transformers
4d1cc69d - Explain why MXFP4 defaults to DeepSeek's layout and W4A4 experts

Commit

9 days ago

Explain why MXFP4 defaults to DeepSeek's layout and W4A4 experts Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

References

#49778 - 🚨 MXFP4 on the fine-grained machinery, and DeepSeek-V4's FP8 + MXFP4 plan

#49779 - NVFP4 on the fine-grained machinery

#49780 - MXFP8 on the fine-grained machinery (`Mxfp8Config`)

#49823 - Load ModelOpt and compressed-tensors exports on the fine-grained formats

Author

IlyasMoutawwakil

IlyasMoutawwakil

Parents

Loading