fork intelligence · huggingface/transformers

What’s living in huggingface/transformers’s forks?

🤗 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. Scanned 34,331 forks, 40 inspected in depth, 10 carrying real unmerged work.

forksplorer.com/repo/huggingface/transformerscached scan · 2026-08-23

▲ scanned huggingface/transformers — 34,331 forks · 40 inspected · 6 features surfaced

#1 added option to not load clip vision model midjourney/transformers · 2 ahead · last push 2023-09-01

ac71359 added option to not load clip vision model — gadgetsam $ git cherry-pick ac71359 → view fork on GitHub

#2 Added missing continuity.py file. samuelemarro/continuous-llm-experiments · 16 ahead · last push 2025-05-12

6c60964 Added missing continuity.py file. — samuelemarro 7001af5 Added arXiv link. — samuelemarro cca073a Added charts for the various analysis results. — samuelemarro 3da6950 Added counting events data. — samuelemarro 6886e6f Added datasets. — samuelemarro $ git cherry-pick 6886e6f^..6c60964 → view fork on GitHub

#3 Add swiglu_limit = 7.0 unslothai/transformers · 2 ahead · last push 2025-08-15

411ab01 Add swiglu_limit = 7.0 — danielhanchen $ git cherry-pick 411ab01 → view fork on GitHub

#4 add the possibility to disable custom kernels thomwolf/transformers · 84 ahead · last push 2024-02-20

166dfa8 add the possibility to disable custom kernels — OlivierDehaene 6266815 add the possibility to disable custom kernels — OlivierDehaene f6d7746 support tensor parallelism for t5 — OlivierDehaene 76a35c5 add custom kernel for gpt-neox — OlivierDehaene 536e759 make it possible to install with pip install -e . — OlivierDehaene $ git cherry-pick 536e759^..166dfa8 → view fork on GitHub

#5 mixed fp16 init bug ROCm/transformers · 4 ahead · last push 2026-07-13

6978a60 mixed fp16 init bug — Cemberk fe25ce6 backward compatibility with HybridCache — Cemberk c929fd1 rocm skips — Cemberk 35f3f91 core changes — Cemberk $ git cherry-pick 35f3f91^..6978a60 → view fork on GitHub

#6 Customize Llama modeling code for OpenVLA-OFT: Full bidirectional attention moojink/transformers-openvla-oft · 4 ahead · last push 2025-02-25

bc339d9 Customize Llama modeling code for OpenVLA-OFT: Full bidirectional attention — moojink 9fe3f58 v4.40.1 — ArthurZucker f8fec6b Make EosTokenCriteria compatible with mps (#30376) — pcuenca 745bbfe Release: v4.40.0 — LysandreJik $ git cherry-pick 745bbfe^..bc339d9 → view fork on GitHub

View huggingface/transformers on GitHub · Cached scans refresh weekly — not a live per-request re-scan.