🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
COMMITS
January 28, 2026
Y
check PR bot permission - part 3 (try content attribute) (#43555)
Yih-Dar committed
Y
check PR bot permission - part 2 (style only) (#43554)
Yih-Dar committed
Y
check PR bot permission - part 1 (#43553)
Yih-Dar committed
S
Fix failing tests due to no attribute `pad_token_id` (#43453)
Sai-Suraj-27 committed
K
fix: GPT OSS Conversion Script Enhancements (#42901)
Kyle Mylonakis committed
M
[Quantization] Fix triton_kernels name after being renamed to gpt-oss-triton-kernels (#43528)
Mohamed Mekkouri committed
M
[Quantization] Add cutlass kernel for FP8 (#43304)
Mohamed Mekkouri committed
R
[CB] Minor perf improvements and ty compatibility (#43521)
Rémi Ouazan committed
L
Add Youtu-LLM model (#43166)
Lu junru committed
A
Fix tiles mixing for batched input, add tie_word_embeddings to LFM2VL config (#43379)
Anna Banaszak committed
January 27, 2026
S
fix: return labels instead of label in reduce_label method in BeitImageProcessorFast (#43527)
StevenBucaille committed
A
[`RoPE`] Make explicit inheritance (#43410)
Anton Vlasjuk committed
M
Fix for #43530 (#43535)
Matt committed
M
Operation Green CI (#43530)
Matt committed
C
Tie the weights even if initializing from a config on meta device (#43523)
Cyril Vallez committed
M
[kernels] Update cv_utils name (#43529)
Mohamed Mekkouri committed
M
add trackio to training notebooks (#43442)
Merve Noyan committed
M
Mark test_prompt_lookup_decoding as flaky (#42184)
Matt committed
I
Fix some MoE routers (#43445)
Ilyas Moutawwakil committed
I
batched_mm is slow on cpu (#43438)
Ilyas Moutawwakil committed
T
fix: initialize BatchNorm2d buffers only when needed (#43520)
Tarek Ziade committed
Y
[GLM-OCR] GLM-OCR Support (#43391)
Yuxuan Zhang committed
G
Fix loading of Qwen3 FP8 (#43494)
githubnemo committed
W
fix `ShieldGemma2IntegrationTest::test_model` (#43343)
Wang, Yi committed
January 26, 2026
Y
Revert utils files changes from PR #42845 (#43507)
Yih-Dar committed
M
Move hardcoded time_step params to config for Bamba, FalconH1, GraniteMoeHybrid (#43461)
Murat Raimbekov committed
R
Prepare inputs for generation is called from `super()` (#43280)
Raushan Turganbay committed
Y
Enhance repo. consistency bot (#43503)
Yih-Dar committed
T
Add `pytest-random-order` for reproducible test randomization (#43483)
Tarek Ziade committed