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