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