View new feature information and release downloads for the latest and previous releases on GitHub. Validated configurations and distribution sites located here as well:
Contact email@example.com if you need additional assistance.
The MSE tuning strategy does not work with the PyTorch adaptor layer. This strategy requires a comparison between the FP32 and INT8 tensors to decide which op impacts the final quantization accuracy. The PyTorch adaptor layer does not implement this inspect tensor interface. Therefore, do not choose the MSE tuning strategy for PyTorch models.
The diagnosis function does not work with ONNX Runtime 1.13.1 for QDQ format quantization of ONNX models. It can not dump the output value of QDQ pairs since framework limitation.
Neural Compressor v1.2.1 solves this backward compatible issues introduced in v1.2 by moving new user facing APIs to neural_compressor.experimental package and keep old one as is. Please refer to API documentation to know the details of user-facing APIs.
Neural Compressor v1.7 renames the pip/conda package name from lpot to neural_compressor. To run old examples on latest software, please replace package name for compatibility with
sed -i "s|lpot|neural_compressor|g" your_script.py .
Neural Compressor v2.2 from this release, binary
neural-compressor-full is deprecated, we deliver 3 binaries named