evalstate/diffusers-dashboard
1
1[2 {3 "additions": 86,4 "author": "yiyixuxu",5 "author_association": "MEMBER",6 "body_excerpt": null,7 "changed_files": 4,8 "cluster_id": null,9 "cluster_ids": [],10 "cluster_role": null,11 "comments_count": 0,12 "conversation_url": "https://github.com/huggingface/diffusers/pull/13567",13 "created_at": "2026-04-27T06:09:54Z",14 "deletions": 22,15 "draft": false,16 "files_url": "https://github.com/huggingface/diffusers/pull/13567/files",17 "html_url": "https://github.com/huggingface/diffusers/pull/13567",18 "labels": [19 "size/M"20 ],21 "merged": false,22 "number": 13567,23 "review_comments_count": 4,24 "state": "open",25 "title": "[agents docs] add pipelines.md etc",26 "updated_at": "2026-04-27T06:42:46Z"27 },28 {29 "additions": 27,30 "author": "ViktoriiaRomanova",31 "author_association": "FIRST_TIME_CONTRIBUTOR",32 "body_excerpt": "Fixes performance issues identified by profiling LTX2 with torch.profiler as part of #13401. Optimises LTX2 by removing unnecessary GPU synchronisation points and replacing CPU tensor creation with on-device tensor operations across the de\u2026",33 "changed_files": 4,34 "cluster_id": null,35 "cluster_ids": [],36 "cluster_role": null,37 "comments_count": 0,38 "conversation_url": "https://github.com/huggingface/diffusers/pull/13564",39 "created_at": "2026-04-26T21:37:51Z",40 "deletions": 16,41 "draft": false,42 "files_url": "https://github.com/huggingface/diffusers/pull/13564/files",43 "html_url": "https://github.com/huggingface/diffusers/pull/13564",44 "labels": [45 "models",46 "size/S",47 "pipelines",48 "schedulers"49 ],50 "merged": false,51 "number": 13564,52 "review_comments_count": 0,53 "state": "open",54 "title": "Eliminate GPU sync overhead and CPU\u2192GPU transfers across LTX2 pipeline",55 "updated_at": "2026-04-26T21:38:06Z"56 },57 {58 "additions": 688,59 "author": "MukundaKatta",60 "author_association": "FIRST_TIME_CONTRIBUTOR",61 "body_excerpt": "## Summary Refs #7219. Scaffold and design doc for `SUPIRPipeline`. The full SUPIR pipeline (degradation-robust SDXL VAE encoder, trimmed ControlNet with ZeroSFT, restoration-guided EDM sampler, optional LLaVA captioning, custom modules, w\u2026",62 "changed_files": 8,63 "cluster_id": null,64 "cluster_ids": [],65 "cluster_role": null,66 "comments_count": 0,67 "conversation_url": "https://github.com/huggingface/diffusers/pull/13563",68 "created_at": "2026-04-25T18:34:23Z",69 "deletions": 0,70 "draft": true,71 "files_url": "https://github.com/huggingface/diffusers/pull/13563/files",72 "html_url": "https://github.com/huggingface/diffusers/pull/13563",73 "labels": [74 "tests",75 "size/L",76 "utils",77 "pipelines"78 ],79 "merged": false,80 "number": 13563,81 "review_comments_count": 0,82 "state": "open",83 "title": "SUPIR: scaffold pipeline + design doc (refs #7219)",84 "updated_at": "2026-04-25T18:34:37Z"85 },86 {87 "additions": 31,88 "author": "sayakpaul",89 "author_association": "MEMBER",90 "body_excerpt": "# What does this PR do? Fixes https://github.com/huggingface/diffusers/actions/runs/24914743793/job/72964477860?pr=13329",91 "changed_files": 6,92 "cluster_id": null,93 "cluster_ids": [],94 "cluster_role": null,95 "comments_count": 0,96 "conversation_url": "https://github.com/huggingface/diffusers/pull/13562",97 "created_at": "2026-04-25T13:36:09Z",98 "deletions": 8,99 "draft": false,100 "files_url": "https://github.com/huggingface/diffusers/pull/13562/files",101 "html_url": "https://github.com/huggingface/diffusers/pull/13562",102 "labels": [103 "size/S",104 "CI"105 ],106 "merged": false,107 "number": 13562,108 "review_comments_count": 0,109 "state": "open",110 "title": "[ci] use tokenizers stable installtion in CI.",111 "updated_at": "2026-04-25T13:36:25Z"112 },113 {114 "additions": 0,115 "author": "cartesian-plane",116 "author_association": "FIRST_TIME_CONTRIBUTOR",117 "body_excerpt": "# What does this PR do? Fixes the `disable_free_noise` function in `AnimateDiffFreeNoiseMixin` The `self.unet.mid_block` was being unconditionally included as a result of the new `if/else` branch introduced in #9288. This means that `self.\u2026",118 "changed_files": 1,119 "cluster_id": null,120 "cluster_ids": [],121 "cluster_role": null,122 "comments_count": 0,123 "conversation_url": "https://github.com/huggingface/diffusers/pull/13561",124 "created_at": "2026-04-25T10:57:58Z",125 "deletions": 1,126 "draft": false,127 "files_url": "https://github.com/huggingface/diffusers/pull/13561/files",128 "html_url": "https://github.com/huggingface/diffusers/pull/13561",129 "labels": [130 "size/S",131 "pipelines"132 ],133 "merged": false,134 "number": 13561,135 "review_comments_count": 0,136 "state": "open",137 "title": "Fix the `disable_free_noise` function in `AnimateDiffFreeNoiseMixin`",138 "updated_at": "2026-04-25T10:58:12Z"139 },140 {141 "additions": 1,142 "author": "cartesian-plane",143 "author_association": "FIRST_TIME_CONTRIBUTOR",144 "body_excerpt": "# What does this PR do? Previously it was set up to show `__setattr__`, likely due to a copy-paste error. This is a minor fix as it just makes the error message correct. ## Before submitting - [x] This PR fixes a typo or improves the docs\u2026",145 "changed_files": 1,146 "cluster_id": null,147 "cluster_ids": [],148 "cluster_role": null,149 "comments_count": 0,150 "conversation_url": "https://github.com/huggingface/diffusers/pull/13560",151 "created_at": "2026-04-25T09:38:56Z",152 "deletions": 1,153 "draft": false,154 "files_url": "https://github.com/huggingface/diffusers/pull/13560/files",155 "html_url": "https://github.com/huggingface/diffusers/pull/13560",156 "labels": [157 "size/S"158 ],159 "merged": false,160 "number": 13560,161 "review_comments_count": 0,162 "state": "open",163 "title": "Fix Exception message for `set_item` call on FrozenDict",164 "updated_at": "2026-04-25T09:39:09Z"165 },166 {167 "additions": 90,168 "author": "NazarKozak",169 "author_association": "FIRST_TIME_CONTRIBUTOR",170 "body_excerpt": "# What does this PR do? Detect Git LFS pointer stubs in `load_state_dict` *before* attempting to deserialize them, and raise a clear error pointing at the actual fix instead of failing with a cryptic safetensors / pickle deserialization er\u2026",171 "changed_files": 4,172 "cluster_id": null,173 "cluster_ids": [],174 "cluster_role": null,175 "comments_count": 0,176 "conversation_url": "https://github.com/huggingface/diffusers/pull/13559",177 "created_at": "2026-04-25T07:22:55Z",178 "deletions": 1,179 "draft": false,180 "files_url": "https://github.com/huggingface/diffusers/pull/13559/files",181 "html_url": "https://github.com/huggingface/diffusers/pull/13559",182 "labels": [183 "models",184 "tests",185 "size/M",186 "utils"187 ],188 "merged": false,189 "number": 13559,190 "review_comments_count": 0,191 "state": "open",192 "title": "feat(loading): detect Git LFS pointer files and raise a clear error",193 "updated_at": "2026-04-25T07:23:10Z"194 },195 {196 "additions": 51,197 "author": "NazarKozak",198 "author_association": "FIRST_TIME_CONTRIBUTOR",199 "body_excerpt": "# What does this PR do? Adds a \"Production deployment notes\" section to `examples/server/README.md` covering one of the easier traps to fall into when adapting the example for orchestrated production environments (Kubernetes, Cloud Run, Ve\u2026",200 "changed_files": 1,201 "cluster_id": null,202 "cluster_ids": [],203 "cluster_role": null,204 "comments_count": 0,205 "conversation_url": "https://github.com/huggingface/diffusers/pull/13558",206 "created_at": "2026-04-25T06:49:39Z",207 "deletions": 0,208 "draft": false,209 "files_url": "https://github.com/huggingface/diffusers/pull/13558/files",210 "html_url": "https://github.com/huggingface/diffusers/pull/13558",211 "labels": [212 "size/M",213 "examples"214 ],215 "merged": false,216 "number": 13558,217 "review_comments_count": 0,218 "state": "open",219 "title": "docs(examples/server): add health-check trap when switching to background load",220 "updated_at": "2026-04-25T06:49:53Z"221 },222 {223 "additions": 38,224 "author": "NazarKozak",225 "author_association": "FIRST_TIME_CONTRIBUTOR",226 "body_excerpt": "# What does this PR do? The `image=` parameter on `Flux2Pipeline` and `Flux2KleinPipeline` is currently described as a \"starting point\", which mirrors the SD/Flux.1 img2img wording: > `Image`, numpy array or tensor representing an image ba\u2026",227 "changed_files": 3,228 "cluster_id": null,229 "cluster_ids": [],230 "cluster_role": null,231 "comments_count": 0,232 "conversation_url": "https://github.com/huggingface/diffusers/pull/13557",233 "created_at": "2026-04-25T06:35:12Z",234 "deletions": 10,235 "draft": false,236 "files_url": "https://github.com/huggingface/diffusers/pull/13557/files",237 "html_url": "https://github.com/huggingface/diffusers/pull/13557",238 "labels": [239 "documentation",240 "size/S",241 "pipelines"242 ],243 "merged": false,244 "number": 13557,245 "review_comments_count": 0,246 "state": "open",247 "title": "docs(flux2): clarify image= is reference conditioning, not img2img",248 "updated_at": "2026-04-25T06:37:29Z"249 },250 {251 "additions": 34,252 "author": "HaoyuLi-Nova",253 "author_association": "FIRST_TIME_CONTRIBUTOR",254 "body_excerpt": "# What does this PR do? This PR fixes `FlaxLMSDiscreteScheduler.step` so it maps the timestep **value** passed from Flax pipelines (the label stored in `state.timesteps[i]`) to a **step index** via `jnp.where`, consistent with `FlaxEulerDi\u2026",255 "changed_files": 1,256 "cluster_id": null,257 "cluster_ids": [],258 "cluster_role": null,259 "comments_count": 0,260 "conversation_url": "https://github.com/huggingface/diffusers/pull/13556",261 "created_at": "2026-04-24T15:24:20Z",262 "deletions": 11,263 "draft": false,264 "files_url": "https://github.com/huggingface/diffusers/pull/13556/files",265 "html_url": "https://github.com/huggingface/diffusers/pull/13556",266 "labels": [267 "size/S",268 "schedulers"269 ],270 "merged": false,271 "number": 13556,272 "review_comments_count": 0,273 "state": "open",274 "title": "Fix FlaxLMS step timestep\u2192index; complete get_lms_coefficient docs",275 "updated_at": "2026-04-24T16:31:47Z"276 },277 {278 "additions": 1,279 "author": "rtrompier",280 "author_association": "MEMBER",281 "body_excerpt": "Bump the pinned doc-builder SHA so that main documentation builds also sync to the HF bucket (dual-write introduced in doc-builder PR #780, 2026-04-15). The current pin predates that change, so release docs never land in the bucket \u2014 only\u2026",282 "changed_files": 1,283 "cluster_id": null,284 "cluster_ids": [],285 "cluster_role": null,286 "comments_count": 0,287 "conversation_url": "https://github.com/huggingface/diffusers/pull/13555",288 "created_at": "2026-04-24T12:27:53Z",289 "deletions": 1,290 "draft": false,291 "files_url": "https://github.com/huggingface/diffusers/pull/13555/files",292 "html_url": "https://github.com/huggingface/diffusers/pull/13555",293 "labels": [294 "size/S",295 "CI"296 ],297 "merged": true,298 "number": 13555,299 "review_comments_count": 0,300 "state": "closed",301 "title": "chore: bump doc-builder SHA for main doc build workflow",302 "updated_at": "2026-04-24T14:10:48Z"303 },304 {305 "additions": 27,306 "author": "DN6",307 "author_association": "MEMBER",308 "body_excerpt": "# What does this PR do? Change the PR labeller workflow for size, missing-tests and fixes-issue so that it only relabels a PR when the set of labels changes. Currently it relabels regardless of whether or not the same labels are being appl\u2026",309 "changed_files": 1,310 "cluster_id": null,311 "cluster_ids": [],312 "cluster_role": null,313 "comments_count": 0,314 "conversation_url": "https://github.com/huggingface/diffusers/pull/13554",315 "created_at": "2026-04-24T10:55:03Z",316 "deletions": 10,317 "draft": false,318 "files_url": "https://github.com/huggingface/diffusers/pull/13554/files",319 "html_url": "https://github.com/huggingface/diffusers/pull/13554",320 "labels": [321 "size/S",322 "CI"323 ],324 "merged": false,325 "number": 13554,326 "review_comments_count": 0,327 "state": "open",328 "title": "[CI] QOL improvement for PR size labeler",329 "updated_at": "2026-04-24T10:55:20Z"330 },331 {332 "additions": 35,333 "author": "racyjelly",334 "author_association": "FIRST_TIMER",335 "body_excerpt": "# What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflec\u2026",336 "changed_files": 7,337 "cluster_id": null,338 "cluster_ids": [],339 "cluster_role": null,340 "comments_count": 0,341 "conversation_url": "https://github.com/huggingface/diffusers/pull/13553",342 "created_at": "2026-04-23T14:40:30Z",343 "deletions": 11,344 "draft": false,345 "files_url": "https://github.com/huggingface/diffusers/pull/13553/files",346 "html_url": "https://github.com/huggingface/diffusers/pull/13553",347 "labels": [348 "models",349 "size/S",350 "utils",351 "pipelines"352 ],353 "merged": false,354 "number": 13553,355 "review_comments_count": 0,356 "state": "closed",357 "title": "Custom v0.12.1",358 "updated_at": "2026-04-23T14:44:14Z"359 },360 {361 "additions": 885,362 "author": "xin3he",363 "author_association": "FIRST_TIME_CONTRIBUTOR",364 "body_excerpt": "# What does this PR do? This pull request introduces support for the AutoRound quantization algorithm in the Diffusers library. AutoRound is a weight-only quantization method that enables efficient inference by optimizing weight rounding a\u2026",365 "changed_files": 12,366 "cluster_id": null,367 "cluster_ids": [],368 "cluster_role": null,369 "comments_count": 8,370 "conversation_url": "https://github.com/huggingface/diffusers/pull/13552",371 "created_at": "2026-04-23T13:37:55Z",372 "deletions": 1,373 "draft": false,374 "files_url": "https://github.com/huggingface/diffusers/pull/13552/files",375 "html_url": "https://github.com/huggingface/diffusers/pull/13552",376 "labels": [377 "documentation",378 "quantization",379 "tests",380 "size/L",381 "utils"382 ],383 "merged": false,384 "number": 13552,385 "review_comments_count": 37,386 "state": "open",387 "title": "Integrate AutoRound into Diffusers",388 "updated_at": "2026-04-27T03:26:05Z"389 },390 {391 "additions": 3466,392 "author": "waitingcheung",393 "author_association": "FIRST_TIME_CONTRIBUTOR",394 "body_excerpt": "# What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflec\u2026",395 "changed_files": 19,396 "cluster_id": null,397 "cluster_ids": [],398 "cluster_role": null,399 "comments_count": 4,400 "conversation_url": "https://github.com/huggingface/diffusers/pull/13551",401 "created_at": "2026-04-23T05:54:56Z",402 "deletions": 6,403 "draft": false,404 "files_url": "https://github.com/huggingface/diffusers/pull/13551/files",405 "html_url": "https://github.com/huggingface/diffusers/pull/13551",406 "labels": [407 "documentation",408 "models",409 "tests",410 "size/L",411 "utils",412 "pipelines",413 "single-file",414 "guiders"415 ],416 "merged": false,417 "number": 13551,418 "review_comments_count": 0,419 "state": "open",420 "title": "feat: Add Motif-Video model and pipelines",421 "updated_at": "2026-04-27T05:58:30Z"422 },423 {424 "additions": 3,425 "author": "Ricardo-M-L",426 "author_association": "CONTRIBUTOR",427 "body_excerpt": "## What does this PR do? Replaces 3 bare `except:` clauses with `except Exception:`. Bare `except:` catches `SystemExit` and `KeyboardInterrupt`, which prevents graceful shutdown (e.g. `Ctrl+C` during a download in `hub_utils.py` would be\u2026",428 "changed_files": 2,429 "cluster_id": null,430 "cluster_ids": [],431 "cluster_role": null,432 "comments_count": 1,433 "conversation_url": "https://github.com/huggingface/diffusers/pull/13550",434 "created_at": "2026-04-22T16:44:18Z",435 "deletions": 3,436 "draft": false,437 "files_url": "https://github.com/huggingface/diffusers/pull/13550/files",438 "html_url": "https://github.com/huggingface/diffusers/pull/13550",439 "labels": [440 "size/S",441 "utils",442 "experimental"443 ],444 "merged": false,445 "number": 13550,446 "review_comments_count": 0,447 "state": "closed",448 "title": "Replace bare except with except Exception",449 "updated_at": "2026-04-25T06:35:12Z"450 },451 {452 "additions": 1,453 "author": "Ricardo-M-L",454 "author_association": "CONTRIBUTOR",455 "body_excerpt": "## What does this PR do? Fixes a bug in `is_torch_xla_version()` where `is_torch_xla_available` (a function) is referenced without parentheses. This checks the truthiness of the function object itself (always `True`) rather than calling it\u2026",456 "changed_files": 1,457 "cluster_id": null,458 "cluster_ids": [],459 "cluster_role": null,460 "comments_count": 0,461 "conversation_url": "https://github.com/huggingface/diffusers/pull/13549",462 "created_at": "2026-04-22T16:16:39Z",463 "deletions": 1,464 "draft": false,465 "files_url": "https://github.com/huggingface/diffusers/pull/13549/files",466 "html_url": "https://github.com/huggingface/diffusers/pull/13549",467 "labels": [468 "size/S",469 "utils"470 ],471 "merged": false,472 "number": 13549,473 "review_comments_count": 0,474 "state": "open",475 "title": "Fix missing function call in is_torch_xla_version",476 "updated_at": "2026-04-22T16:16:52Z"477 },478 {479 "additions": 30,480 "author": "sayakpaul",481 "author_association": "MEMBER",482 "body_excerpt": null,483 "changed_files": 1,484 "cluster_id": null,485 "cluster_ids": [],486 "cluster_role": null,487 "comments_count": 0,488 "conversation_url": "https://github.com/huggingface/diffusers/pull/13548",489 "created_at": "2026-04-22T15:59:58Z",490 "deletions": 0,491 "draft": false,492 "files_url": "https://github.com/huggingface/diffusers/pull/13548/files",493 "html_url": "https://github.com/huggingface/diffusers/pull/13548",494 "labels": [495 "size/S",496 "CI"497 ],498 "merged": true,499 "number": 13548,500 "review_comments_count": 0,501 "state": "closed",502 "title": "[ci] feat: have pr labeler label for closing issues.",503 "updated_at": "2026-04-22T20:53:23Z"504 },505 {506 "additions": 199,507 "author": "songh11",508 "author_association": "CONTRIBUTOR",509 "body_excerpt": "# What does this PR do? Adds a new \"Ring Anything\" context-parallel attention mode that supports arbitrary (non-evenly divisible) KV sequence lengths across ring-degree workers. ### Motivation Existing `TemplatedRingAttention` requires KV\u2026",510 "changed_files": 4,511 "cluster_id": null,512 "cluster_ids": [],513 "cluster_role": null,514 "comments_count": 9,515 "conversation_url": "https://github.com/huggingface/diffusers/pull/13545",516 "created_at": "2026-04-22T06:25:46Z",517 "deletions": 16,518 "draft": false,519 "files_url": "https://github.com/huggingface/diffusers/pull/13545/files",520 "html_url": "https://github.com/huggingface/diffusers/pull/13545",521 "labels": [522 "documentation",523 "performance",524 "models",525 "size/L",526 "hooks"527 ],528 "merged": false,529 "number": 13545,530 "review_comments_count": 14,531 "state": "open",532 "title": "feat: support ring attention with arbitrary KV sequence lengths",533 "updated_at": "2026-04-26T17:06:09Z"534 },535 {536 "additions": 1,537 "author": "Ricardo-M-L",538 "author_association": "CONTRIBUTOR",539 "body_excerpt": "## What does this PR do? In `examples/research_projects/pytorch_xla/inference/flux/flux_inference_spmd.py`, the `NotImplementedError` exception object is created but never raised when `num_devices < 4`: ```python # Before (bug): else: NotI\u2026",540 "changed_files": 1,541 "cluster_id": null,542 "cluster_ids": [],543 "cluster_role": null,544 "comments_count": 0,545 "conversation_url": "https://github.com/huggingface/diffusers/pull/13544",546 "created_at": "2026-04-22T02:52:55Z",547 "deletions": 1,548 "draft": false,549 "files_url": "https://github.com/huggingface/diffusers/pull/13544/files",550 "html_url": "https://github.com/huggingface/diffusers/pull/13544",551 "labels": [552 "size/S",553 "examples"554 ],555 "merged": false,556 "number": 13544,557 "review_comments_count": 0,558 "state": "open",559 "title": "Fix missing raise before NotImplementedError in Flux SPMD example",560 "updated_at": "2026-04-22T02:53:12Z"561 },562 {563 "additions": 1,564 "author": "Ricardo-M-L",565 "author_association": "CONTRIBUTOR",566 "body_excerpt": "## What does this PR do? `AdaLayerNormZeroSingle.forward()` has a return type annotation declaring a 5-element tuple: ```python def forward(self, x, emb=None) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:\u2026",567 "changed_files": 1,568 "cluster_id": null,569 "cluster_ids": [],570 "cluster_role": null,571 "comments_count": 1,572 "conversation_url": "https://github.com/huggingface/diffusers/pull/13543",573 "created_at": "2026-04-22T02:40:15Z",574 "deletions": 1,575 "draft": false,576 "files_url": "https://github.com/huggingface/diffusers/pull/13543/files",577 "html_url": "https://github.com/huggingface/diffusers/pull/13543",578 "labels": [579 "models",580 "size/S"581 ],582 "merged": false,583 "number": 13543,584 "review_comments_count": 0,585 "state": "closed",586 "title": "Fix AdaLayerNormZeroSingle.forward return type annotation",587 "updated_at": "2026-04-25T06:35:23Z"588 },589 {590 "additions": 1,591 "author": "Ricardo-M-L",592 "author_association": "CONTRIBUTOR",593 "body_excerpt": "## What does this PR do? In `FromSingleFileMixin.from_single_file()`, when an invalid `torch_dtype` is passed, the warning message always displays `torch.float32` instead of the actual invalid value: ```python # Before (bug): if torch_dtyp\u2026",594 "changed_files": 1,595 "cluster_id": null,596 "cluster_ids": [],597 "cluster_role": null,598 "comments_count": 0,599 "conversation_url": "https://github.com/huggingface/diffusers/pull/13542",600 "created_at": "2026-04-22T02:28:43Z",601 "deletions": 1,602 "draft": false,603 "files_url": "https://github.com/huggingface/diffusers/pull/13542/files",604 "html_url": "https://github.com/huggingface/diffusers/pull/13542",605 "labels": [606 "size/S",607 "single-file"608 ],609 "merged": false,610 "number": 13542,611 "review_comments_count": 0,612 "state": "open",613 "title": "Fix torch_dtype warning showing wrong value in from_single_file",614 "updated_at": "2026-04-22T02:28:57Z"615 },616 {617 "additions": 3,618 "author": "jiqing-feng",619 "author_association": "CONTRIBUTOR",620 "body_excerpt": "## What does this PR do? Shrinks dummy inputs in `TestWanAnimateTransformer3DTorchAo` to avoid OOM on devices without FlashAttention (e.g. XPU, which falls back to math SDPA and materializes the full O(S\u00b2) attention matrix). Reduced `hidde\u2026",621 "changed_files": 1,622 "cluster_id": null,623 "cluster_ids": [],624 "cluster_role": null,625 "comments_count": 1,626 "conversation_url": "https://github.com/huggingface/diffusers/pull/13541",627 "created_at": "2026-04-22T01:50:46Z",628 "deletions": 3,629 "draft": false,630 "files_url": "https://github.com/huggingface/diffusers/pull/13541/files",631 "html_url": "https://github.com/huggingface/diffusers/pull/13541",632 "labels": [633 "tests",634 "size/S"635 ],636 "merged": false,637 "number": 13541,638 "review_comments_count": 0,639 "state": "open",640 "title": "Reduce WanAnimate TorchAO test input sizes to prevent OOM",641 "updated_at": "2026-04-27T02:37:48Z"642 },643 {644 "additions": 33,645 "author": "jiqing-feng",646 "author_association": "CONTRIBUTOR",647 "body_excerpt": "## What does this PR do? Introduces `_get_dummy_inputs_for_model(model)` helper in both `QuantizationTesterMixin` and `QuantizationCompileTesterMixin` to automatically cast floating-point input tensors to the model's parameter dtype, preve\u2026",648 "changed_files": 1,649 "cluster_id": null,650 "cluster_ids": [],651 "cluster_role": null,652 "comments_count": 0,653 "conversation_url": "https://github.com/huggingface/diffusers/pull/13540",654 "created_at": "2026-04-22T01:50:18Z",655 "deletions": 10,656 "draft": false,657 "files_url": "https://github.com/huggingface/diffusers/pull/13540/files",658 "html_url": "https://github.com/huggingface/diffusers/pull/13540",659 "labels": [660 "tests",661 "size/S"662 ],663 "merged": false,664 "number": 13540,665 "review_comments_count": 0,666 "state": "open",667 "title": "Add _get_dummy_inputs_for_model helper for quantization tests",668 "updated_at": "2026-04-22T08:19:50Z"669 },670 {671 "additions": 2,672 "author": "jiqing-feng",673 "author_association": "CONTRIBUTOR",674 "body_excerpt": "## What does this PR do? Changes autocast dtype from `float16` to `bfloat16` in `_test_quantization_training`. Float16's limited dynamic range (max ~65504, min subnormal ~5.96e-8) causes gradients to underflow to zero when passing through\u2026",675 "changed_files": 1,676 "cluster_id": null,677 "cluster_ids": [],678 "cluster_role": null,679 "comments_count": 0,680 "conversation_url": "https://github.com/huggingface/diffusers/pull/13539",681 "created_at": "2026-04-22T01:49:25Z",682 "deletions": 1,683 "draft": false,684 "files_url": "https://github.com/huggingface/diffusers/pull/13539/files",685 "html_url": "https://github.com/huggingface/diffusers/pull/13539",686 "labels": [687 "tests",688 "size/S"689 ],690 "merged": false,691 "number": 13539,692 "review_comments_count": 0,693 "state": "open",694 "title": "Fix training gradient underflow in quantization tests",695 "updated_at": "2026-04-22T08:19:50Z"696 },697 {698 "additions": 19,699 "author": "jiqing-feng",700 "author_association": "CONTRIBUTOR",701 "body_excerpt": "## What does this PR do? Implements the `_dequantize()` method for `TorchAoHfQuantizer`, enabling `model.dequantize()` to convert TorchAO-quantized models back to standard float weights. ### Changes 1. **Add `_dequantize()` method**: Itera\u2026",702 "changed_files": 2,703 "cluster_id": null,704 "cluster_ids": [],705 "cluster_role": null,706 "comments_count": 0,707 "conversation_url": "https://github.com/huggingface/diffusers/pull/13538",708 "created_at": "2026-04-22T01:41:35Z",709 "deletions": 0,710 "draft": false,711 "files_url": "https://github.com/huggingface/diffusers/pull/13538/files",712 "html_url": "https://github.com/huggingface/diffusers/pull/13538",713 "labels": [714 "quantization",715 "tests",716 "size/S"717 ],718 "merged": false,719 "number": 13538,720 "review_comments_count": 0,721 "state": "open",722 "title": "Implement _dequantize for TorchAO quantizer",723 "updated_at": "2026-04-22T08:19:51Z"724 },725 {726 "additions": 7,727 "author": "jiqing-feng",728 "author_association": "CONTRIBUTOR",729 "body_excerpt": "## What does this PR do? Enables TorchAO `int4wo` (Int4WeightOnlyConfig) quantization tests on Intel XPU devices. ### Changes 1. **Remove `_int4wo_skip` marker**: The `int4wo` tests were previously restricted to CUDA only. Removed this ski\u2026",730 "changed_files": 1,731 "cluster_id": null,732 "cluster_ids": [],733 "cluster_role": null,734 "comments_count": 0,735 "conversation_url": "https://github.com/huggingface/diffusers/pull/13537",736 "created_at": "2026-04-22T01:40:50Z",737 "deletions": 7,738 "draft": false,739 "files_url": "https://github.com/huggingface/diffusers/pull/13537/files",740 "html_url": "https://github.com/huggingface/diffusers/pull/13537",741 "labels": [742 "tests",743 "size/S"744 ],745 "merged": false,746 "number": 13537,747 "review_comments_count": 0,748 "state": "open",749 "title": "Enable TorchAO int4wo quantization tests on XPU",750 "updated_at": "2026-04-22T08:19:56Z"751 },752 {753 "additions": 117,754 "author": "sayakpaul",755 "author_association": "MEMBER",756 "body_excerpt": "# What does this PR do? Potentially fixes https://github.com/huggingface/diffusers/pull/13472#discussion_r3082680279",757 "changed_files": 1,758 "cluster_id": null,759 "cluster_ids": [],760 "cluster_role": null,761 "comments_count": 0,762 "conversation_url": "https://github.com/huggingface/diffusers/pull/13536",763 "created_at": "2026-04-22T00:37:13Z",764 "deletions": 12,765 "draft": false,766 "files_url": "https://github.com/huggingface/diffusers/pull/13536/files",767 "html_url": "https://github.com/huggingface/diffusers/pull/13536",768 "labels": [769 "size/M",770 "CI"771 ],772 "merged": false,773 "number": 13536,774 "review_comments_count": 2,775 "state": "open",776 "title": "[ci] allow claude to open PRs for certain instructions.",777 "updated_at": "2026-04-23T12:02:24Z"778 },779 {780 "additions": 25,781 "author": "sywangyi",782 "author_association": "CONTRIBUTOR",783 "body_excerpt": "following case fail tests/models/transformers/test_models_transformer_bria.py::BriaTransformerLoRAHotSwapTests::test_hotswapping_compiled_model_linear_2 tests/models/transformers/test_models_transformer_bria.py::BriaTransformerLoRAHotSwapT\u2026",784 "changed_files": 1,785 "cluster_id": null,786 "cluster_ids": [],787 "cluster_role": null,788 "comments_count": 5,789 "conversation_url": "https://github.com/huggingface/diffusers/pull/13535",790 "created_at": "2026-04-21T12:22:47Z",791 "deletions": 6,792 "draft": false,793 "files_url": "https://github.com/huggingface/diffusers/pull/13535/files",794 "html_url": "https://github.com/huggingface/diffusers/pull/13535",795 "labels": [796 "lora",797 "size/S"798 ],799 "merged": false,800 "number": 13535,801 "review_comments_count": 1,802 "state": "closed",803 "title": "fix the test_hotswapping_compiled_model_xxx case failure in xpu",804 "updated_at": "2026-04-22T02:04:16Z"805 },806 {807 "additions": 1,808 "author": "kaixuanliu",809 "author_association": "CONTRIBUTOR",810 "body_excerpt": "This PR tries to fix 2 failed test cases for hiream pipeline tests: ``` tests/pipelines/hidream_image/test_pipeline_hidream.py::HiDreamImagePipelineFastTests::test_cpu_offload_forward_pass_twice tests/pipelines/hidream_image/test_pipeline_\u2026",811 "changed_files": 1,812 "cluster_id": null,813 "cluster_ids": [],814 "cluster_role": null,815 "comments_count": 2,816 "conversation_url": "https://github.com/huggingface/diffusers/pull/13534",817 "created_at": "2026-04-21T07:41:02Z",818 "deletions": 1,819 "draft": false,820 "files_url": "https://github.com/huggingface/diffusers/pull/13534/files",821 "html_url": "https://github.com/huggingface/diffusers/pull/13534",822 "labels": [823 "tests",824 "size/S"825 ],826 "merged": true,827 "number": 13534,828 "review_comments_count": 0,829 "state": "closed",830 "title": "Fix non-deterministic T5 outputs in HiDream pipeline tests",831 "updated_at": "2026-04-21T16:28:41Z"832 },833 {834 "additions": 2,835 "author": "Ricardo-M-L",836 "author_association": "CONTRIBUTOR",837 "body_excerpt": "## What does this PR do? Both `AuraFlowAttnProcessor2_0` and `FusedAuraFlowAttnProcessor2_0` in `attention_processor.py` contain a copy-paste error where `attn.norm_added_q` is called on the encoder **key** projection while guarded by a ch\u2026",838 "changed_files": 1,839 "cluster_id": null,840 "cluster_ids": [],841 "cluster_role": null,842 "comments_count": 1,843 "conversation_url": "https://github.com/huggingface/diffusers/pull/13533",844 "created_at": "2026-04-21T06:18:39Z",845 "deletions": 2,846 "draft": false,847 "files_url": "https://github.com/huggingface/diffusers/pull/13533/files",848 "html_url": "https://github.com/huggingface/diffusers/pull/13533",849 "labels": [850 "models",851 "size/S"852 ],853 "merged": true,854 "number": 13533,855 "review_comments_count": 0,856 "state": "closed",857 "title": "Fix AuraFlow attn processors applying norm_added_q to key projection",858 "updated_at": "2026-04-21T19:20:55Z"859 },860 {861 "additions": 2,862 "author": "Ricardo-M-L",863 "author_association": "CONTRIBUTOR",864 "body_excerpt": "## What this PR does When a user passes pre-computed `prompt_embeds` (or `negative_prompt_embeds`) alongside `num_images_per_prompt > 1`, `ErnieImagePipeline.__call__` did not replicate the provided embeddings \u2014 the embeds list kept its or\u2026",865 "changed_files": 1,866 "cluster_id": null,867 "cluster_ids": [],868 "cluster_role": null,869 "comments_count": 1,870 "conversation_url": "https://github.com/huggingface/diffusers/pull/13532",871 "created_at": "2026-04-21T06:07:21Z",872 "deletions": 2,873 "draft": false,874 "files_url": "https://github.com/huggingface/diffusers/pull/13532/files",875 "html_url": "https://github.com/huggingface/diffusers/pull/13532",876 "labels": [877 "close-to-merge",878 "size/S",879 "pipelines"880 ],881 "merged": false,882 "number": 13532,883 "review_comments_count": 0,884 "state": "open",885 "title": "Fix ErnieImagePipeline pre-computed prompt_embeds + num_images_per_prompt shape mismatch",886 "updated_at": "2026-04-21T20:01:13Z"887 },888 {889 "additions": 0,890 "author": "Ricardo-M-L",891 "author_association": "FIRST_TIME_CONTRIBUTOR",892 "body_excerpt": "## What this PR does `train_sana_sprint_diffusers.py` ends training (under `--push_to_hub`) with: https://github.com/huggingface/diffusers/blob/main/examples/research_projects/sana/train_sana_sprint_diffusers.py#L1758-L1766 ```python if ar\u2026",893 "changed_files": 1,894 "cluster_id": null,895 "cluster_ids": [],896 "cluster_role": null,897 "comments_count": 0,898 "conversation_url": "https://github.com/huggingface/diffusers/pull/13531",899 "created_at": "2026-04-21T05:59:24Z",900 "deletions": 1,901 "draft": false,902 "files_url": "https://github.com/huggingface/diffusers/pull/13531/files",903 "html_url": "https://github.com/huggingface/diffusers/pull/13531",904 "labels": [905 "size/S",906 "examples"907 ],908 "merged": false,909 "number": 13531,910 "review_comments_count": 0,911 "state": "open",912 "title": "Fix AttributeError/TypeError in sana_sprint model-card save",913 "updated_at": "2026-04-21T05:59:35Z"914 },915 {916 "additions": 75,917 "author": "jiqing-feng",918 "author_association": "CONTRIBUTOR",919 "body_excerpt": "## What does this PR do? This PR improves the TorchAO quantization testing infrastructure with several fixes: enabling `int4wo` tests on Intel XPU, implementing `_dequantize` for TorchAO, fixing input dtype mismatches, and fixing training\u2026",920 "changed_files": 3,921 "cluster_id": null,922 "cluster_ids": [],923 "cluster_role": null,924 "comments_count": 5,925 "conversation_url": "https://github.com/huggingface/diffusers/pull/13530",926 "created_at": "2026-04-21T05:59:22Z",927 "deletions": 19,928 "draft": false,929 "files_url": "https://github.com/huggingface/diffusers/pull/13530/files",930 "html_url": "https://github.com/huggingface/diffusers/pull/13530",931 "labels": [932 "quantization",933 "tests",934 "size/M"935 ],936 "merged": false,937 "number": 13530,938 "review_comments_count": 6,939 "state": "open",940 "title": "Improve TorchAO quantization test coverage and XPU support",941 "updated_at": "2026-04-27T01:07:07Z"942 },943 {944 "additions": 4,945 "author": "Ricardo-M-L",946 "author_association": "FIRST_TIME_CONTRIBUTOR",947 "body_excerpt": "## What this PR does Both Wuerstchen prior training scripts build their HF Hub model card with an f-string that references `{args.weight_dtype}`: **`train_text_to_image_prior.py`** (lines 104-105): ```python pipe_prior = DiffusionPipeline.\u2026",948 "changed_files": 2,949 "cluster_id": null,950 "cluster_ids": [],951 "cluster_role": null,952 "comments_count": 0,953 "conversation_url": "https://github.com/huggingface/diffusers/pull/13529",954 "created_at": "2026-04-21T05:57:33Z",955 "deletions": 4,956 "draft": false,957 "files_url": "https://github.com/huggingface/diffusers/pull/13529/files",958 "html_url": "https://github.com/huggingface/diffusers/pull/13529",959 "labels": [960 "size/S",961 "examples"962 ],963 "merged": false,964 "number": 13529,965 "review_comments_count": 0,966 "state": "open",967 "title": "Fix AttributeError in Wuerstchen prior training model card",968 "updated_at": "2026-04-21T05:57:49Z"969 },970 {971 "additions": 9,972 "author": "Ricardo-M-L",973 "author_association": "FIRST_TIME_CONTRIBUTOR",974 "body_excerpt": "## What this PR does The Intel/BF16 `textual_inversion_bf16.py` script reads `args.report_to` at the top of `main()`: https://github.com/huggingface/diffusers/blob/main/examples/research_projects/intel_opts/textual_inversion/textual_invers\u2026",975 "changed_files": 1,976 "cluster_id": null,977 "cluster_ids": [],978 "cluster_role": null,979 "comments_count": 0,980 "conversation_url": "https://github.com/huggingface/diffusers/pull/13528",981 "created_at": "2026-04-21T05:56:11Z",982 "deletions": 0,983 "draft": false,984 "files_url": "https://github.com/huggingface/diffusers/pull/13528/files",985 "html_url": "https://github.com/huggingface/diffusers/pull/13528",986 "labels": [987 "size/S",988 "examples"989 ],990 "merged": false,991 "number": 13528,992 "review_comments_count": 0,993 "state": "open",994 "title": "Fix AttributeError in textual_inversion_bf16 by adding --report_to arg",995 "updated_at": "2026-04-21T05:56:27Z"996 },997 {998 "additions": 4,999 "author": "Ricardo-M-L",1000 "author_association": "CONTRIBUTOR",1001 "body_excerpt": "## What this PR does `ZImageOmniPipeline.__call__` defines `negative_condition_siglip_embeds` only inside the CFG guard: https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/z_image/pipeline_z_image_omni.py#L581-L582\u2026",1002 "changed_files": 1,1003 "cluster_id": null,1004 "cluster_ids": [],1005 "cluster_role": null,1006 "comments_count": 2,1007 "conversation_url": "https://github.com/huggingface/diffusers/pull/13527",1008 "created_at": "2026-04-21T05:42:57Z",1009 "deletions": 3,1010 "draft": false,1011 "files_url": "https://github.com/huggingface/diffusers/pull/13527/files",1012 "html_url": "https://github.com/huggingface/diffusers/pull/13527",1013 "labels": [1014 "size/S",1015 "pipelines"1016 ],1017 "merged": false,1018 "number": 13527,1019 "review_comments_count": 0,1020 "state": "open",1021 "title": "Fix NameError in ZImageOmniPipeline when guidance_scale=0",1022 "updated_at": "2026-04-24T09:16:09Z"1023 },1024 {1025 "additions": 1,1026 "author": "Ricardo-M-L",1027 "author_association": "FIRST_TIME_CONTRIBUTOR",1028 "body_excerpt": "## What this PR does `LongCatImageEditPipeline._encode_prompt` calls `len()` twice on an `int` when building its truncation warning message: https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/longcat_image/pipeline_\u2026",1029 "changed_files": 1,1030 "cluster_id": null,1031 "cluster_ids": [],1032 "cluster_role": null,1033 "comments_count": 0,1034 "conversation_url": "https://github.com/huggingface/diffusers/pull/13526",1035 "created_at": "2026-04-21T05:37:58Z",1036 "deletions": 1,1037 "draft": false,1038 "files_url": "https://github.com/huggingface/diffusers/pull/13526/files",1039 "html_url": "https://github.com/huggingface/diffusers/pull/13526",1040 "labels": [1041 "size/S",1042 "pipelines"1043 ],1044 "merged": false,1045 "number": 13526,1046 "review_comments_count": 0,1047 "state": "open",1048 "title": "Fix TypeError in LongCatImageEditPipeline truncation warning",1049 "updated_at": "2026-04-21T05:38:14Z"1050 },1051 {1052 "additions": 4,1053 "author": "Ricardo-M-L",1054 "author_association": "FIRST_TIME_CONTRIBUTOR",1055 "body_excerpt": "## What this PR does In `LongCatAudioDiTPipeline.__call__`, positive prompts are normalized via `_normalize_text` before being encoded: ```python normalized_prompts = [_normalize_text(text) for text in prompt] ... prompt_embeds, prompt_emb\u2026",1056 "changed_files": 1,1057 "cluster_id": null,1058 "cluster_ids": [],1059 "cluster_role": null,1060 "comments_count": 0,1061 "conversation_url": "https://github.com/huggingface/diffusers/pull/13525",1062 "created_at": "2026-04-21T05:32:36Z",1063 "deletions": 1,1064 "draft": false,1065 "files_url": "https://github.com/huggingface/diffusers/pull/13525/files",1066 "html_url": "https://github.com/huggingface/diffusers/pull/13525",1067 "labels": [1068 "size/S",1069 "pipelines"1070 ],1071 "merged": false,1072 "number": 13525,1073 "review_comments_count": 0,1074 "state": "open",1075 "title": "Normalize negative_prompt in LongCatAudioDiT pipeline for CFG symmetry",1076 "updated_at": "2026-04-21T05:32:50Z"1077 },1078 {1079 "additions": 3,1080 "author": "Ricardo-M-L",1081 "author_association": "FIRST_TIME_CONTRIBUTOR",1082 "body_excerpt": "## What does this PR do? The ORT variant of the unconditional training script, `examples/research_projects/onnxruntime/unconditional_image_generation/train_unconditional.py`, defines the tracker flag as `--logger` on the parser: ```python\u2026",1083 "changed_files": 1,1084 "cluster_id": null,1085 "cluster_ids": [],1086 "cluster_role": null,1087 "comments_count": 0,1088 "conversation_url": "https://github.com/huggingface/diffusers/pull/13524",1089 "created_at": "2026-04-21T04:05:52Z",1090 "deletions": 3,1091 "draft": false,1092 "files_url": "https://github.com/huggingface/diffusers/pull/13524/files",1093 "html_url": "https://github.com/huggingface/diffusers/pull/13524",1094 "labels": [1095 "size/S",1096 "examples"1097 ],1098 "merged": false,1099 "number": 13524,1100 "review_comments_count": 0,1101 "state": "open",1102 "title": "Fix AttributeError in onnxruntime train_unconditional (args.report_to \u2192 args.logger)",1103 "updated_at": "2026-04-21T04:06:06Z"1104 },1105 {1106 "additions": 1,1107 "author": "Ricardo-M-L",1108 "author_association": "FIRST_TIME_CONTRIBUTOR",1109 "body_excerpt": "## What does this PR do? `extract_glyph_texts` in the HunyuanVideo 1.5 modular pipeline had a copy-paste typo in its regex: ```python # src/diffusers/modular_pipelines/hunyuan_video1_5/encoders.py pattern = r\"\\\"(.*?)\\\"|\\\"(.*?)\\\"\" ``` Both\u2026",1110 "changed_files": 1,1111 "cluster_id": null,1112 "cluster_ids": [],1113 "cluster_role": null,1114 "comments_count": 0,1115 "conversation_url": "https://github.com/huggingface/diffusers/pull/13523",1116 "created_at": "2026-04-21T04:02:54Z",1117 "deletions": 1,1118 "draft": false,1119 "files_url": "https://github.com/huggingface/diffusers/pull/13523/files",1120 "html_url": "https://github.com/huggingface/diffusers/pull/13523",1121 "labels": [1122 "size/S",1123 "modular-pipelines"1124 ],1125 "merged": false,1126 "number": 13523,1127 "review_comments_count": 0,1128 "state": "open",1129 "title": "Fix HunyuanVideo 1.5 modular glyph regex for curly quotes",1130 "updated_at": "2026-04-21T04:03:06Z"1131 },1132 {1133 "additions": 1,1134 "author": "Ricardo-M-L",1135 "author_association": "FIRST_TIME_CONTRIBUTOR",1136 "body_excerpt": "## What does this PR do? Fixes an `AttributeError` in `examples/research_projects/diffusion_dpo/train_diffusion_dpo.py` that crashes the IPO loss path. The IPO branch references `args.beta`: ```python elif args.loss_type == \"ipo\": losses =\u2026",1137 "changed_files": 1,1138 "cluster_id": null,1139 "cluster_ids": [],1140 "cluster_role": null,1141 "comments_count": 0,1142 "conversation_url": "https://github.com/huggingface/diffusers/pull/13522",1143 "created_at": "2026-04-21T03:57:51Z",1144 "deletions": 1,1145 "draft": false,1146 "files_url": "https://github.com/huggingface/diffusers/pull/13522/files",1147 "html_url": "https://github.com/huggingface/diffusers/pull/13522",1148 "labels": [1149 "size/S",1150 "examples"1151 ],1152 "merged": false,1153 "number": 13522,1154 "review_comments_count": 0,1155 "state": "open",1156 "title": "Fix AttributeError in diffusion_dpo --loss_type ipo",1157 "updated_at": "2026-04-21T03:58:04Z"1158 },1159 {1160 "additions": 1,1161 "author": "Ricardo-M-L",1162 "author_association": "FIRST_TIME_CONTRIBUTOR",1163 "body_excerpt": "## What does this PR do? Fixes a latent `AttributeError` in `examples/t2i_adapter/train_t2i_adapter_sdxl.py` that breaks `--resume_from_checkpoint`. The accelerate `load_model_hook` references `args.control_type`, which is never declared a\u2026",1164 "changed_files": 1,1165 "cluster_id": null,1166 "cluster_ids": [],1167 "cluster_role": null,1168 "comments_count": 0,1169 "conversation_url": "https://github.com/huggingface/diffusers/pull/13521",1170 "created_at": "2026-04-21T03:55:25Z",1171 "deletions": 3,1172 "draft": false,1173 "files_url": "https://github.com/huggingface/diffusers/pull/13521/files",1174 "html_url": "https://github.com/huggingface/diffusers/pull/13521",1175 "labels": [1176 "size/S",1177 "examples"1178 ],1179 "merged": false,1180 "number": 13521,1181 "review_comments_count": 0,1182 "state": "open",1183 "title": "Fix AttributeError in train_t2i_adapter_sdxl checkpoint resume",1184 "updated_at": "2026-04-21T03:55:37Z"1185 },1186 {1187 "additions": 1,1188 "author": "Ricardo-M-L",1189 "author_association": "FIRST_TIME_CONTRIBUTOR",1190 "body_excerpt": "## What does this PR do? Fixes an `AttributeError` in `examples/dreambooth/train_dreambooth_lora_flux2_img2img.py` that makes the validation flow unusable. The validation block references `args.validation_image_path`: ```python if args.val\u2026",1191 "changed_files": 1,1192 "cluster_id": null,1193 "cluster_ids": [],1194 "cluster_role": null,1195 "comments_count": 0,1196 "conversation_url": "https://github.com/huggingface/diffusers/pull/13520",1197 "created_at": "2026-04-21T03:50:12Z",1198 "deletions": 1,1199 "draft": false,1200 "files_url": "https://github.com/huggingface/diffusers/pull/13520/files",