RepoJournal
Hugging Face @huggingface

Hugging Face — full archive

139 daily briefings · 9 reviews.

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

  1. week Diffusers 0.41.0 ships Qwen-Image 2.1, TRL 1.15 goes breaking
  2. week Transformers 5.18 ships Nemotron 3 diarization
  3. week TRL 1.14 removes trl.losses, drops GSPO-token
  4. week OpenEnv 0.5.0 ships as typed clients get fixed
  5. week TRL 1.13.0 adds training beyond 1M tokens
  6. month TRL ships 1.13 and 1.14, removing trl.losses
  7. week Transformers deprecations cleaned, NeoMME and offload land
  8. week Transformers v5.16.0 ships Qwen4-Exp, MoE loss cut 99.7%, TRL unblocks MoE LoRA
  9. month Cache-billing leak fixed, and everything from robots to diffusers gets harder to break
  1. MoE expert sharding gets a test, lerobot swaps vendor SDKs for rustypot
  2. Transformers 5.20 ships Muse Spark Open, TRL adds AWS Trainium
  3. TRL defaults padding to "auto" on Neuron, drops vLLM 0.21.0
  4. AsyncGRPO packs rollouts as prefix trees, TRL types tool arguments by schema
  5. Transformers drops legacy data helpers, TRL rebuilds GOLD tests on real trainers
  6. serge PR triage tightens, TPU and multi-GPU test fixes land
  7. Repo2RLEnv parsers learn cargo-nextest and Django, Tasksmith ships to PyPI
  8. OpenRouter joins Inference Providers in huggingface.js
  9. TRL 1.15 lands a fused LM head, transformers finishes its RoPE cleanup
  10. TRL moves every trainer onto the fused LM head
  11. Multi-eos completions in TRL, Kandinsky 6 lands in diffusers
  12. Diffusers 0.41.0 ships Qwen-Image 2.1, TRL drops Python 3.10
  13. TRL widens template matching for LFM2, LFM2.5, and Gemma 4
  14. AnyLanguageModel catches up to Foundation Models 26 and 27
  15. QLoRA keeps DoRA in float32, flash-attention kwargs stop leaking into vision encoders
  16. transformers clears out deprecated mask, cache, and rotary paths
  17. Transformers 5.18 ships Nemotron 3 diarization, TRL drops vLLM 0.20.0
  18. hf-mount duration defaults silently ignored, funes update now bumps integrations
  19. Repo2RLEnv learns mocha's output format, transformers cleans up MPS and Mamba leftovers
  20. TRL seeds before adapter init, docs links repaired
  21. timm fixes small-dataset sampling, TRL patches vLLM weight-sync crash
  22. Funes batches scanner runs, HfFileSystem downloads Xet files directly
  23. OpenEnv 0.6.0 drops implicit port 8000, TRL 1.14 removes trl.losses
  24. TRL drops an experimental trainer as DPO moves to chunked log-probs
  25. Fused logprob kernels in TRL, rustls pinned past RUSTSEC-2026-0285
  26. lerobot fp16 training lands on every topology
  27. Presets reach reviewed endpoints, Hub revisions resolved once per load
  28. Flash and sage varlen prep work under torch.compile dynamic shapes
  29. OpenEnv closes soupsieve CVEs in two more envs, TRL reshapes GRPO log-probs
  30. OpenEnv 0.5.0 ships new public surface, capture proxy gains an admin key
  31. pwc-cli exposes every read-only command as an MCP tool
  32. SageMaker S3 ownership guard, kernels breaks setuptools builds
  33. OpenEnv typed clients silently dropped every reward
  34. Tau fixes resume loading and Herdr mouse input in 0.4.4
  35. FSDP2 plus expert parallelism, and a unified device mesh for PP + TP inference
  36. HyperClovaX Vision lands in transformers, XPU test baselines reworked
  37. ghlore locks down wide binds, kernels restores env config
  38. Tau 0.4.2 trims session files, chat-ui gates deep links, TRL ships v1.13.0
  39. Fix expert-parallel NaN gradients and honor shift_labels
  40. HF Mount hardens against Hub failures, funes detects secrets per block
  41. Transformers.js adds structured output and docs overhaul
  42. Transformers adds Kimi linear support
  43. TRL removes PPOTrainer, vendors Liger-Kernel losses
  44. TRL documents 1M-token training, diffusers fixes group offloading, kernels fix Intel heuristics
  45. Jinja engine aligns with Jinja2; hub-kernel fallbacks now warn once
  46. kernels gains minver checks, diffusers deprecates ONNX
  47. Transformers drops deprecated code, adds NeoMME and gradient checkpointing offload
  48. Transformers tightens integration tests, TRL pins hybrid attention coverage
  49. moon-ide adds write deadlines and container support
  50. Diffusers pins revisions once, TRL fixes vLLM weight-sync hangs
  51. SAGE ATTENTION GETS ITS SECOND WIND
  52. TRANSFORMERS V5.16.0 SHIPS WITH QWEN4-EXP
  53. TRL UNBLOCKS MOE LORA AS PEFT PATCHES AUTOCAST
  54. Transformers MoE loss cut 99.7% as TRL gains tool calling
  55. Transformers linter hunts nn.Buffer misuse as kernels ships hotfix
  56. Harbor-hf sinks broken Sandbox cleanup
  57. LEROBOT CUTS FASTWAM OVERHEAD WITH COMPILED INFERENCE
  58. Transformers Brings Biology to v5 Main
  59. TRL DROPS ASYNC DISTILLATION, DIFFUSERS GO TENSOR PARALLEL
  60. TWO NEW TOOLCHAINS AND A PATCH THAT MATTERS

Never miss a Hugging Face briefing.

Get Hugging Face's week in review every Monday - and security advisories the day they land. Free, unsubscribe in one click.

One click. Every Monday: what mattered across the projects you've been reading (or our AI picks) — security news same-day. You refine anytime. Choose specific journals · Read a sample issue →