{ "per_channel": false, "reduce_range": false, "quantize_mode": "q8", "per_model_config": { "decoder_model": { "op_types": [ "Where", "Gather", "LayerNormalization", "Less", "Unsqueeze", "Reshape", "Squeeze", "Range", "Constant", "Tanh", "Shape", "Transpose", "Sub", "Concat", "ConstantOfShape", "MatMul", "Neg", "Slice", "Add", "Expand", "Softmax", "Div", "Cast", "Mul", "Pow", "Equal" ], "weight_type": "QInt8" }, "decoder_model_merged": { "op_types": [ "Where", "Gather", "LayerNormalization", "Less", "Unsqueeze", "Reshape", "Squeeze", "Range", "Constant", "Tanh", "Shape", "Transpose", "If", "Sub", "Concat", "ConstantOfShape", "MatMul", "Neg", "Slice", "Add", "Expand", "Softmax", "Div", "Cast", "Mul", "Pow", "Equal" ], "weight_type": "QInt8" }, "decoder_with_past_model": { "op_types": [ "Where", "Gather", "LayerNormalization", "Unsqueeze", "Reshape", "Range", "Constant", "Tanh", "Shape", "Transpose", "Sub", "Concat", "ConstantOfShape", "MatMul", "Neg", "Slice", "Add", "Expand", "Softmax", "Div", "Cast", "Mul", "Pow", "Equal" ], "weight_type": "QInt8" } } }