diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,3705 @@ +{ + "metadata": { + "ParamSize": 315, + "ParamBytes": 1470917664.0, + "BitsPerParam": 4.501064158865038 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 294914304, + "records": [ + { + "name": "model.embed_tokens.q_weight", + "shape": [ + 256002, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 294914304, + "byteOffset": 0 + } + ], + "md5sum": "be13c36eb75b1f061ce66374699fe9af" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 36864288, + "records": [ + { + "name": "model.embed_tokens.q_scale", + "shape": [ + 256002, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 36864288, + "byteOffset": 0 + } + ], + "md5sum": "95e87c44b980eeffa3c966836fe2fc05" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 33182208, + "records": [ + { + "name": "model.layers.0.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 0 + }, + { + "name": "model.layers.0.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 4608 + }, + { + "name": "model.layers.0.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10621440 + }, + { + "name": "model.layers.0.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11948544 + } + ], + "md5sum": "042e9d36c69b62cd45958aab1fd8d7a6" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.1.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "30cbaa72ad212914334e47fc8b92746f" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.0.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.0.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.0.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.0.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.0.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.0.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.0.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.0.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.1.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.1.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.1.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.1.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.1.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.1.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.1.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.1.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.1.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.1.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.1.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.10.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "36ad615704805641ea37151c7f64c70b" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.10.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.10.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.10.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "e29a40b60b2fcc5136db1a5c9d2096db" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.11.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "12d51f5f962dbefecff6c5b31d583df6" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.10.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.10.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.10.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.10.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.10.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.10.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.10.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.10.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.11.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.11.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.11.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.11.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.11.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.11.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.11.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.11.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.11.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.11.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.11.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.12.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "749df52b0b7190252c1be3d030419643" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.12.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.12.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.12.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "621f1aec3b20215b00c6c1d0215e926e" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.13.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "b1eb960fb74edabb98ecd08f54c7f2b2" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.12.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.12.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.12.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.12.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.12.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.12.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.12.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.12.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.13.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.13.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.13.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.13.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.13.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.13.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.13.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.13.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.13.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.13.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.13.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.14.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "cacce6b20e6f25acb3f26186c302de00" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.14.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.14.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.14.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "7badda5547ceef26ec44ffa930471cf1" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.15.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "107532e2c5c60c12ba5149aaeb3f610a" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.14.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.14.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.14.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.14.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.14.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.14.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.14.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.14.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.15.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.15.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.15.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.15.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.15.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.15.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.15.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.15.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.15.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.15.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.15.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.16.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "a1b7aa1e247995397438e1e72425ac2a" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.16.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.16.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.16.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "2ff0d4c8c83cb3d18f9b7d0e8abe9f22" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.17.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "21077c826ca532ccbd09cd0c1b193beb" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.16.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.16.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.16.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.16.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.16.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.16.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.16.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.16.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.17.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.17.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.17.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.17.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.17.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.17.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.17.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.17.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.17.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.17.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.17.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.18.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "2bef4981751d881a1eba3f58ea31d3af" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.18.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.18.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.18.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "d6859b1fa620dfb79c1c533950a2637e" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.19.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "dcea8743d8ec1a66c77885b9b66fbf73" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.18.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.18.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.18.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.18.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.18.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.18.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.18.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.18.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.19.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.19.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.19.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.19.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.19.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.19.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.19.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.19.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.19.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.19.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.19.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.2.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "5cd79680f58729d3d2a9ea3517ac9cf9" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.2.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.2.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.2.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "4730967683d7b6554ef9708e22a983cc" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.20.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "e2687846f89c07322d5ab55d55a9f6d3" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.2.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.2.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.2.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.2.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.2.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.2.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.2.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.2.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.20.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.20.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.20.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.20.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.20.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.20.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.20.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.20.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.20.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.20.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.20.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.21.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "be6fc05c5d0d02e57168359665663563" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.21.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.21.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.21.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "4332f98979d0604e63e38be065265b21" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.22.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "6f8fe14eef03a49dc7c4f76ab11226e0" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.21.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.21.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.21.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.21.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.21.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.21.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.21.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.21.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.22.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.22.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.22.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.22.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.22.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.22.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.22.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.22.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.22.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.22.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.22.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.23.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "4ac08a779071276922176ab8fa296834" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.23.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.23.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.23.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "771e74b1e5eb6f5ab1e631f706944245" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 31864320, + "records": [ + { + "name": "model.layers.23.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.23.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.23.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.23.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.23.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.23.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.23.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.23.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.24.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10630656 + } + ], + "md5sum": "2d157a920ca47988825a81333510b9ac" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.3.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "41a49f16845f1cbb91c552584330f83b" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 33200640, + "records": [ + { + "name": "model.layers.24.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.24.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2654208 + }, + { + "name": "model.layers.24.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7372800 + }, + { + "name": "model.layers.24.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7962624 + }, + { + "name": "model.layers.24.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10321920 + }, + { + "name": "model.layers.3.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10616832 + }, + { + "name": "model.layers.3.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10621440 + }, + { + "name": "model.layers.3.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21238272 + }, + { + "name": "model.layers.3.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22565376 + }, + { + "name": "model.layers.3.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25219584 + }, + { + "name": "model.layers.3.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25224192 + }, + { + "name": "model.layers.3.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25228800 + }, + { + "name": "model.layers.3.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25233408 + }, + { + "name": "model.layers.3.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29952000 + }, + { + "name": "model.layers.3.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30541824 + }, + { + "name": "model.layers.3.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32901120 + }, + { + "name": "model.layers.4.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33196032 + } + ], + "md5sum": "d055987bffaa0ae00b9df102316f8f2a" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.4.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.4.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.4.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "cdb681496c5fac03d6928770591b6d97" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.5.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "b5e1d09cafb10fd45d99acc3aa8b888b" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.4.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.4.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.4.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.4.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.4.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.4.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.4.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.4.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.5.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.5.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.5.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.5.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.5.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.5.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.5.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.5.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.5.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.5.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.5.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.6.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "ed3430d92c6cd362f38e9dccf2a3abf9" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.6.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.6.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.6.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "efce26070207cf7ed7fa13297924008e" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.7.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "2faf1b719fdb35d0b85e8c74fabc5a73" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.6.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.6.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.6.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.6.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.6.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.6.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.6.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.6.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.7.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.7.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.7.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.7.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.7.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.7.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.7.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.7.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.7.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.7.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.7.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.8.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "b2c2a3ff24e9de877dcec0e14738dbcc" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 33177600, + "records": [ + { + "name": "model.layers.8.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.8.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.8.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 11943936 + } + ], + "md5sum": "149670ab25bee6be377ab6029ff7b8fe" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.9.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "5e323ad7d26a2741381d9e12a3f42da7" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 33214464, + "records": [ + { + "name": "model.layers.8.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 0 + }, + { + "name": "model.layers.8.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "model.layers.8.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2658816 + }, + { + "name": "model.layers.8.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2663424 + }, + { + "name": "model.layers.8.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 2668032 + }, + { + "name": "model.layers.8.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 7386624 + }, + { + "name": "model.layers.8.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 7976448 + }, + { + "name": "model.layers.8.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 10335744 + }, + { + "name": "model.layers.9.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 10630656 + }, + { + "name": "model.layers.9.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 10635264 + }, + { + "name": "model.layers.9.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 21252096 + }, + { + "name": "model.layers.9.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 22579200 + }, + { + "name": "model.layers.9.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25233408 + }, + { + "name": "model.layers.9.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25238016 + }, + { + "name": "model.layers.9.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 25242624 + }, + { + "name": "model.layers.9.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 25247232 + }, + { + "name": "model.layers.9.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 29965824 + }, + { + "name": "model.layers.9.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 30555648 + }, + { + "name": "model.layers.9.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 32914944 + }, + { + "name": "model.layers.24.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 33209856 + } + ], + "md5sum": "a05c2a9bcfa7d26005833a473d81dc25" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.25.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "471f6747cb5d782d4c9693fb63f4f081" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 31882752, + "records": [ + { + "name": "model.layers.24.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.24.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 10616832 + }, + { + "name": "model.layers.24.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 11943936 + }, + { + "name": "model.layers.24.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 11948544 + }, + { + "name": "model.layers.24.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 11953152 + }, + { + "name": "model.layers.25.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 11957760 + }, + { + "name": "model.layers.25.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 11962368 + }, + { + "name": "model.layers.25.mlp.down_proj.q_scale", + "shape": [ + 2304, + 288 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1327104, + "byteOffset": 22579200 + }, + { + "name": "model.layers.25.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2654208, + "byteOffset": 23906304 + }, + { + "name": "model.layers.25.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26560512 + }, + { + "name": "model.layers.25.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26565120 + }, + { + "name": "model.layers.25.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26569728 + }, + { + "name": "model.layers.25.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 26574336 + }, + { + "name": "model.layers.25.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 589824, + "byteOffset": 31292928 + } + ], + "md5sum": "a81f75e4639071df7acbd7059791cb47" + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 2658816, + "records": [ + { + "name": "model.layers.25.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 0 + }, + { + "name": "model.layers.25.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 2359296 + }, + { + "name": "model.norm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + } + ], + "md5sum": "2288aeb8d0296331bcc21a23a7b42f3a" + } + ] +} \ No newline at end of file