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

@@ -13,7 +13,7 @@ packaging==23.2
huggingface_hub>=0.33.0
peft>=0.17.0
transformers==4.55.3
transformers==4.55.4
tokenizers>=0.21.1
accelerate==1.10.0
datasets==4.0.0