Fix mamba caches#40198

Closed

manueldeprada wants to merge 18 commits into

huggingface:mainfrom

manueldeprada:fix-mamba-caches2

Commits on Aug 5, 2025

Release: v4.55.0
LysandreJik
committed

Commits on Aug 7, 2025

[CI] post-`GptOss` fixes for green CI (#39929)

gante
authored and
LysandreJik
committed

Commits on Aug 13, 2025

Fix MXFP4 quantizer validation to allow CPU inference with dequantize option (#39953)

returnL
authored and
ArthurZucker
committed
Enable gpt-oss mxfp4 on older hardware (sm75+) (#39940)

authored and
ArthurZucker
committed
remove `triton_kernels` dep with `kernels` instead (#39926)

SunMarc
authored and
ArthurZucker
committed
[Idefics] fix device mismatch (#39981)

zucchini-nlp
authored and
ArthurZucker
committed
Fix missing video inputs for PerceptionLM. (#39971)

shuminghu
authored and
ArthurZucker
committed
Fix broken image inference for Fuyu model (#39915)

Isotr0py
authored and
ArthurZucker
committed
Fix missing None default values for Gemma3n model in get_placeholder_mask (#39991) (#40024)

authored and
ArthurZucker
committed
fix: resolve triton version check compatibility on windows (#39986)

authored and
ArthurZucker
committed
[`GPT Big Code`] Fix attention scaling (#40041)

vasqu
authored and
ArthurZucker
committed
Default to dequantize if cpu in device_map for mxfp4 (#39993)

MekkCyber
authored and
ArthurZucker
committed
fix merge conlicts

vasqu
authored and
ArthurZucker
committed
[bugfix] Fix tensor device in Idefics2, Idefics3, and SmolVLM (#39975)

qgallouedec
authored and
ArthurZucker
committed
v4.55.1
ArthurZucker
committed
qfix bad cherry-pick
ArthurZucker
committed
v4.55.2
ArthurZucker
committed

Commits on Aug 15, 2025

fix mamba models caches inheritance
manueldeprada
committed