support for tiledmlp for GPT-OSS (#3116)

* fix use of flex attn kwargs and add support for tiledmlp for GPT-OSS

* add logging back

* update deps
This commit is contained in:
Wing Lian
2025-08-29 13:52:49 -04:00
committed by GitHub
parent 7ed40f1d70
commit 0094a2d744
6 changed files with 144 additions and 162 deletions

View File

@@ -127,7 +127,7 @@ extras_require = {
"yunchang==0.6.0",
],
"deepspeed": [
"deepspeed==0.17.2",
"deepspeed==0.17.5",
"deepspeed-kernels",
],
"mamba-ssm": [