MXFP8

4 posts · 3 sources

Hugging FaceBlog1w

Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs

NVIDIABlog2w

Efficient MoE Training for Biological Foundation Models

CursorBlog6 Apr

Better MoE model inference with warp decode

CursorBlog29 Aug

1.5x faster MoE training with custom MXFP8 kernels