Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit History

Commits on Feb 26, 2026

• Fix Flash Attention 3 interface for new FA3 return format (#13173)

  Show description for 47455bd
  veeceey

  authored
  47455bd

  Copy full SHA for 47455bd

• Fix wrong `do_classifier_free_guidance` threshold in ZImagePipeline (#13183)

  Show description for 97c2c6e
  

  kirillsst

  and

  Hezlich2

  authored
  97c2c6e

  Copy full SHA for 97c2c6e

• Cosmos Transfer2.5 Auto-Regressive Inference Pipeline (#13114)

  Show description for 212db7b
  miguelmartin75

  authored
  212db7b

  Copy full SHA for 212db7b

Commits on Feb 25, 2026

• [attention backends] use dedicated wrappers from fa3 for cp. (#13165)

  Show description for 3105848
  sayakpaul

  authored
  3105848

  Copy full SHA for 3105848

Commits on Feb 24, 2026

• fix: graceful fallback when attention backends fail to import (#13060)

  Show description for 1f6ac1c
  

  3 peopleauthored
  1f6ac1c

  Copy full SHA for 1f6ac1c

• migrate to `transformers` v5 (#12976)

  Show description for 5e94d62
  sayakpaul

  authored
  5e94d62

  Copy full SHA for 5e94d62

• Fix AutoModel `typing` Import Error (#13178)

  Show description for 7ab2011
  dg845

  authored
  7ab2011

  Copy full SHA for 7ab2011

Commits on Feb 23, 2026

• Allow Automodel to use `from_config` with custom code. (#13123)

  Show description for 4890e9b
  DN6

  authored
  4890e9b

  Copy full SHA for 4890e9b

• Fix T5GemmaEncoder loading for transformers 5.x composite T5GemmaConfig (#13143)

  DavidBert

  authored
  f1e5914

  Copy full SHA for f1e5914

Commits on Feb 21, 2026

• Support Flux Klein peft (fal) lora format (#13169)

  Show description for a80b192
  asomoza

  authored
  a80b192

  Copy full SHA for a80b192

Commits on Feb 20, 2026

• [gguf][torch.compile time] Convert to plain tensor earlier in dequantize_gguf_tensor (#13166)

  Show description for 01de02e
  anijain2305

  authored
  01de02e

  Copy full SHA for 01de02e

Commits on Feb 19, 2026

• [core] Enable CP for kernels-based attention backends (#12812)

  Show description for 99daaa8
  

  sayakpaul

  and

  DN6

  authored
  99daaa8

  Copy full SHA for 99daaa8

• Fix `ftfy` import for PRX Pipeline (#13154)

  Show description for fe78a7b
  dg845

  authored
  fe78a7b

  Copy full SHA for fe78a7b

• Flux2: Tensor tuples can cause issues for checkpointing (#12777)

  Show description for a577ec3
  

  4 peopleauthored
  a577ec3

  Copy full SHA for a577ec3

Commits on Feb 18, 2026

• docs: improve docstring scheduling_flow_match_lcm.py (#13160)

  Show description for 64734b2
  delmalih

  authored
  64734b2

  Copy full SHA for 64734b2

• [CI] Add ftfy as a test dependency (#13155)

  Show description for f81e653
  DN6

  authored
  f81e653

  Copy full SHA for f81e653

Commits on Feb 17, 2026

• [Bug] Fix QwenImageEditPlus Series on NPU (#13017)

  Show description for bcbbded
  

  5 peopleauthored
  bcbbded

  Copy full SHA for bcbbded

Commits on Feb 16, 2026

• [core] support device type device_maps to work with offloading. (#12811)

  Show description for 35086ac
  sayakpaul

  authored
  35086ac

  Copy full SHA for 35086ac

• [Pipelines] Remove k-diffusion (#13152)

  Show description for 59e7a46
  DN6

  authored
  59e7a46

  Copy full SHA for 59e7a46

Commits on Feb 15, 2026

• [LTX2] Fix wrong lora mixin (#13144)

  Show description for b0dc51d
  

  asomoza

  and

  sayakpaul

  authored
  b0dc51d

  Copy full SHA for b0dc51d

• [Modular] add explicit workflow support (#13028)

  Show description for c919ec0
  

  6 peopleauthored
  c919ec0

  Copy full SHA for c919ec0

Commits on Feb 14, 2026

• [Modular] update doc for `ModularPipeline` (#13100)

  Show description for 3c7506b
  

  5 peopleauthored
  3c7506b

  Copy full SHA for 3c7506b

• fix guider (#13147)

  Show description for 19ab0ec
  yiyixuxu

  authored
  19ab0ec

  Copy full SHA for 19ab0ec

• fix MT5Tokenizer (#13146)

  Show description for 5b00a18
  yiyixuxu

  authored
  5b00a18

  Copy full SHA for 5b00a18

• [Modular] add different pipeine blocks to init (#13145)

  Show description for 6141ae2
  

  yiyixuxu

  and

  [email protected]

  authored
  6141ae2

  Copy full SHA for 6141ae2

• [docs] fix ltx2 i2v docstring. (#13135)

  Show description for 3c1c62e
  sayakpaul

  authored
  3c1c62e

  Copy full SHA for 3c1c62e

Commits on Feb 13, 2026

• feat: implement apply_lora_scale to remove boilerplate. (#12994)

  Show description for 8abcf35
  sayakpaul

  authored
  8abcf35

  Copy full SHA for 8abcf35

• Sunset Python 3.8 & get rid of explicit `typing` exports where possible (#12524)

  Show description for 2843b3d
  

  sayakpaul

  and

  DN6

  authored
  2843b3d

  Copy full SHA for 2843b3d

• fix cosmos transformer typing. (#13134)

  sayakpaul

  authored
  76af013

  Copy full SHA for 76af013

Commits on Feb 12, 2026

• docs: improve docstring scheduling_flow_match_heun_discrete.py (#13130)

  Show description for 5f3ea22
  delmalih

  authored
  5f3ea22

  Copy full SHA for 5f3ea22

• Fix LTX-2 Inference when `num_videos_per_prompt > 1` and CFG is Enabled (#13121)

  Show description for 985d83c
  dg845

  authored
  985d83c

  Copy full SHA for 985d83c

• Cosmos Transfer2.5 inference pipeline: general/{seg, depth, blur, edge} (#13066)

  Show description for a181616
  miguelmartin75

  authored
  a181616

  Copy full SHA for a181616

• docs: improve docstring scheduling_flow_match_euler_discrete.py (#13127)

  Show description for 06a0f98
  delmalih

  authored
  06a0f98

  Copy full SHA for 06a0f98

Commits on Feb 11, 2026

• [Fix]Allow `prompt` and `prior_token_ids` to be provided simultaneously in `GlmImagePipeline` (#13092)

  Show description for d324839
  JaredforReal

  authored
  d324839

  Copy full SHA for d324839

• docs: improve docstring scheduling_edm_dpmsolver_multistep.py (#13122)

  Show description for 64e2adf
  delmalih

  authored
  64e2adf

  Copy full SHA for 64e2adf