intel neural-compressor 2722 Python https://avatars.githubusercontent.com/u/17888862?v=4 73 SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime
创建于 6 年前
最后更新 5 天前
AI 综合评分
73 B 级
代码质量 67
文档质量 76
社区健康 70
项目活跃 79
创新性 64
项目简介
SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression techniques on PyTorch, TensorFlow, and ONNX Runtime