Intel Neural Compressor

license

Apache-2.0(OSI)

source

public

core features withheld

no

The LICENSE file contains the Apache-2.0 text. The repository is public and the README describes no paid tier, enterprise edition or license-gated build, so the core is ungated.

• https://raw.githubusercontent.com/intel/neural-compressor/main/LICENSE recorded 2026-09-30

  LICENSE file is the verbatim Apache-2.0 text, opening with "Apache License" and "Version 2.0, January 2004" and "TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION".

• https://raw.githubusercontent.com/intel/neural-compressor/main/README.md recorded 2026-09-30

  README opens with "An open-source Python library supporting popular model compression techniques on mainstream deep learning frameworks (PyTorch, TensorFlow, and JAX)" and carries the badge "license-Apache%202". Searched the body for enterprise, license key, paid, commercial, proprietary and closed: none.

• https://raw.githubusercontent.com/intel/neural-compressor/main/setup.py recorded 2026-09-30

  setup.py declares license="Apache 2.0" and the classifier "License :: OSI Approved :: Apache Software License". Searched the body for enterprise, license key, paid, commercial and proprietary: none.

Adoption is measured from PyPI downloads of the `neural-compressor` package.

• https://pypistats.org/api/packages/neural-compressor/recent recorded 2026-08-18

  last_month downloads = 33,576 for neural-compressor

Intel Neural Compressor covers static, dynamic, weight-only and quantization-aware quantization, SmoothQuant and mixed precision across PyTorch, TensorFlow and JAX, with an auto-tune workflow. That sits one step below Apache TVM, whose scope covers more of the category.

• https://raw.githubusercontent.com/intel/neural-compressor/main/docs/source/design.md recorded 2026-09-30

  Design doc says "Intel® Neural Compressor provides two workflows: Quantization and Auto-tune."

• https://raw.githubusercontent.com/intel/neural-compressor/main/docs/source/PT_QAT.md recorded 2026-09-30

  QAT page says "QAT explicitly simulates quantization during training, significantly narrowing the accuracy gap compared to pure Post-Training Quantization (PTQ)" and integrates "Microscaling (MX) formats" such as MXFP4 and MXFP8.

• https://raw.githubusercontent.com/intel/neural-compressor/main/docs/source/PT_WeightOnlyQuant.md recorded 2026-09-30

  Weight-only page opens with "weight only quantization is probably a better trade-off to balance the performance and the accuracy" and lists "GPTQ", "AutoRound", "AWQ" and "HQQ".

• https://raw.githubusercontent.com/intel/neural-compressor/main/docs/source/TF_Quant.md recorded 2026-09-30

  TensorFlow page has the section "With Accuracy Aware Tuning" for tuning out the best quantized model, and its intro reads "supports quantizing both TensorFlow and Keras model".

• https://raw.githubusercontent.com/intel/neural-compressor/main/README.md recorded 2026-09-30

  README says Intel Neural Compressor "aims to provide popular model compression techniques such as Static Quantization, Dynamic Quantization, SmoothQuant, Weight-Only Quantization, Quantization-Aware Training, Mixed Precision, etc." and offers "advanced quantization of Large Language Models (LLMs) and Vision-Language Models (VLMs)" with "Support a wide range of Intel hardware" and "support AMD CPU, ARM CPU, and NVidia GPU with limited testing".