neural_compressor.onnxrt.algorithms

Package Contents

Classes

Smoother

Fake input channel quantization.

Functions

apply_rtn_on_model(→ onnx.ModelProto)	Apply RTN on onnx model.
apply_gptq_on_model(→ onnx.ModelProto)	Apply GPTQ on onnx model.
apply_awq_on_model(→ onnx.ModelProto)	Apply Activation-aware Weight quantization(AWQ) on onnx model.
layer_wise_quant(...)	Quantize model layer by layer to save memory.