quark.onnx.finetuning.auto_mixprecision

Contents

`quark.onnx.finetuning.auto_mixprecision`#

Module Contents#

Functions#

quark.onnx.finetuning.auto_mixprecision.auto_mixprecision(float_model_path: str, quant_model_path: str, dr: Any, activation_type: Any, weight_type: Any, extra_options: Any) → Any#: Automatic apply low precision quantization on Q/DQ.