Who contributed to Hugging Face on July 30, 2026?

13 developers shipped this update, including delmalih, stevhliu, akshan-main, jiqing-feng, Cyrilvallez, daviswer, Harry Mellor, and Hameed Ibrahim, and 5 more.

What were the notable Hugging Face updates?

byebye torch 2.4, Fix CUDA Graph breaking host to device copy from scalar tensor allocation (#47547), and Vectorize NoRepeatNGramLogitsProcessor and remove its host sync (#47571).

@huggingface

Transformers, Datasets, and the open AI-model layer

github ↗

Pick a date

Topics: Python AI / ML Full archive →

The Wire · Showcase

TRANSFORMERS DROPS TORCH 2.4 SUPPORT; DIFFUSERS FIXES LORA COMPILATION BUGS

By RepoJournal · Filed 07:00 UTC on July 30, 2026 · About Hugging Face

13 people shipped this

qgallouedec @qgallouedec 2 cited

jiqing-feng @jiqing-feng 1 cited

Cyrilvallez @Cyrilvallez 1 cited

daviswer @daviswer 1 cited

Harry Mellor @hmellor 1 cited

Hameed Ibrahim @hameedibrh 1 cited

ydshieh @ydshieh 1 cited

pjdurden @pjdurden 1 cited

delmalih @delmalih

stevhliu @stevhliu

akshan-main @akshan-main

thomwolf @thomwolf

lvwerra @lvwerra

Hugging Face is narrowing its PyTorch version floor while shipping critical fixes across diffusers, transformers, candle, and trl that address tensor allocation, batching, and generation pipeline bugs.

Transformers is ending torch 2.4 support [1], a housekeeping move that signals the team is moving past legacy PyTorch versions. More urgent: a CUDA Graph fix now makes Gemma 4 compilable by solving host-to-device copy issues from scalar tensor allocation [2], and a new vectorized NoRepeatNGramLogitsProcessor removes a performance-blocking host sync point [3]. Meanwhile, diffusers fixed a critical LoRA hot-swapping bug where models with `different_shapes_for_compilation` failed on recompile with tensor size mismatches [4]. Candle patched Qwen3 batching: causal masks were built batch-independently but reshaped as if fully batch-aware, causing batch rows past the first to read garbage and return wrong output [5]. On the feature side, transformers now supports Granite-swa and Granitemoe-swa models with sliding window attention and learnable per-head sinks [6], and a major Rotary module simplification removes redundant device/dtype casting [7]. TRL's DistillationTrainer refactor is in motion: generation switched to GRPO's stack with a deletion-heavy 193/537 PR [8], and a new `_get_last_hidden_state` method lands unwired but ready for chunked JSD loss [9].

Action items

→ Patch diffusers immediately if you use LoRA hot-swapping with different_shapes_for_compilation huggingface/diffusers [immediate]
→ Recompile Candle Qwen3 models after the batch mask fix if you rely on batch inference huggingface/candle [immediate]
→ Review Rotary module simplification in transformers; breaking change removes device/seq_len arguments huggingface/transformers [plan]
→ Watch TRL DistillationTrainer refactor stacking; generation cutover now live huggingface/trl [monitor]

References

[1] byebye torch 2.4 ↗ huggingface/transformers
[2] Fix CUDA Graph breaking host to device copy from scalar tensor allocation (#47547) huggingface/transformers
[3] Vectorize NoRepeatNGramLogitsProcessor and remove its host sync (#47571) huggingface/transformers
[4] Fix LoRA hot-swapping recompilation with `different_shapes_for_compilation` (#14297) huggingface/diffusers
[5] fix(qwen3): build causal mask batch-independently (#3582) ↗ huggingface/candle
[6] Add Granite-swa and Granitemoe-swa model support ↗ huggingface/transformers
[7] Simplify all Rotary modules ↗ huggingface/transformers
[8] [DistillationTrainer refactor] Switch generation to GRPO's stack; delete the buffer ↗ huggingface/trl
[9] [DistillationTrainer refactor] Add `_get_last_hidden_state` (unwired) ↗ huggingface/trl

Quick answers

What shipped in Hugging Face on July 30, 2026?: Hugging Face is narrowing its PyTorch version floor while shipping critical fixes across diffusers, transformers, candle, and trl that address tensor allocation, batching, and generation pipeline bugs. In total, 36 commits and 33 pull requests landed.
Who contributed to Hugging Face on July 30, 2026?: 13 developers shipped this update, including delmalih, stevhliu, akshan-main, jiqing-feng, Cyrilvallez, daviswer, Harry Mellor, and Hameed Ibrahim, and 5 more.
What were the notable Hugging Face updates?: byebye torch 2.4, Fix CUDA Graph breaking host to device copy from scalar tensor allocation (#47547), and Vectorize NoRepeatNGramLogitsProcessor and remove its host sync (#47571).

SPEECH-TO-SPEECH 0.2.12 SHIPS WITH SMARTER TURN-TAKING AND WEBRTC SUPPORT

The final 0.2.x release lands Smart Turn v3.2 endpointing, WebRTC support, and direct audio input for audio-capable LLMs, while diffusers onboards MiniMax-H3 for joint video and audio generation.

python 57 shipped

@huggingface Aug 5

KERNELS REFACTOR CLEANS UP LOADING PATH, TOKENIZERS KILLS DEPENDABOT NOISE

Hugging Face is untangling kernel loading internals across multiple repos while simultaneously eliminating 50% of its Dependabot burden through surgical dependency cleanup.

python 26 shipped

@huggingface Aug 4

LEROBOT BREAKS IMPORTS, FIXES TRAINING PIPELINE SILENTLY BREAKING

LeRobot 0.6.1 renames the types module and patches a critical bug where multi-stage training produced all-zero targets when episode annotations were malformed.

+11

python 69 shipped

@huggingface Aug 3

LEROBOT FIXES D405 REALSENSE STARTUP HANGS WITH AUTO-RECOVERY

LeRobot's camera stack now survives RealSense D405 hardware lockups that used to require a physical replug.

python 5 shipped

Elsewhere on the wire

AI Agents Aug 6

LANGCHAIN HARDENS ANTHROPIC TOOL CALLING AND GIT CREDENTIALS AS OPENHANDS SHIPS 1.10.0

LangChain fixed a critical credential leak in evaluation runs while Anthropic tool schemas get proper fallback handling, as three major agent frameworks push releases across the ecosystem.

ai-agents 54 shipped

Local LLMs Aug 6

OLLAMA FIXES KV CACHE CORRUPTION, LLAMA.CPP HARDENS FILE SEARCH, VLLM KILLS MOE LEGACY CODE

Ollama patched a critical cache corruption bug that corrupts lazy snapshots during multi-request prefills, while llama.cpp hardened Windows junction traversal and vLLM removed 14 hardcoded MoE shards across quantization methods.