RepoJournal
Hugging Face

@huggingface

Transformers, Datasets, and the open AI-model layer

Pick a date

Topics: Python AI / ML Full archive →

The Wire · Showcase

DIFFUSERS SHIPS COSMOS3 FEW-STEP INFERENCE, LEROBOT STRIPS 3K LINES OF VENDORED CODE

By RepoJournal · Filed · About Hugging Face

Cosmos3 distilled checkpoints hit diffusers with 4-step T2I and I2V support, while lerobot kills its Qwen2.5-VL fork and rebases onto native transformers.

The diffusers pipeline now supports nvidia's Cosmos3-Super distilled variants [1], both 4-step checkpoints shipping with fixed sigma schedules already baked in for immediate inference without tuning. Parallel to that, diffusers is gutting its test layer [2] and adding granular control over pipeline signatures [3], letting test classes declare which canonical parameters a pipeline deliberately drops (like guidance-distilled models skipping negative_prompt) so regressions fail hard. Over in lerobot, the Wall-X policy got a major haircut [7]: the team ripped out 2.8k lines of vendored Qwen2.5-VL code and subclassed the native transformers version instead, killing the hand-sync tax every time transformers updates. That refactor is part of a broader policy cleanup [8] [9] [10] extracting shared model components and standardizing how policies register themselves by convention rather than boilerplate. trl moved fast on housekeeping: KTO trainer got redundant tensor re-wrapping removed [4] [6], and both GRPO and RLOO trainers now document the max_steps requirement for iterable datasets [5], closing the docs gap that already existed in DPO and SFT.

Action items

References

  1. [1] Cosmos3 Distilled support ↗ huggingface/diffusers
  2. [2] [tests] Pipeline test refactor ↗ huggingface/diffusers
  3. [3] Add not_params to ModularPipelineTesterMixin (#14207) huggingface/diffusers
  4. [4] Remove redundant torch.tensor re-wrapping of batch labels in KTO (#6431) huggingface/trl
  5. [5] Document max_steps requirement for iterable datasets in GRPO and RLOO ↗ huggingface/trl
  6. [6] Remove redundant torch.tensor re-wrapping of batch labels in KTO ↗ huggingface/trl
  7. [7] refactor(wall-x): subclass native Transformers Qwen2.5-VL instead of vendoring it ↗ huggingface/lerobot
  8. [8] refactor(policies): extract shared model components ↗ huggingface/lerobot
  9. [9] refactor(processors): share policy pipeline builders ↗ huggingface/lerobot
  10. [10] refactor(policies): resolve policy components by convention ↗ huggingface/lerobot

Quick answers

What shipped in Hugging Face on July 18, 2026?
Cosmos3 distilled checkpoints hit diffusers with 4-step T2I and I2V support, while lerobot kills its Qwen2.5-VL fork and rebases onto native transformers. In total, 20 commits and 20 pull requests landed.
Who contributed to Hugging Face on July 18, 2026?
6 developers shipped this update, including yzhautouskay, sayakpaul, YiYi Xu, Albert Villanova del Moral, Quentin Gallouédec, and imstevenpmwork.
What were the notable Hugging Face updates?
Cosmos3 Distilled support, [tests] Pipeline test refactor, and Add not_params to ModularPipelineTesterMixin (#14207).

More from @huggingface

Daily updates, in your inbox

Follow Hugging Face

Transformers, Datasets, and the open AI-model layer We'll email you a link to confirm first.

Free. Confirm via email. Unsubscribe in one click.

— or follow the whole beat:

Elsewhere on the wire