{ "per_channel": false, "reduce_range": false, "per_model_config": { "encoder_model": { "op_types": [ "ReduceMean", "Pow", "Gather", "MatMul", "Shape", "Mul", "Softmax", "Reshape", "Conv", "Add", "Erf", "Sub", "Constant", "Div", "Sqrt", "Unsqueeze", "Concat", "Transpose" ], "weight_type": "QUInt8" }, "decoder_model": { "op_types": [ "Pow", "Expand", "Squeeze", "Sub", "MatMul", "Concat", "Equal", "Mul", "Cast", "Sqrt", "ConstantOfShape", "Softmax", "Add", "Constant", "Where", "Range", "Transpose", "Gather", "Slice", "ReduceMean", "Shape", "Reshape", "Erf", "Unsqueeze", "Div", "Less" ], "weight_type": "QInt8" }, "decoder_with_past_model": { "op_types": [ "Gather", "Slice", "ReduceMean", "Pow", "MatMul", "Shape", "Mul", "Softmax", "Reshape", "Erf", "Add", "Sub", "Sqrt", "Unsqueeze", "Constant", "Div", "Concat", "Transpose" ], "weight_type": "QInt8" }, "decoder_model_merged": { "op_types": [ "Pow", "Expand", "Squeeze", "Sub", "MatMul", "Concat", "If", "Equal", "Mul", "Cast", "Sqrt", "ConstantOfShape", "Softmax", "Add", "Constant", "Where", "Range", "Transpose", "Gather", "Slice", "ReduceMean", "Shape", "Reshape", "Erf", "Unsqueeze", "Div", "Less" ], "weight_type": "QInt8" } } }