What shipped in Hugging Face on August 20, 2026?

HuggingFace's training and inference stacks moved in lockstep overnight. In total, 11 commits and 11 pull requests landed.

Who contributed to Hugging Face on August 20, 2026?

7 developers shipped this update, including kashif, Quentin Gallouédec, Albert Villanova del Moral, atharvajoshi10, JingyaHuang, Dhruv Nair, and sayakpaul.

What were the notable Hugging Face updates?

Add AsyncDistillationTrainer, Simplify AsyncGRPOTrainer: drop redundant metric guards, merge train-begin callbacks (#6378), and [core] Support tensor parallelism for model inference (CUDA, Neuron).

@huggingface

Transformers, Datasets, and the open AI-model layer

github ↗

Pick a date

Topics: Python AI / ML Full archive →

The Wire · Showcase

TRL DROPS ASYNC DISTILLATION, DIFFUSERS GO TENSOR PARALLEL

By RepoJournal · Filed 06:59 UTC on August 20, 2026 · About Hugging Face

7 people shipped this

kashif @kashif 2 cited

JingyaHuang @JingyaHuang 2 cited

Quentin Gallouédec @qgallouedec 1 cited

Albert Villanova del Moral @albertvillanova 1 cited

sayakpaul @sayakpaul 1 cited

atharvajoshi10 @atharvajoshi10

Dhruv Nair @DN6

HuggingFace's training and inference stacks moved in lockstep overnight.

TRL's new AsyncDistillationTrainer brings on-policy distillation online, with an architecture that mirrors AsyncGRPOTrainer: a background worker generates student completions and scores them against a teacher served over HTTP, so generation and training overlap instead of alternating [1]. The trainer also supports multi-teacher on-policy distillation, routing each sample to a specific teacher via a `teacher_id` column [1]. Meanwhile, AsyncGRPOTrainer got a simplification pass that drops redundant metric guards and merges train-begin callbacks, tightening the core loop [2]. Diffusers stepped up for production inference: tensor parallelism (TP) is now supported for model inference on CUDA and AWS Neuron, exposed through `model.enable_parallelism(config=TensorParallelConfig(...))`, the same API used for context parallelism [3]. TP shards from a flat `_tp_plan` and is validated on three pipelines on trn2 with TP=8 in both eager and `torch.compile` mode [4]. For those running kernels, the tiled MLP backward now accumulates each weight gradient once per pass instead of once per shard, fixing DDP's `Expected to mark a variable ready only once` failure [5]. Kernels also added an autotuning example, closing issue #733 [6]. Liger-kernel gets a bump to 0.8.2, required in the TRL dependency, and the SAPO warning filter is dropped [7]. Over the period, 11 commits and 11 PRs moved across the four repos.

Action items

→ Review AsyncDistillationTrainer for your distillation pipeline huggingface/trl [plan]
→ Test tensor parallelism on your Neuron or CUDA inference setup huggingface/diffusers [plan]
→ Upgrade liger-kernel to 0.8.2 to match TRL requirement huggingface/trl [immediate]

References

[1] Add AsyncDistillationTrainer ↗ huggingface/trl
[2] Simplify AsyncGRPOTrainer: drop redundant metric guards, merge train-begin callbacks (#6378) huggingface/trl
[3] [core] Support tensor parallelism for model inference (CUDA, Neuron) ↗ huggingface/diffusers
[4] [core] Support tensor parallelism for model inference (CUDA, Neuron) (#13718) huggingface/diffusers
[5] Accumulate tiled MLP weight grads once per backward ↗ huggingface/kernels-community
[6] feat: add an example on using autotuning. ↗ huggingface/kernels
[7] Require liger-kernel 0.8.2 and drop the SAPO warning filter (#6768) huggingface/trl

Quick answers

What shipped in Hugging Face on August 20, 2026?: HuggingFace's training and inference stacks moved in lockstep overnight. In total, 11 commits and 11 pull requests landed.
Who contributed to Hugging Face on August 20, 2026?: 7 developers shipped this update, including kashif, Quentin Gallouédec, Albert Villanova del Moral, atharvajoshi10, JingyaHuang, Dhruv Nair, and sayakpaul.
What were the notable Hugging Face updates?: Add AsyncDistillationTrainer, Simplify AsyncGRPOTrainer: drop redundant metric guards, merge train-begin callbacks (#6378), and [core] Support tensor parallelism for model inference (CUDA, Neuron).

Transformers Brings Biology to v5 Main

Hugging Face's flagship repo hauled two protein-folding models onto the v5 mainline, and a wav2vec2 attention fix is ready to unblock your SDPA workflows.

python 33 shipped

@huggingface Aug 19

TWO NEW TOOLCHAINS AND A PATCH THAT MATTERS

Hugging Face's kernels repo brings Intel's next-gen GPU into the fold, while Transformers quietly fixes a cross-attention bug that could have been eating your encoder context.

python 84 shipped

@huggingface Aug 18

Harbor-HF Walls Off Sandboxes While Serge Patrols Its Gates

Overnight, Hugging Face's internal infrastructure hardened its edge: Harbor-HF locks down sandbox capabilities, Serge's verify workflow stops dispatching with empty inputs, and dataset-viewer workers finally log their own startups.

python 76 shipped

@huggingface Aug 17

HARBOR-HF SHIPS UNIFIED TYPESCRIPT CONTROL SERVICE, TRANSFORMERS PATCHES DEVICE MISMATCHES

Harbor-HF migrated from distributed Python services to a single hosted TypeScript control plane, while Transformers fixed critical device placement bugs and unlocked tokenizers 0.23.1 compatibility.

python 22 shipped

Elsewhere on the wire

AI Agents Aug 21

CLINE V4.1.11 SHIPS IMAGE GENERATION AND AGENT-UI FIXES

Cline's latest release lets models generate images mid-task, while LlamaIndex quietly breaks frontend tool handlers.

ai-agents 78 shipped

Local LLMs Aug 21

DeepSeek V4 fuses shared experts, as vLLM ships a security guard

The biggest local-LLM repos are moving fast: DeepSeek V4's MoE kernel gets a massive fusion, while vLLM patches a security hole.

+12

llms 274 shipped

@CachyOS Aug 21

CachyOS Scheduler Hotfixes and Kernel Upgrades Lead the Day

A hotfix for the scx_cake scheduler and the latest Linux 7.2 kernel are rolling out across CachyOS repos.

infra 66 shipped

Elixir & Phoenix Aug 21

Phoenix 1.8.12 patches channel message drop

Phoenix shipped a patch release that fixes a channel bug silently dropping messages without a join_ref, and LiveView is quietly rewriting the rules of navigation.