RepoJournal
Hugging Face @huggingface
139 wires and counting

$ follow Hugging Face

Keep up with Hugging Face in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

or

fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.

Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.

$ status

wire 2026-06-09
stories 44
GitHub RSS Archive Reviews

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

AI REVIEWER GAINS INFERENCE CONTROL, DIFFUSERS TEST SUITE OVERHAUL UNDERWAY

By RepoJournal · Filed · About Hugging Face · Composed from the cited sources · methodology

akshan-main sywangyi zhtmike
3 people shipped this
akshan-main @akshan-main 2 cited sywangyi @sywangyi 1 cited zhtmike @zhtmike 1 cited

The AI reviewer bot now lets you pick your inference provider and manage token budgets with headroom, while diffusers ships a major modeling-test migration across four autoencoder variants.

AI-reviewer shipped two critical UX improvements overnight. The dropdown for Hugging Face inference provider selection [1] fixes a blocking usability gap, and token reduction with headroom support [2] gives you fine-grained control over context windows. Both changes need UI propagation to be useful at scale. Over in diffusers, the modeling-test migration continues at pace: four autoencoder test suites (temporal decoder, cosmos, kvae, mochi) moved to the mixin-based structure [3], fixing a return_dict propagation bug in AutoencoderKLMochi along the way. CosmosControlNetModel tests also migrated [6]. The marigold depth pipeline stopped failing on XPU and A100 [4], and QwenImage attention masking under Ulysses sequence parallelism is fixed [5]. huggingface.js released tasks-v0.21.10 and v0.21.9 [7] [8], added MMMU-Pro as an evaluation framework [9], and updated CheXmix download tracking [10]. Transformers-ci infrastructure gained staging environment mirroring, protocol flexibility, and memory optimizations [11] [12] [13].

Action items

References

  1. [1] implement dropdown when hf inference provider is selected. ↗ huggingface/ai-reviewer
  2. [2] implement token reduction with `headroom` ↗ huggingface/ai-reviewer
  3. [3] refactor autoencoder tests (temporal decoder, cosmos, kvae, mochi) ↗ huggingface/diffusers
  4. [4] fix marigold depth failure in xpu and A100 ↗ huggingface/diffusers
  5. [5] Fix the QwenImage Attention mask under Ulysses SP ↗ huggingface/diffusers
  6. [6] refactor controlnet_cosmos tests ↗ huggingface/diffusers
  7. [7] tasks-v0.21.10 ↗ huggingface/huggingface.js
  8. [8] tasks-v0.21.9 ↗ huggingface/huggingface.js
  9. [9] Add mmmu-pro eval framework (#2213) ↗ huggingface/huggingface.js
  10. [10] Updated CheXmix Download Counts (#2219) ↗ huggingface/huggingface.js
  11. [11] added a mirroring option for a staging env ↗ huggingface/transformers-ci
  12. [12] protect calls around staging ↗ huggingface/transformers-ci
  13. [13] allow setting the protocol ↗ huggingface/transformers-ci

$ ls huggingface/reviews/ # the step-back reads

the week in review Diffusers 0.41.0 ships Qwen-Image 2.1, TRL 1.15 goes breaking the month in review TRL ships 1.13 and 1.14, removing trl.losses
Full archive → Weekly reviews →

Quick answers

What shipped in Hugging Face on June 9, 2026?
The AI reviewer bot now lets you pick your inference provider and manage token budgets with headroom, while diffusers ships a major modeling-test migration across four autoencoder variants. In total, 28 commits, 14 pull requests, and 2 releases landed.
Who contributed to Hugging Face on June 9, 2026?
3 developers shipped this update, including akshan-main, sywangyi, and zhtmike.
What were the notable Hugging Face updates?
implement dropdown when hf inference provider is selected, implement token reduction with `headroom`, and refactor autoencoder tests (temporal decoder, cosmos, kvae, mochi).
♥

$ log @huggingface --recent

MoE expert sharding gets a test, lerobot swaps vendor SDKs for rustypot 59 shipped Transformers 5.20 ships Muse Spark Open, TRL adds AWS Trainium 97 shipped TRL defaults padding to "auto" on Neuron, drops vLLM 0.21.0 164 shipped AsyncGRPO packs rollouts as prefix trees, TRL types tool arguments by schema 80 shipped

$ log --all # elsewhere on the wire

Alignment & Post-training verl rejects conflicting vLLM engine_kwargs, open-instruct adds MILES OPD behind a Megatron learner Interpretability nnsight 0.8.0 cuts import time from 7.6s to 1.6s

$ ls showcases/

Safety Evals AI Red Teaming Agent Security Interpretability Alignment & Post-training

$ init your-desk

RepoJournal can write one of these for any repo you care about. Sign in with GitHub, pick your repos and sources, and your desk gets a wire like this one every morning.