Who contributed to Hugging Face on May 28, 2026?

3 developers shipped this update, including atharvajoshi10, sywangyi, and ydshieh.

What were the notable Hugging Face updates?

Adding Cosmos 3 to Diffusers, This PR updates the Stable Diffusion IP-Adapter integration, and Use tmp dataset cache + tmp HF modules cache.

@huggingface

Transformers, Datasets, and the open AI-model layer

github ↗

Pick a date

Topics: Python AI / ML Full archive →

The Wire · Showcase

COSMOS 3 LANDS IN DIFFUSERS, PYTORCH 2.12 BREAKS IP-ADAPTER TESTS

By RepoJournal · Filed 06:04 UTC on May 28, 2026 · About Hugging Face

3 people shipped this

ydshieh @ydshieh 2 cited

atharvajoshi10 @atharvajoshi10 1 cited

sywangyi @sywangyi 1 cited

NVIDIA's unified world foundation model is now integrated into Diffusers, but the PyTorch randomness shift is forcing test rewrites across the stack.

Cosmos 3 [1] is shipping as a single Mixture-of-Transformers pipeline that consolidates world generation, physical reasoning, and action prediction into one forward pass, replacing the three-model approach from earlier releases. The IP-Adapter tests for Stable Diffusion are failing on PyTorch 2.12 [2] because torch.randn no longer produces deterministic outputs for the same seed, forcing engineers to patch pipeline-level noise generation directly and update expected output slices. Meanwhile, the CI infrastructure across transformers-test-ci is stabilizing with cache fixes [3] and branch ref corrections [4] to prevent lock file creation errors and git checkout conflicts on main pushes. The ai-reviewer tooling picked up performance improvements in git operations [5] while documentation cleanup continues across the board [6] [7] [8].

Action items

→ Review Cosmos 3 pipeline integration [ref:7] for production readiness before merging huggingface/diffusers [plan]
→ Test IP-Adapter workflows against PyTorch 2.12 deterministic patches [ref:6] huggingface/diffusers [plan]
→ Merge cache and branch ref fixes [ref:3] [ref:5] to unblock CI pipeline huggingface/transformers-test-ci [immediate]

References

[1] Adding Cosmos 3 to Diffusers ↗ huggingface/diffusers
[2] This PR updates the Stable Diffusion IP-Adapter integration ↗ huggingface/diffusers
[3] Use tmp dataset cache + tmp HF modules cache ↗ huggingface/transformers-test-ci
[4] use main instead of main:main ↗ huggingface/transformers-test-ci
[5] improve git operations performance huggingface/ai-reviewer
[6] useless in the README huggingface/ai-reviewer
[7] pesky BRs huggingface/ai-reviewer
[8] useless link huggingface/ai-reviewer

Quick answers

What shipped in Hugging Face on May 28, 2026?: NVIDIA's unified world foundation model is now integrated into Diffusers, but the PyTorch randomness shift is forcing test rewrites across the stack. In total, 22 commits and 7 pull requests landed.
Who contributed to Hugging Face on May 28, 2026?: 3 developers shipped this update, including atharvajoshi10, sywangyi, and ydshieh.
What were the notable Hugging Face updates?: Adding Cosmos 3 to Diffusers, This PR updates the Stable Diffusion IP-Adapter integration, and Use tmp dataset cache + tmp HF modules cache.

TRANSFORMERS OVERHAULS LINEAR ATTENTION WHILE DEPRECATING LEGACY RESPONSE SCHEMA

The transformers library is retiring its fragile response_schema prototype in favor of streaming-compatible parsing, while simultaneously refactoring every linear attention model to use standardized convolution patterns.

python 70 shipped 2-min read

@huggingface 1 day ago

TRANSFORMERS SHIPS FSDP DISTRIBUTED TRAINING STACK, HUB LIBRARY PLUGS REDOS HOLE

Hugging Face landed distributed training orchestration in transformers while plugging a regex vulnerability that could stall untrusted card parsing for minutes.

+10

python 61 shipped 1-min read

@huggingface 2 days ago

TRANSFORMERS HARDENS AGAINST PYTORCH FRAGMENTATION WHILE TRL SIMPLIFIES DISTILLATION

Transformers plugged a cascading import failure that breaks downstream CI on older PyTorch versions, while TRL rips out dead code to lock DistillationTrainer into prompt-only datasets.

python 70 shipped 2-min read

@huggingface 3 days ago

DATASET VIEWER LOCKS DOWN ARROW, FUNES SHIPS GROUNDED ASK

Hugging Face security teams moved overnight to contain a critical Arrow IPC parsing vulnerability in dataset-viewer while shipping three production hardening releases across Repo2RLEnv, optimum-executorch, and funes.

python 91 shipped 2-min read

Elsewhere on the wire

AI Agents about 9 hours ago

CLAUDE OPUS 5 LANDS ACROSS THE STACK

The newest Anthropic model is now live in langchain, Cline, and llama-index, with native support for extended reasoning and 1M context windows.

ai-agents 28 shipped 1-min read

Local LLMs about 9 hours ago

OLLAMA LANDS LAGUNA SUPPORT AND CRUSHES MEMORY LEAKS WHILE SGLANG HITS V0.5.16 WITH CONFIDENCE-DRIVEN SPECULATIVE DECODING

Ollama shipped three critical performance and reliability fixes for Metal residency and concurrent access patterns, while SGL-Lang released 0.5.16 with a new speculative algorithm hitting 383.7 tok/s on DeepSeek-V4.