quark.onnx.tools.convert_quant_to_float

Contents

`quark.onnx.tools.convert_quant_to_float`#

Convert quantized model to FP32 model.

Module Contents#

Functions#

convert_initializers_to_float(→ onnx.ModelProto)

Convert integer initializers used by DequantizeLinear nodes to float initializers.

quark.onnx.tools.convert_quant_to_float.convert_initializers_to_float(model: onnx.ModelProto, initializers_to_convert: Dict[str, Dict[str, str]]) → onnx.ModelProto#: Convert integer initializers used by DequantizeLinear nodes to float initializers.