All pull requests

Pull requests

• [ESM/Persimmon/NllbMoe] use safetensors repos to fix Xet FUSE SIGBUS on bucket CI runners

  #49194·ydshieh opened Sep 29, 2026Collaborator·

  ·
  9

• Fix AutoTokenizer backend selection for CohereLabs tiny-aya checkpoints

  #49192·Subodh-17 opened Sep 29, 2026·

  ·
  1

  3

• [Parakeet] Convert NeMo's stochastic depth to layerdrop

  #49191·Deep-unlearning opened Sep 29, 2026Contributor·

  ·
  3

• Clip grad norm dtensor

  #49189·michaelbenayoun opened Sep 29, 2026Member·

  ·
  2

• Fix BLT entropy patcher missing its 512-token sliding window

  #49188·hyunW3 opened Sep 29, 2026·

  ·
  1

  2

• Add a linear attention interface, starting with Mamba2's SSD core

  #49186·hmellor opened Sep 29, 2026Member·

  ·
  3

• [MPS] Let mps sdpa handle grouped query attention directly

  #49187·Isalia20 opened Sep 29, 2026Contributor·

  ·
  3

• QA: fix noisy comment checker

  #49184·tarekziade opened Sep 29, 2026Collaborator·

  ·
  6

• Fix missing router_logits in Qwen3.5-MoE and other MoE models

  #49179·zucchini-nlp opened Sep 29, 2026Member·

  ·
  1

  10

• Fix deepstack features for mixed-input

  #49177·zucchini-nlp opened Sep 29, 2026Member·

  ·
  4

• docs: percent-encode parentheses in DETR notebook links

  #49174·ump45nose opened Sep 29, 2026·

  ·
  3

• Qualcomm AI Engine Direct - Support QNN quantization and lowering flow in the ExecuTorch exporter

  #49173·shewu-quic opened Sep 29, 2026·

  ·
  1

• [MiMo-V2-Flash] Count the query token in the sliding window

  #49170·guapisolo opened Sep 28, 2026·

  ·
  4

• [docs] Add translation workflow

  #49168·stevhliu opened Sep 28, 2026Member·

  ·
  2

• [Nemotron3Diarization] fix streaming last stft frame dropped

  #49167·eustlb opened Sep 28, 2026Collaborator·

  ·
  1

  4

• Fix MiniMax M3 partial 3D vision rotary embeddings

  #49164·cuichenx opened Sep 28, 2026·

  ·
  6

• [distributed] remaining ep_plan to update

  #49162·3outeille opened Sep 28, 2026Member·7/7·

  ·
  3

• [distributed] Default MoE ep_plans to token dispatch

  #49160·3outeille opened Sep 28, 2026Member·6/7·

  ·
  12

• [CB] Make CB more device agnostic and add support for XPU

  #49156·remi-or opened Sep 28, 2026Collaborator·

  ·
  3

• Serialize MPS transfers in async model loading (fix #48029)

  #49154·Irfanwani opened Sep 28, 2026·

  ·
  1

  2

• fix bug for DebertaV2ForMaskedLM non-legacy load issue

  #49152·kaixuanliu opened Sep 28, 2026Contributor·

  ·
  5

• feat(ci): scan the docker images with Wiz before they are pushed

  #49150·rtrompier opened Sep 28, 2026Contributor·

  ·
  3

• Fix backend attribute in documentation and default backend in MistralCommonBackend

  #49149·samyuktahegde opened Sep 28, 2026Contributor·

  ·
  5

• This PR fixes Gemma3n fp16 loading when missing weights need to be in…

  #49147·sywangyi opened Sep 28, 2026Contributor·

  ·
  4

• Map bare list, tuple and dict annotations to the right JSON schema type in get_json_schema

  #49145·825pranav opened Sep 27, 2026·

  ·
  1

  6