finegrained-fp8: fused moe kernels#530

Draft

IlyasMoutawwakil wants to merge 3 commits intohuggingface:mainfrom

IlyasMoutawwakil:fp8-fused-moe

Member

IlyasMoutawwakil commented Apr 7, 2026 •

edited

Loading

moe_tflops_grouped

moe_tflops_batched

faster and more accurate fused fp8 for both long context (grouped) and decode (batched)

IlyasMoutawwakil added 3 commits

April 1, 2026 21:28


          first experiment

33bb509


          more configs


          final fused kernels

2e639af

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

danieldk Awaiting requested review from danieldk danieldk will be requested when the pull request is marked ready for review danieldk is a code owner

drbh Awaiting requested review from drbh drbh will be requested when the pull request is marked ready for review drbh is a code owner

At least 1 approving review is required to merge this pull request.

Labels

None yet