="9860" to-port="2" /> <edge from-layer="9860" from-port="3" to-layer="9861" to-port="0" /> <edge from-layer="9861" from-port="2" to-layer="9862" to-port="0" /> <edge from-layer="9862" from-port="1" to-layer="9864" to-port="0" /> <edge from-layer="9863" from-port="0" to-layer="9864" to-port="1" /> <edge from-layer="9864" from-port="2" to-layer="9866" to-port="2" /> <edge from-layer="9865" from-port="0" to-layer="9866" to-port="3" /> <edge from-layer="9866" from-port="4" to-layer="9874" to-port="2" /> <edge from-layer="9866" from-port="4" to-layer="9869" to-port="1" /> <edge from-layer="9867" from-port="0" to-layer="9869" to-port="2" /> <edge from-layer="9868" from-port="0" to-layer="9869" to-port="3" /> <edge from-layer="9869" from-port="4" to-layer="9871" to-port="0" /> <edge from-layer="9870" from-port="0" to-layer="9871" to-port="1" /> <edge from-layer="9871" from-port="2" to-layer="9875" to-port="0" /> <edge from-layer="9872" from-port="0" to-layer="9874" to-port="1" /> <edge from-layer="9873" from-port="0" to-layer="9874" to-port="3" /> <edge from-layer="9874" from-port="4" to-layer="9875" to-port="1" /> <edge from-layer="9875" from-port="2" to-layer="9883" to-port="0" /> <edge from-layer="9876" from-port="0" to-layer="9879" to-port="0" /> <edge from-layer="9877" from-port="0" to-layer="9879" to-port="1" /> <edge from-layer="9878" from-port="0" to-layer="9879" to-port="3" /> <edge from-layer="9879" from-port="4" to-layer="9881" to-port="0" /> <edge from-layer="9880" from-port="0" to-layer="9881" to-port="2" /> <edge from-layer="9881" from-port="3" to-layer="9882" to-port="0" /> <edge from-layer="9882" from-port="2" to-layer="9915" to-port="1" /> <edge from-layer="9882" from-port="2" to-layer="9883" to-port="1" /> <edge from-layer="9883" from-port="2" to-layer="9884" to-port="1" /> <edge from-layer="9884" from-port="2" to-layer="9889" to-port="0" /> <edge from-layer="9885" from-port="0" to-layer="9888" to-port="1" /> <edge from-layer="9886" from-port="0" to-layer="9888" to-port="2" /> <edge from-layer="9887" from-port="0" to-layer="9888" to-port="3" /> <edge from-layer="9888" from-port="4" to-layer="9889" to-port="1" /> <edge from-layer="9889" from-port="2" to-layer="9983" to-port="0" /> <edge from-layer="9889" from-port="2" to-layer="9945" to-port="0" /> <edge from-layer="9890" from-port="0" to-layer="9893" to-port="1" /> <edge from-layer="9891" from-port="0" to-layer="9893" to-port="2" /> <edge from-layer="9892" from-port="0" to-layer="9893" to-port="3" /> <edge from-layer="9893" from-port="4" to-layer="9908" to-port="0" /> <edge from-layer="9893" from-port="4" to-layer="9897" to-port="0" /> <edge from-layer="9893" from-port="4" to-layer="9913" to-port="0" /> <edge from-layer="9893" from-port="4" to-layer="9894" to-port="0" /> <edge from-layer="9894" from-port="2" to-layer="9916" to-port="0" /> <edge from-layer="9895" from-port="0" to-layer="9905" to-port="0" /> <edge from-layer="9896" from-port="0" to-layer="9905" to-port="1" /> <edge from-layer="9897" from-port="1" to-layer="9899" to-port="0" /> <edge from-layer="9898" from-port="0" to-layer="9899" to-port="2" /> <edge from-layer="9899" from-port="3" to-layer="9900" to-port="0" /> <edge from-layer="9900" from-port="2" to-layer="9901" to-port="0" /> <edge from-layer="9901" from-port="1" to-layer="9903" to-port="0" /> <edge from-layer="9902" from-port="0" to-layer="9903" to-port="1" /> <edge from-layer="9903" from-port="2" to-layer="9905" to-port="2" /> <edge from-layer="9904" from-port="0" to-layer="9905" to-port="3" /> <edge from-layer="9905" from-port="4" to-layer="9908" to-port="1" /> <edge from-layer="9905" from-port="4" to-layer="9913" to-port="2" /> <edge from-layer="9906" from-port="0" to-layer="9908" to-port="2" /> <edge from-layer="9907" from-port="0" to-layer="9908" to-port="3" /> <edge from-layer="9908" from-port="4" to-layer="9910" to-port="0" /> <edge from-layer="9909" from-port="0" to-layer="9910" to-port="1" /> <edge from-layer="9910" from-port="2" to-layer="9914" to-port="0" /> <edge from-layer="9911" from-port="0" to-layer="9913" to-port="1" /> <edge from-layer="9912" from-port="0" to-layer="9913" to-port="3" /> <edge from-layer="9913" from-port="4" to-layer="9914" to-port="1" /> <edge from-layer="9914" from-port="2" to-layer="9915" to-port="0" /> <edge from-layer="9915" from-port="2" to-layer="9916" to-port="1" /> <edge from-layer="9916" from-port="2" to-layer="9921" to-port="0" /> <edge from-layer="9917" from-port="0" to-layer="9920" to-port="1" /> <edge from-layer="9918" from-port="0" to-layer="9920" to-port="2" /> <edge from-layer="9919" from-port="0" to-layer="9920" to-port="3" /> <edge from-layer="9920" from-port="4" to-layer="9921" to-port="1" /> <edge from-layer="9921" from-port="2" to-layer="9922" to-port="1" /> <edge from-layer="9922" from-port="2" to-layer="9983" to-port="1" /> <edge from-layer="9922" from-port="2" to-layer="9942" to-port="0" /> <edge from-layer="9922" from-port="2" to-layer="10099" to-port="0" /> <edge from-layer="9923" from-port="0" to-layer="9928" to-port="0" /> <edge from-layer="9924" from-port="0" to-layer="9927" to-port="1" /> <edge from-layer="9925" from-port="0" to-layer="9927" to-port="2" /> <edge from-layer="9926" from-port="0" to-layer="9927" to-port="3" /> <edge from-layer="9927" from-port="4" to-layer="9928" to-port="1" /> <edge from-layer="9928" from-port="2" to-layer="9929" to-port="0" /> <edge from-layer="9929" from-port="1" to-layer="9931" to-port="0" /> <edge from-layer="9930" from-port="0" to-layer="9931" to-port="2" /> <edge from-layer="9931" from-port="3" to-layer="9938" to-port="0" /> <edge from-layer="9932" from-port="0" to-layer="9935" to-port="1" /> <edge from-layer="9933" from-port="0" to-layer="9935" to-port="2" /> <edge from-layer="9934" from-port="0" to-layer="9935" to-port="3" /> <edge from-layer="9935" from-port="4" to-layer="9937" to-port="0" /> <edge from-layer="9936" from-port="0" to-layer="9937" to-port="1" /> <edge from-layer="9937" from-port="2" to-layer="9938" to-port="1" /> <edge from-layer="9938" from-port="2" to-layer="9983" to-port="2" /> <edge from-layer="9938" from-port="2" to-layer="10098" to-port="0" /> <edge from-layer="9939" from-port="0" to-layer="9957" to-port="0" /> <edge from-layer="9940" from-port="0" to-layer="9952" to-port="0" /> <edge from-layer="9941" from-port="0" to-layer="9952" to-port="1" /> <edge from-layer="9941" from-port="0" to-layer="9955" to-port="1" /> <edge from-layer="9942" from-port="1" to-layer="9944" to-port="0" /> <edge from-layer="9943" from-port="0" to-layer="9944" to-port="2" /> <edge from-layer="9944" from-port="3" to-layer="9954" to-port="0" /> <edge from-layer="9944" from-port="3" to-layer="9948" to-port="0" /> <edge from-layer="9945" from-port="1" to-layer="9947" to-port="0" /> <edge from-layer="9945" from-port="1" to-layer="9972" to-port="0" /> <edge from-layer="9946" from-port="0" to-layer="9947" to-port="2" /> <edge from-layer="9947" from-port="3" to-layer="9948" to-port="1" /> <edge from-layer="9948" from-port="2" to-layer="9950" to-port="0" /> <edge from-layer="9949" from-port="0" to-layer="9950" to-port="1" /> <edge from-layer="9949" from-port="0" to-layer="9954" to-port="1" /> <edge from-layer="9950" from-port="2" to-layer="9952" to-port="2" /> <edge from-layer="9951" from-port="0" to-layer="9952" to-port="3" /> <edge from-layer="9951" from-port="0" to-layer="9955" to-port="3" /> <edge from-layer="9952" from-port="4" to-layer="9957" to-port="1" /> <edge from-layer="9953" from-port="0" to-layer="9955" to-port="0" /> <edge from-layer="9954" from-port="2" to-layer="9955" to-port="2" /> <edge from-layer="9954" from-port="2" to-layer="9962" to-port="2" /> <edge from-layer="9955" from-port="4" to-layer="9957" to-port="2" /> <edge from-layer="9956" from-port="0" to-layer="9957" to-port="3" /> <edge from-layer="9957" from-port="4" to-layer="9964" to-port="0" /> <edge from-layer="9958" from-port="0" to-layer="9964" to-port="1" /> <edge from-layer="9959" from-port="0" to-layer="9962" to-port="0" /> <edge from-layer="9960" from-port="0" to-layer="9962" to-port="1" /> <edge from-layer="9961" from-port="0" to-layer="9962" to-port="3" /> <edge from-layer="9962" from-port="4" to-layer="9964" to-port="2" /> <edge from-layer="9963" from-port="0" to-layer="9964" to-port="3" /> <edge from-layer="9964" from-port="4" to-layer="9967" to-port="0" /> <edge from-layer="9965" from-port="0" to-layer="9967" to-port="1" /> <edge from-layer="9966" from-port="0" to-layer="9967" to-port="2" /> <edge from-layer="9967" from-port="3" to-layer="9968" to-port="0" /> <edge from-layer="9968" from-port="2" to-layer="9982" to-port="0" /> <edge from-layer="9968" from-port="2" to-layer="9969" to-port="0" /> <edge from-layer="9969" from-port="1" to-layer="9981" to-port="0" /> <edge from-layer="9970" from-port="0" to-layer="9972" to-port="1" /> <edge from-layer="9971" from-port="0" to-layer="9972" to-port="2" /> <edge from-layer="9972" from-port="3" to-layer="9976" to-port="0" /> <edge from-layer="9973" from-port="0" to-layer="9976" to-port="1" /> <edge from-layer="9974" from-port="0" to-layer="9976" to-port="2" /> <edge from-layer="9975" from-port="0" to-layer="9976" to-port="3" /> <edge from-layer="9976" from-port="4" to-layer="9978" to-port="0" /> <edge from-layer="9976" from-port="4" to-layer="9980" to-port="2" /> <edge from-layer="9977" from-port="0" to-layer="9978" to-port="1" /> <edge from-layer="9978" from-port="2" to-layer="9980" to-port="0" /> <edge from-layer="9979" from-port="0" to-layer="9980" to-port="1" /> <edge from-layer="9980" from-port="3" to-layer="9981" to-port="1" /> <edge from-layer="9981" from-port="2" to-layer="9982" to-port="1" /> <edge from-layer="9982" from-port="2" to-layer="9983" to-port="3" /> <edge from-layer="9983" from-port="4" to-layer="9985" to-port="0" /> <edge from-layer="9984" from-port="0" to-layer="9985" to-port="1" /> <edge from-layer="9985" from-port="2" to-layer="9987" to-port="0" /> <edge from-layer="9986" from-port="0" to-layer="9987" to-port="1" /> <edge from-layer="9987" from-port="2" to-layer="9995" to-port="0" /> <edge from-layer="9988" from-port="0" to-layer="9989" to-port="0" /> <edge from-layer="9989" from-port="1" to-layer="9992" to-port="0" /> <edge from-layer="9990" from-port="0" to-layer="9991" to-port="0" /> <edge from-layer="9991" from-port="1" to-layer="9992" to-port="1" /> <edge from-layer="9992" from-port="2" to-layer="9994" to-port="0" /> <edge from-layer="9993" from-port="0" to-layer="9994" to-port="1" /> <edge from-layer="9994" from-port="2" to-layer="9995" to-port="1" /> <edge from-layer="9995" from-port="2" to-layer="9997" to-port="0" /> <edge from-layer="9996" from-port="0" to-layer="9997" to-port="1" /> <edge from-layer="9997" from-port="2" to-layer="9998" to-port="1" /> <edge from-layer="9998" from-port="2" to-layer="9999" to-port="0" /> <edge from-layer="9999" from-port="2" to-layer="10001" to-port="0" /> <edge from-layer="9999" from-port="2" to-layer="10175" to-port="0" /> <edge from-layer="9999" from-port="2" to-layer="10299" to-port="1" /> <edge from-layer="10000" from-port="0" to-layer="10001" to-port="1" /> <edge from-layer="10001" from-port="2" to-layer="10003" to-port="0" /> <edge from-layer="10002" from-port="0" to-layer="10003" to-port="1" /> <edge from-layer="10003" from-port="2" to-layer="10005" to-port="0" /> <edge from-layer="10004" from-port="0" to-layer="10005" to-port="1" /> <edge from-layer="10005" from-port="2" to-layer="10013" to-port="0" /> <edge from-layer="10006" from-port="0" to-layer="10007" to-port="0" /> <edge from-layer="10007" from-port="1" to-layer="10010" to-port="0" /> <edge from-layer="10008" from-port="0" to-layer="10009" to-port="0" /> <edge from-layer="10009" from-port="1" to-layer="10010" to-port="1" /> <edge from-layer="10010" from-port="2" to-layer="10012" to-port="0" /> <edge from-layer="10011" from-port="0" to-layer="10012" to-port="1" /> <edge from-layer="10012" from-port="2" to-layer="10013" to-port="1" /> <edge from-layer="10013" from-port="2" to-layer="10015" to-port="0" /> <edge from-layer="10014" from-port="0" to-layer="10015" to-port="1" /> <edge from-layer="10015" from-port="2" to-layer="10017" to-port="0" /> <edge from-layer="10016" from-port="0" to-layer="10017" to-port="1" /> <edge from-layer="10017" from-port="2" to-layer="10021" to-port="0" /> <edge from-layer="10017" from-port="2" to-layer="10038" to-port="0" /> <edge from-layer="10017" from-port="2" to-layer="10204" to-port="0" /> <edge from-layer="10018" from-port="0" to-layer="10021" to-port="1" /> <edge from-layer="10019" from-port="0" to-layer="10021" to-port="2" /> <edge from-layer="10020" from-port="0" to-layer="10021" to-port="3" /> <edge from-layer="10021" from-port="4" to-layer="10023" to-port="0" /> <edge from-layer="10022" from-port="0" to-layer="10023" to-port="1" /> <edge from-layer="10023" from-port="2" to-layer="10024" to-port="1" /> <edge from-layer="10024" from-port="2" to-layer="10283" to-port="2" /> <edge from-layer="10024" from-port="2" to-layer="10025" to-port="0" /> <edge from-layer="10026" from-port="0" to-layer="10031" to-port="0" /> <edge from-layer="10027" from-port="0" to-layer="10030" to-port="1" /> <edge from-layer="10028" from-port="0" to-layer="10030" to-port="2" /> <edge from-layer="10029" from-port="0" to-layer="10030" to-port="3" /> <edge from-layer="10030" from-port="4" to-layer="10031" to-port="1" /> <edge from-layer="10031" from-port="2" to-layer="10032" to-port="0" /> <edge from-layer="10032" from-port="1" to-layer="10034" to-port="0" /> <edge from-layer="10033" from-port="0" to-layer="10034" to-port="2" /> <edge from-layer="10034" from-port="3" to-layer="10096" to-port="0" /> <edge from-layer="10034" from-port="3" to-layer="10050" to-port="0" /> <edge from-layer="10035" from-port="0" to-layer="10038" to-port="1" /> <edge from-layer="10036" from-port="0" to-layer="10038" to-port="2" /> <edge from-layer="10037" from-port="0" to-layer="10038" to-port="3" /> <edge from-layer="10038" from-port="4" to-layer="10040" to-port="0" /> <edge from-layer="10039" from-port="0" to-layer="10040" to-port="1" /> <edge from-layer="10040" from-port="2" to-layer="10047" to-port="0" /> <edge from-layer="10040" from-port="2" to-layer="10044" to-port="0" /> <edge from-layer="10040" from-port="2" to-layer="10094" to-port="0" /> <edge from-layer="10041" from-port="0" to-layer="10044" to-port="1" /> <edge from-layer="10042" from-port="0" to-layer="10044" to-port="2" /> <edge from-layer="10043" from-port="0" to-layer="10044" to-port="3" /> <edge from-layer="10044" from-port="4" to-layer="10064" to-port="0" /> <edge from-layer="10044" from-port="4" to-layer="10061" to-port="0" /> <edge from-layer="10044" from-port="4" to-layer="10075" to-port="0" /> <edge from-layer="10044" from-port="4" to-layer="10080" to-port="0" /> <edge from-layer="10045" from-port="0" to-layer="10057" to-port="0" /> <edge from-layer="10046" from-port="0" to-layer="10057" to-port="1" /> <edge from-layer="10047" from-port="1" to-layer="10049" to-port="0" /> <edge from-layer="10048" from-port="0" to-layer="10049" to-port="2" /> <edge from-layer="10049" from-port="3" to-layer="10053" to-port="0" /> <edge from-layer="10050" from-port="1" to-layer="10052" to-port="0" /> <edge from-layer="10051" from-port="0" to-layer="10052" to-port="2" /> <edge from-layer="10052" from-port="3" to-layer="10053" to-port="1" /> <edge from-layer="10053" from-port="2" to-layer="10055" to-port="0" /> <edge from-layer="10054" from-port="0" to-layer="10055" to-port="1" /> <edge from-layer="10055" from-port="2" to-layer="10085" to-port="2" /> <edge from-layer="10055" from-port="2" to-layer="10057" to-port="2" /> <edge from-layer="10056" from-port="0" to-layer="10057" to-port="3" /> <edge from-layer="10057" from-port="4" to-layer="10059" to-port="0" /> <edge from-layer="10058" from-port="0" to-layer="10059" to-port="2" /> <edge from-layer="10059" from-port="3" to-layer="10060" to-port="0" /> <edge from-layer="10060" from-port="2" to-layer="10061" to-port="1" /> <edge from-layer="10060" from-port="2" to-layer="10211" to-port="1" /> <edge from-layer="10061" from-port="2" to-layer="10090" to-port="0" /> <edge from-layer="10062" from-port="0" to-layer="10072" to-port="0" /> <edge from-layer="10063" from-port="0" to-layer="10072" to-port="1" /> <edge from-layer="10064" from-port="1" to-layer="10066" to-port="0" /> <edge from-layer="10065" from-port="0" to-layer="10066" to-port="2" /> <edge from-layer="10066" from-port="3" to-layer="10067" to-port="0" /> <edge from-layer="10067" from-port="2" to-layer="10068" to-port="0" /> <edge from-layer="10068" from-port="1" to-layer="10070" to-port="0" /> <edge from-layer="10069" from-port="0" to-layer="10070" to-port="1" /> <edge from-layer="10070" from-port="2" to-layer="10072" to-port="2" /> <edge from-layer="10071" from-port="0" to-layer="10072" to-port="3" /> <edge from-layer="10072" from-port="4" to-layer="10080" to-port="2" /> <edge from-layer="10072" from-port="4" to-layer="10075" to-port="1" /> <edge from-layer="10073" from-port="0" to-layer="10075" to-port="2" /> <edge from-layer="10074" from-port="0" to-layer="10075" to-port="3" /> <edge from-layer="10075" from-port="4" to-layer="10077" to-port="0" /> <edge from-layer="10076" from-port="0" to-layer="10077" to-port="1" /> <edge from-layer="10077" from-port="2" to-layer="10081" to-port="0" /> <edge from-layer="10078" from-port="0" to-layer="10080" to-port="1" /> <edge from-layer="10079" from-port="0" to-layer="10080" to-port="3" /> <edge from-layer="10080" from-port="4" to-layer="10081" to-port="1" /> <edge from-layer="10081" from-port="2" to-layer="10089" to-port="0" /> <edge from-layer="10082" from-port="0" to-layer="10085" to-port="0" /> <edge from-layer="10083" from-port="0" to-layer="10085" to-port="1" /> <edge from-layer="10084" from-port="0" to-layer="10085" to-port="3" /> <edge from-layer="10085" from-port="4" to-layer="10087" to-port="0" /> <edge from-layer="10086" from-port="0" to-layer="10087" to-port="2" /> <edge from-layer="10087" from-port="3" to-layer="10088" to-port="0" /> <edge from-layer="10088" from-port="2" to-layer="10089" to-port="1" /> <edge from-layer="10088" from-port="2" to-layer="10232" to-port="1" /> <edge from-layer="10089" from-port="2" to-layer="10090" to-port="1" /> <edge from-layer="10090" from-port="2" to-layer="10095" to-port="0" /> <edge from-layer="10091" from-port="0" to-layer="10094" to-port="1" /> <edge from-layer="10092" from-port="0" to-layer="10094" to-port="2" /> <edge from-layer="10093" from-port="0" to-layer="10094" to-port="3" /> <edge from-layer="10094" from-port="4" to-layer="10095" to-port="1" /> <edge from-layer="10095" from-port="2" to-layer="10096" to-port="1" /> <edge from-layer="10096" from-port="2" to-layer="10097" to-port="0" /> <edge from-layer="10096" from-port="2" to-layer="10242" to-port="0" /> <edge from-layer="10096" from-port="2" to-layer="10283" to-port="1" /> <edge from-layer="10174" from-port="0" to-layer="10175" to-port="1" /> <edge from-layer="10175" from-port="2" to-layer="10177" to-port="0" /> <edge from-layer="10176" from-port="0" to-layer="10177" to-port="1" /> <edge from-layer="10177" from-port="2" to-layer="10179" to-port="0" /> <edge from-layer="10178" from-port="0" to-layer="10179" to-port="1" /> <edge from-layer="10179" from-port="2" to-layer="10187" to-port="0" /> <edge from-layer="10180" from-port="0" to-layer="10181" to-port="0" /> <edge from-layer="10181" from-port="1" to-layer="10184" to-port="0" /> <edge from-layer="10182" from-port="0" to-layer="10183" to-port="0" /> <edge from-layer="10183" from-port="1" to-layer="10184" to-port="1" /> <edge from-layer="10184" from-port="2" to-layer="10186" to-port="0" /> <edge from-layer="10185" from-port="0" to-layer="10186" to-port="1" /> <edge from-layer="10186" from-port="2" to-layer="10187" to-port="1" /> <edge from-layer="10187" from-port="2" to-layer="10189" to-port="0" /> <edge from-layer="10188" from-port="0" to-layer="10189" to-port="1" /> <edge from-layer="10189" from-port="2" to-layer="10190" to-port="0" /> <edge from-layer="10190" from-port="1" to-layer="10198" to-port="0" /> <edge from-layer="10191" from-port="0" to-layer="10192" to-port="0" /> <edge from-layer="10192" from-port="1" to-layer="10195" to-port="0" /> <edge from-layer="10193" from-port="0" to-layer="10194" to-port="0" /> <edge from-layer="10194" from-port="1" to-layer="10195" to-port="1" /> <edge from-layer="10195" from-port="2" to-layer="10197" to-port="0" /> <edge from-layer="10196" from-port="0" to-layer="10197" to-port="1" /> <edge from-layer="10197" from-port="2" to-layer="10198" to-port="1" /> <edge from-layer="10198" from-port="2" to-layer="10200" to-port="0" /> <edge from-layer="10199" from-port="0" to-layer="10200" to-port="1" /> <edge from-layer="10200" from-port="2" to-layer="10298" to-port="0" /> <edge from-layer="10201" from-port="0" to-layer="10204" to-port="1" /> <edge from-layer="10202" from-port="0" to-layer="10204" to-port="2" /> <edge from-layer="10203" from-port="0" to-layer="10204" to-port="3" /> <edge from-layer="10204" from-port="4" to-layer="10206" to-port="0" /> <edge from-layer="10205" from-port="0" to-layer="10206" to-port="1" /> <edge from-layer="10206" from-port="2" to-layer="10237" to-port="0" /> <edge from-layer="10206" from-port="2" to-layer="10210" to-port="0" /> <edge from-layer="10207" from-port="0" to-layer="10210" to-port="1" /> <edge from-layer="10208" from-port="0" to-layer="10210" to-port="2" /> <edge from-layer="10209" from-port="0" to-layer="10210" to-port="3" /> <edge from-layer="10210" from-port="4" to-layer="10225" to-port="0" /> <edge from-layer="10210" from-port="4" to-layer="10230" to-port="0" /> <edge from-layer="10210" from-port="4" to-layer="10211" to-port="0" /> <edge from-layer="10210" from-port="4" to-layer="10214" to-port="0" /> <edge from-layer="10211" from-port="2" to-layer="10233" to-port="0" /> <edge from-layer="10212" from-port="0" to-layer="10222" to-port="0" /> <edge from-layer="10213" from-port="0" to-layer="10222" to-port="1" /> <edge from-layer="10214" from-port="1" to-layer="10216" to-port="0" /> <edge from-layer="10215" from-port="0" to-layer="10216" to-port="2" /> <edge from-layer="10216" from-port="3" to-layer="10217" to-port="0" /> <edge from-layer="10217" from-port="2" to-layer="10218" to-port="0" /> <edge from-layer="10218" from-port="1" to-layer="10220" to-port="0" /> <edge from-layer="10219" from-port="0" to-layer="10220" to-port="1" /> <edge from-layer="10220" from-port="2" to-layer="10222" to-port="2" /> <edge from-layer="10221" from-port="0" to-layer="10222" to-port="3" /> <edge from-layer="10222" from-port="4" to-layer="10230" to-port="2" /> <edge from-layer="10222" from-port="4" to-layer="10225" to-port="1" /> <edge from-layer="10223" from-port="0" to-layer="10225" to-port="2" /> <edge from-layer="10224" from-port="0" to-layer="10225" to-port="3" /> <edge from-layer="10225" from-port="4" to-layer="10227" to-port="0" /> <edge from-layer="10226" from-port="0" to-layer="10227" to-port="1" /> <edge from-layer="10227" from-port="2" to-layer="10231" to-port="0" /> <edge from-layer="10228" from-port="0" to-layer="10230" to-port="1" /> <edge from-layer="10229" from-port="0" to-layer="10230" to-port="3" /> <edge from-layer="10230" from-port="4" to-layer="10231" to-port="1" /> <edge from-layer="10231" from-port="2" to-layer="10232" to-port="0" /> <edge from-layer="10232" from-port="2" to-layer="10233" to-port="1" /> <edge from-layer="10233" from-port="2" to-layer="10238" to-port="0" /> <edge from-layer="10234" from-port="0" to-layer="10237" to-port="1" /> <edge from-layer="10235" from-port="0" to-layer="10237" to-port="2" /> <edge from-layer="10236" from-port="0" to-layer="10237" to-port="3" /> <edge from-layer="10237" from-port="4" to-layer="10238" to-port="1" /> <edge from-layer="10238" from-port="2" to-layer="10245" to-port="0" /> <edge from-layer="10238" from-port="2" to-layer="10283" to-port="0" /> <edge from-layer="10239" from-port="0" to-layer="10257" to-port="0" /> <edge from-layer="10240" from-port="0" to-layer="10252" to-port="0" /> <edge from-layer="10241" from-port="0" to-layer="10252" to-port="1" /> <edge from-layer="10241" from-port="0" to-layer="10255" to-port="1" /> <edge from-layer="10242" from-port="1" to-layer="10244" to-port="0" /> <edge from-layer="10243" from-port="0" to-layer="10244" to-port="2" /> <edge from-layer="10244" from-port="3" to-layer="10254" to-port="0" /> <edge from-layer="10244" from-port="3" to-layer="10248" to-port="0" /> <edge from-layer="10245" from-port="1" to-layer="10247" to-port="0" /> <edge from-layer="10245" from-port="1" to-layer="10272" to-port="0" /> <edge from-layer="10246" from-port="0" to-layer="10247" to-port="2" /> <edge from-layer="10247" from-port="3" to-layer="10248" to-port="1" /> <edge from-layer="10248" from-port="2" to-layer="10250" to-port="0" /> <edge from-layer="10249" from-port="0" to-layer="10254" to-port="1" /> <edge from-layer="10249" from-port="0" to-layer="10250" to-port="1" /> <edge from-layer="10250" from-port="2" to-layer="10252" to-port="2" /> <edge from-layer="10251" from-port="0" to-layer="10255" to-port="3" /> <edge from-layer="10251" from-port="0" to-layer="10252" to-port="3" /> <edge from-layer="10252" from-port="4" to-layer="10257" to-port="1" /> <edge from-layer="10253" from-port="0" to-layer="10255" to-port="0" /> <edge from-layer="10254" from-port="2" to-layer="10262" to-port="2" /> <edge from-layer="10254" from-port="2" to-layer="10255" to-port="2" /> <edge from-layer="10255" from-port="4" to-layer="10257" to-port="2" /> <edge from-layer="10256" from-port="0" to-layer="10257" to-port="3" /> <edge from-layer="10257" from-port="4" to-layer="10264" to-port="0" /> <edge from-layer="10258" from-port="0" to-layer="10264" to-port="1" /> <edge from-layer="10259" from-port="0" to-layer="10262" to-port="0" /> <edge from-layer="10260" from-port="0" to-layer="10262" to-port="1" /> <edge from-layer="10261" from-port="0" to-layer="10262" to-port="3" /> <edge from-layer="10262" from-port="4" to-layer="10264" to-port="2" /> <edge from-layer="10263" from-port="0" to-layer="10264" to-port="3" /> <edge from-layer="10264" from-port="4" to-layer="10267" to-port="0" /> <edge from-layer="10265" from-port="0" to-layer="10267" to-port="1" /> <edge from-layer="10266" from-port="0" to-layer="10267" to-port="2" /> <edge from-layer="10267" from-port="3" to-layer="10268" to-port="0" /> <edge from-layer="10268" from-port="2" to-layer="10282" to-port="0" /> <edge from-layer="10268" from-port="2" to-layer="10269" to-port="0" /> <edge from-layer="10269" from-port="1" to-layer="10281" to-port="0" /> <edge from-layer="10270" from-port="0" to-layer="10272" to-port="1" /> <edge from-layer="10271" from-port="0" to-layer="10272" to-port="2" /> <edge from-layer="10272" from-port="3" to-layer="10276" to-port="0" /> <edge from-layer="10273" from-port="0" to-layer="10276" to-port="1" /> <edge from-layer="10274" from-port="0" to-layer="10276" to-port="2" /> <edge from-layer="10275" from-port="0" to-layer="10276" to-port="3" /> <edge from-layer="10276" from-port="4" to-layer="10278" to-port="0" /> <edge from-layer="10276" from-port="4" to-layer="10280" to-port="2" /> <edge from-layer="10277" from-port="0" to-layer="10278" to-port="1" /> <edge from-layer="10278" from-port="2" to-layer="10280" to-port="0" /> <edge from-layer="10279" from-port="0" to-layer="10280" to-port="1" /> <edge from-layer="10280" from-port="3" to-layer="10281" to-port="1" /> <edge from-layer="10281" from-port="2" to-layer="10282" to-port="1" /> <edge from-layer="10282" from-port="2" to-layer="10283" to-port="3" /> <edge from-layer="10283" from-port="4" to-layer="10285" to-port="0" /> <edge from-layer="10284" from-port="0" to-layer="10285" to-port="1" /> <edge from-layer="10285" from-port="2" to-layer="10287" to-port="0" /> <edge from-layer="10286" from-port="0" to-layer="10287" to-port="1" /> <edge from-layer="10287" from-port="2" to-layer="10295" to-port="0" /> <edge from-layer="10288" from-port="0" to-layer="10289" to-port="0" /> <edge from-layer="10289" from-port="1" to-layer="10292" to-port="0" /> <edge from-layer="10290" from-port="0" to-layer="10291" to-port="0" /> <edge from-layer="10291" from-port="1" to-layer="10292" to-port="1" /> <edge from-layer="10292" from-port="2" to-layer="10294" to-port="0" /> <edge from-layer="10293" from-port="0" to-layer="10294" to-port="1" /> <edge from-layer="10294" from-port="2" to-layer="10295" to-port="1" /> <edge from-layer="10295" from-port="2" to-layer="10297" to-port="0" /> <edge from-layer="10296" from-port="0" to-layer="10297" to-port="1" /> <edge from-layer="10297" from-port="2" to-layer="10298" to-port="1" /> <edge from-layer="10298" from-port="2" to-layer="10299" to-port="0" /> <edge from-layer="10299" from-port="2" to-layer="10301" to-port="0" /> <edge from-layer="10300" from-port="0" to-layer="10301" to-port="1" /> <edge from-layer="10301" from-port="2" to-layer="10303" to-port="0" /> <edge from-layer="10302" from-port="0" to-layer="10303" to-port="1" /> <edge from-layer="10303" from-port="2" to-layer="10305" to-port="0" /> <edge from-layer="10304" from-port="0" to-layer="10305" to-port="1" /> <edge from-layer="10305" from-port="2" to-layer="10313" to-port="0" /> <edge from-layer="10306" from-port="0" to-layer="10307" to-port="0" /> <edge from-layer="10307" from-port="1" to-layer="10310" to-port="0" /> <edge from-layer="10308" from-port="0" to-layer="10309" to-port="0" /> <edge from-layer="10309" from-port="1" to-layer="10310" to-port="1" /> <edge from-layer="10310" from-port="2" to-layer="10312" to-port="0" /> <edge from-layer="10311" from-port="0" to-layer="10312" to-port="1" /> <edge from-layer="10312" from-port="2" to-layer="10313" to-port="1" /> <edge from-layer="10313" from-port="2" to-layer="10314" to-port="0" /> </edges> <rt_info> <Runtime_version value="2023.3.0-13775-ceeafaf64f3-releases/2023/3" /> <conversion_parameters> <framework value="pytorch" /> <is_python_object value="True" /> </conversion_parameters> <nncf> <weight_compression> <all_layers value="False" /> <group_size value="-1" /> <ignored_scope> <names value="[]" /> <patterns value="[]" /> <types value="[]" /> <validate value="True" /> </ignored_scope> <mode value="int8_asym" /> <ratio value="1" /> <sensitivity_metric value="weight_quantization_error" /> </weight_compression> </nncf> </rt_info> </net>