Support gpt-oss#22259

Closed

zyongye wants to merge 71 commits intovllm-project:mainfrom

Commits on Jul 25, 2025

working integration with flash attention with sinks

LiuXiaoxuanPKU
authored and
simon-mo
committed
mxfp4 triton kernel integration

zyongye
authored and
simon-mo
committed
code clean up
simon-mo
committed
ran use_existing_torch.py
simon-mo
committed
new weight format works for tp1 and non-swizzling
simon-mo
committed
unpadded tp2 works

zyongye
authored and
simon-mo
committed
new format swizzling

zyongye
authored and
simon-mo
committed
add non-uniform tp sharding for moe layer

zyongye
authored and
simon-mo
committed
fotmating

zyongye
authored and
simon-mo
committed
comply with upstream triton_kernel refactor

zyongye
authored and
simon-mo
committed
working on blackwell

zyongye
authored and
simon-mo
committed
adding constraints to use persistent kernel

zyongye
authored and
simon-mo
committed
add epilogue_subtile constraint

zyongye
authored and
simon-mo
committed
update on latest triton change, use non-swizzling temporially for nan issue

zyongye
authored and
simon-mo
committed
rms default value

LiuXiaoxuanPKU
authored and
simon-mo
committed
swiglu limit

LiuXiaoxuanPKU
authored and
simon-mo
committed
triton kernel fixed, enable swizzling

zyongye
authored and
simon-mo
committed
Remove activation padding

zyongye
authored and
simon-mo
committed
new model path and make it configurable
simon-mo
committed
fix lint
simon-mo
committed

Commits on Jul 28, 2025

Commits on Jul 29, 2025

Commits on Jul 30, 2025

Commits on Aug 1, 2025

Commits on Aug 2, 2025

Commits on Aug 3, 2025

Commits on Aug 5, 2025

Commits on Aug 6, 2025