intel/neural-compressor

• highreadme#1

  Reposition README's main heading to highlight SOTA LLM quantization and Intel optimization.

  原因：

  当前

  <h3> An open-source Python library supporting popular model compression techniques on mainstream deep learning frameworks (PyTorch, TensorFlow, and JAX)</h3>

  复制粘贴的修复

  复制

  <h3> The leading open-source Python library for SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity, optimized for Intel hardware across PyTorch, TensorFlow, and ONNX Runtime.</h3>

  待办完成忽略
• mediumreadme#2

  Add a "Key Differentiators" section to the README.

  原因：

  复制粘贴的修复

  复制

  ## Key Differentiators
  
  Unlike generic quantization tools, Intel® Neural Compressor offers a unified, framework-agnostic approach to model optimization (especially quantization and pruning) with a strong emphasis on maximizing inference performance on Intel CPUs, GPUs, and other Intel hardware. We provide state-of-the-art low-bit LLM quantization techniques and comprehensive model compression for PyTorch, TensorFlow, and ONNX Runtime.

  待办完成忽略
• lowtopics#3

  Add specific LLM names and advanced quantization formats to topics.

  原因：

  当前

  auto-tuning, awq, fp4, gptq, int4, int8, knowledge-distillation, large-language-models, low-precision, mxformat, post-training-quantization, pruning, quantization, quantization-aware-training, smoothquant, sparsegpt, sparsity

  复制粘贴的修复

  复制

  auto-tuning, awq, deepseek, fp4, flux, framepack, gptq, int4, int8, knowledge-distillation, llama, large-language-models, low-precision, mxformat, nvfp4, post-training-quantization, pruning, quantization, quantization-aware-training, qwen, smoothquant, sparsegpt, sparsity

  待办完成忽略

本次扫描解析到的品类 GEO 通道：google/gemini-2.5-flash, deepseek/deepseek-v4-flash

• 品类问题

  How to apply state-of-the-art low-bit quantization for large language models?

  你：未被推荐
  AI 推荐顺序：

  1. bitsandbytes (TimDettmers/bitsandbytes)
  2. AWQ
  3. GPTQ
  4. Hugging Face Optimum (huggingface/optimum)
  5. ONNX Runtime (microsoft/onnxruntime)
  6. Intel OpenVINO (openvinotoolkit/openvino)
  7. NVIDIA TensorRT (NVIDIA/TensorRT)
  8. LLM.int8()
  9. SqueezeLLM
  10. PyTorch (pytorch/pytorch)

  AI 推荐了 10 个替代方案，却始终没点名 intel/neural-compressor。这就是要补上的差距。

  查看 AI 完整回答
• 品类问题

  What are effective model compression techniques for PyTorch and TensorFlow deep learning models?

  你：未被推荐
  AI 推荐顺序：

  1. torch.quantization
  2. tf.lite.TFLiteConverter
  3. TensorFlow Model Optimization Toolkit (tensorflow/model-optimization)
  4. torch.nn.utils.prune
  5. pytorch_model_pruning (IntelLabs/pytorch_model_pruning)
  6. torchdistill (yoshitomo-matsubara/torchdistill)
  7. keras.losses.KLDivergence
  8. tensorly (tensorly/tensorly)
  9. AutoGluon (awslabs/autogluon)
  10. AutoKeras (keras-team/autokeras)
  11. tf.keras.applications
  12. MobileNetV2
  13. EfficientNet
  14. torchvision.models
  15. MobileNetV3
  16. EfficientNet_B0
  17. EfficientNet_B7

  AI 推荐了 17 个替代方案，却始终没点名 intel/neural-compressor。这就是要补上的差距。

  查看 AI 完整回答

• Metadata completeness

  pass

• README presence

  pass

• Compared to common alternatives in this category, what is the core differentiator of intel/neural-compressor?

  pass
  ✓AI 明确点名了 intel/neural-compressor

  AI 的回答可能信誓旦旦却是错的。请按事实核对：技术栈、目标人群、差异化点是不是和你实际的对得上？

• If a team adopts intel/neural-compressor in production, what risks or prerequisites should they evaluate first?

  pass
  ✓AI 明确点名了 intel/neural-compressor

  AI 的回答可能信誓旦旦却是错的。请按事实核对：技术栈、目标人群、差异化点是不是和你实际的对得上？

• In one sentence, what problem does the repo intel/neural-compressor solve, and who is the primary audience?

  pass
  ⚠AI 未点名 intel/neural-compressor —— 很可能在说另一个项目

  AI 的回答可能信誓旦旦却是错的。请按事实核对：技术栈、目标人群、差异化点是不是和你实际的对得上？