| COMMIT |
1.00 |
Fix vllm cis (#45139) |
|
Commit message contains explicit AI assi |
2026-04-08 |
| COMMIT |
1.00 |
Update tiny model creation script (#45241) |
|
Commit message contains explicit AI assi |
2026-04-04 |
| COMMIT |
1.00 |
fix: prefer registered config over remote code in AutoConfig |
|
Commit message contains explicit AI assi |
2026-03-31 |
| COMMIT |
1.00 |
Fix stupid test fetcher (#45140) |
|
Commit message contains explicit AI assi |
2026-03-31 |
| COMMIT |
1.00 |
[Bugfix] Remove incorrect torchvision requirement from PIL b |
|
Commit message contains explicit AI assi |
2026-03-30 |
| COMMIT |
1.00 |
fix: protect torch imports in processing files and fix impor |
|
Commit message contains explicit AI assi |
2026-03-27 |
| COMMIT |
1.00 |
Fix release full (#45029) |
|
Commit message contains explicit AI assi |
2026-03-27 |
| COMMIT |
1.00 |
Add `base_model_tp_plan` to `OlmoeConfig` (#44668) |
|
Commit message contains explicit AI assi |
2026-03-26 |
| PR |
1.00 |
resize_token_embeddings does not effect to output_embeddings |
|
PR body explicitly mentions AI collabora |
2026-04-08 |
| PR |
1.00 |
Add Xiaomi MiMo-V2 |
|
PR body explicitly mentions AI collabora |
2026-03-31 |
| PR |
1.00 |
fix: KV cache sharing |
|
PR body explicitly mentions AI collabora |
2026-04-08 |
| PR |
1.00 |
Update min_lr and max_lr default values to better defaults |
|
PR body explicitly mentions AI collabora |
2026-04-01 |
| PR |
1.00 |
fix: liger unnecessarily materializes logits in VRAM during |
|
PR body explicitly mentions AI collabora |
2026-04-06 |
| PR |
1.00 |
Load adapter with TP |
|
PR body explicitly mentions AI collabora |
2026-03-31 |
| PR |
1.00 |
Fix Sam3Processor missing input_boxes_labels for padded None |
|
PR body explicitly mentions AI collabora |
2026-04-01 |
| PR |
1.00 |
Add cuda compatibility check for using `grouped_mm` |
|
PR body explicitly mentions AI collabora |
2026-03-25 |
| PR |
1.00 |
feat: make timesfm2_5 onnx export compatible |
|
PR body explicitly mentions AI collabora |
2026-04-04 |
| PR |
1.00 |
add expert parallelism for gemma-4-26B-A4B-it |
|
PR body explicitly mentions AI collabora |
2026-04-07 |
| PR |
1.00 |
docs maintenance for transformers repository 979e8 |
|
PR body explicitly mentions AI collabora |
2026-04-07 |
| PR |
1.00 |
[Please ignore] CI Test PR |
|
PR body explicitly mentions AI collabora |
2026-04-07 |
| PR |
1.00 |
Add new qwen2 5 vl |
|
PR body explicitly mentions AI collabora |
2026-04-07 |
| PR |
1.00 |
Fix failing `XCLIPModelIntegrationTest` |
|
PR body explicitly mentions AI collabora |
2026-03-27 |
| PR |
1.00 |
First pull request |
|
PR body explicitly mentions AI collabora |
2026-04-07 |
| PR |
1.00 |
Fix UnboundLocalError in invert_attention_mask by adding pro |
|
PR body explicitly mentions AI collabora |
2026-04-05 |
| PR |
1.00 |
add Qianfan-OCR model definition |
|
PR body explicitly mentions AI collabora |
2026-04-07 |
| PR |
1.00 |
Add Qwen3.5 GGUF loading support |
|
PR body explicitly mentions AI collabora |
2026-04-07 |
| COMMIT |
0.80 |
docs: add PermuteForRope to conversion operations reverse ta |
|
Formal explanatory tone with unusually p |
2026-03-26 |
| PR |
0.60 |
🚨 Refactor ViT to updated standards |
|
Starts with 'This PR aims at…', which is |
2025-10-17 |
| PR |
0.35 |
Fix `text-to-speech` pipeline crash when generation config c |
|
Slightly more formal but still clear, co |
2026-03-30 |
| COMMIT |
0.30 |
[CB] Persistent manager (#44435) |
|
Stacked commits with brief human-like co |
2026-03-26 |
| PR |
0.30 |
feat[vLLM × v5]: Add vLLM compatibility for audio models |
|
Contains domain-specific notes and symbo |
2026-04-08 |
| PR |
0.30 |
Fix: ObjectDetectionPipeline batch inference only returns fi |
|
Mix of technical and explanatory; no sus |
2026-04-03 |
| PR |
0.30 |
Added Sapnous Architecture |
|
Slightly formal but includes domain jarg |
2025-03-25 |
| PR |
0.25 |
[Qwen3MoE] Fix wrong return type annotation in Qwen3MoeSpars |
|
Technical and concise; no clear AI-style |
2026-04-03 |
| COMMIT |
0.20 |
Add cohere asr (#45023) |
|
Standard PR structure with commit messag |
2026-03-26 |
| COMMIT |
0.20 |
Ensure final evaluation runs with step-based evaluation stra |
|
Typical rebase/merge conflict pattern, h |
2026-03-26 |
| PR |
0.20 |
[gemma4] Dissociate kv states sharing from the Cache |
|
Technical and succinct, direct to the ch |
2026-04-08 |
| PR |
0.20 |
Fix resize failure caused by zero-sized masks in PP-DocLayou |
|
Clear technical context, but slightly mo |
2026-04-07 |
| PR |
0.20 |
Fix mutable default arguments in quantization config classes |
|
Fairly formal, but technical specifics a |
2026-04-07 |
| PR |
0.20 |
Add docstring to FFN.forward in DistilBERT |
|
Human style with direct explanation and |
2026-04-06 |
| PR |
0.20 |
Add docstrings to AlbertMLMHead and AlbertSOPHead forward me |
|
Technical context and concise, not overl |
2026-04-06 |
| PR |
0.20 |
add HyperClovaX Vision |
|
Slightly more verbose greeting, but huma |
2026-02-27 |
| PR |
0.20 |
Add THD support in ESM |
|
Brief explanation with technical terms, |
2026-02-19 |
| PR |
0.20 |
Module Fusion API |
|
Somewhat structured, but technical and c |
2026-03-24 |
| PR |
0.20 |
Add HyperCLOVAX SEED Think 14B |
|
Technical language, specific model refer |
2026-03-23 |
| PR |
0.20 |
fix(gemma3, gemma4): default token_type_ids to zeros for tex |
|
Some structured summary, but mainly tech |
2026-04-03 |
| PR |
0.15 |
Fix AttributeError in Gemma3ForConditionalGeneration and Gem |
|
Bullet points, technical detail, no clea |
2026-04-07 |
| PR |
0.15 |
[CB] [Major] Add CPU request offloading |
|
Technical vocabulary and informal tone; |
2026-04-02 |
| PR |
0.15 |
Dynamic auto mapping (PoC) |
|
Informal, creative, and technical discus |
2026-03-26 |
| COMMIT |
0.10 |
Fix `SmolVLM` video processor `resize` using wrong interpola |
|
Technical, detailed explanation with dom |
2026-04-06 |
| COMMIT |
0.10 |
empty (#45261) |
|
Casual tone, technical context, some com |
2026-04-06 |
| COMMIT |
0.10 |
post release tag update |
|
Terse human comment about tag update. |
2026-03-27 |
| COMMIT |
0.10 |
style was missing sorry @ydshieh :) (#45038) |
|
Informal apology with GitHub mention, hu |
2026-03-27 |
| COMMIT |
0.10 |
Add BC for `_further_process_kwargs` (#45033) |
|
Standard BC fix with signed-off-by trail |
2026-03-26 |
| COMMIT |
0.10 |
Use multi runners to check new failing tests in a CI run (#4 |
|
Concise technical change with co‑authore |
2026-03-26 |
| COMMIT |
0.10 |
[`fix`] Use the correct _tied_weights_keys for CamembertForC |
|
Direct fix description, human technical |
2026-03-26 |
| COMMIT |
0.10 |
change dev ver. we forgot to do this when we released 5.3.0 |
|
Casual mention of forgotten version bump |
2026-03-26 |
| PR |
0.10 |
TP refactor for FSDP + TP integration |
|
Terse, uses domain-specific abbreviation |
2026-03-26 |
| PR |
0.10 |
[docs] modular transformers |
|
Casual tone with technical abbreviations |
2026-04-08 |
| PR |
0.10 |
Remove references to torchao's AffineQuantizedTensor |
|
Uses clear summary with domain context; |
2026-04-08 |
| PR |
0.10 |
[PoC] HF exporters |
|
Casual edits, emoji use, and PR referenc |
2025-11-03 |
| PR |
0.10 |
Gemma4 resizing per layer inputs |
|
Terse, includes shorthand and references |
2026-04-08 |
| PR |
0.10 |
Logger has `[transformers]` prefix in non-verbose mode |
|
Casual and direct; addresses reviewers, |
2026-04-08 |
| PR |
0.10 |
fix: dont download artifacts from the test hub |
|
Direct fix, minimal language, friendly r |
2026-04-08 |
| PR |
0.10 |
[CB] Fix capture of max_seqlen |
|
Uses technical language and references, |
2026-04-08 |
| PR |
0.10 |
[MOE] MoE routing capture and replay support |
|
Technical MoE jargon and list structure |
2026-03-22 |
| PR |
0.10 |
typing: rule 15 - checks for tie_word_embeddings presence |
|
Brief, technical rule explanation; lacks |
2026-03-25 |
| PR |
0.10 |
[WIP] Fix FA kernel launch needs correct cuda device ctx in |
|
Direct, technical, somewhat fragmented s |
2026-03-24 |
| PR |
0.10 |
Fix softmaxing router logits |
|
Technical explanation, domain detail, no |
2026-04-08 |
| PR |
0.10 |
Add Molmo2 |
|
Template leakage; minimal free-text, no |
2026-01-23 |
| PR |
0.10 |
fix: leak in tokenizer registry for `test_processors` |
|
Clear technical explanation; informal, n |
2026-04-08 |
| PR |
0.10 |
[AMD CI] Fix torch.compile/export failures on AMD CI due to |
|
Technical explanation, use of code snipp |
2026-04-07 |
| PR |
0.10 |
Fix AttributeError in _patch_mistral_regex when fix_mistral_ |
|
Terse technical bug explanation, free of |
2026-04-08 |
| PR |
0.10 |
Conversion for LLM class loading with VLM ckpt |
|
Brief, ticket references, informal phras |
2026-04-08 |
| PR |
0.10 |
feat/rfc/poc: Agnostic GPU |
|
Direct technical language, exhibits clea |
2026-04-04 |
| PR |
0.10 |
Refactor CLIP-like models |
|
Informal tone, use of 'ppl', and project |
2026-03-04 |
| PR |
0.10 |
Fix: NotebookProgressCallback crash when evaluating with the |
|
Technical and concise bugfix note with i |
2026-03-23 |
| PR |
0.10 |
Add Qwen3.5 support for sequence classification |
|
Uses domain-specific terms and concise e |
2026-03-03 |
| PR |
0.10 |
feat: add Gemma4ForSequenceClassification |
|
Concise domain-specific change descripti |
2026-04-07 |
| PR |
0.10 |
Add Deepseek-OCR-2 model |
|
Factual project/model intro with domain |
2026-03-27 |
| PR |
0.10 |
Fix "AttributeError: NewTokenizer has no attribute special_a |
|
Detailed technical bug fix explanation, |
2026-04-07 |
| PR |
0.10 |
Add SAM3-LiteText |
|
Includes domain terms, uses brief human |
2026-02-27 |
| PR |
0.10 |
Optimize Parakeet feature extraction on CUDA |
|
Detailed technical content, with specifi |
2026-03-31 |
| PR |
0.10 |
Use torchvision `decode_image` to load images in the torchv |
|
Technical wording, concise, no AI hallma |
2026-04-02 |
| PR |
0.10 |
Fix missing image processors backends |
|
Brief, domain-specific discussion, lacks |
2026-04-01 |
| PR |
0.10 |
Less unnecessary RoPE warnings |
|
Direct, issue-linked explanation, brief |
2026-04-07 |
| PR |
0.10 |
Fix unexpected TF32 being enabled in testing |
|
Technical, reference to code and issue, |
2026-04-05 |
| PR |
0.10 |
Fix `Wav2Vec2Config.vocab_size` type to allow `None` |
|
Domain-specific, clear technical explana |
2026-03-30 |
| PR |
0.05 |
[docs] pipeline cleanup |
|
Very terse, corrects docs, review also m |
2026-03-23 |
| PR |
0.05 |
Add MoE to Gemma4 TP plan |
|
Uses domain-specific terms and an inform |
2026-04-03 |
| PR |
0.05 |
[Qwen3_5]Remove unnecessary masked_fill_ in torch_chunk_gate |
|
Domain-specific jargon and terse, techni |
2026-04-03 |
| PR |
0.05 |
[docs] vlm addition |
|
Very terse, domain-jargon, clearly human |
2026-04-06 |
| PR |
0.05 |
Refactor core_model_loading to support FSDP shard-on-read lo |
|
Informal, technical, uses TODOs; very hu |
2026-03-24 |
| COMMIT |
0.00 |
Fix `text-to-speech` pipeline crash when generation config c |
|
Direct, domain-specific commit message w |
2026-04-08 |
| COMMIT |
0.00 |
[docs] pipeline cleanup (#44954) |
|
Extremely terse; typical human doc updat |
2026-04-08 |
| COMMIT |
0.00 |
Add MoE to Gemma4 TP plan (#45219) |
|
Concise, domain-language, signed off by |
2026-04-08 |
| COMMIT |
0.00 |
Fix export for gemma4 and add Integration tests (#45285) |
|
Choppy informal commit breakdowns; clear |
2026-04-08 |
| COMMIT |
0.00 |
[docs] static model rules (#45232) |
|
Minimal, iterative commit style; clearly |
2026-04-08 |
| COMMIT |
0.00 |
fix(security): prevent untrusted users from triggering TRL C |
|
Technical detail, informal, and concise |
2026-04-07 |
| COMMIT |
0.00 |
Fix missing image processors backends (#45165) |
|
Very terse commit summary and message, c |
2026-04-07 |
| COMMIT |
0.00 |
[AMD CI] Fix Qwen2 expectations (#45284) |
|
Terse and informal phrasing; human style |
2026-04-07 |
| COMMIT |
0.00 |
Add `hasattr(torch.backends.cudnn, "conv")` to `conftest.py` |
|
Concise commit message; domain-specific |
2026-04-06 |
| COMMIT |
0.00 |
Fix `Qwen2IntegrationTest` (#45268) |
|
Single word summary; strongly human. |
2026-04-06 |
| COMMIT |
0.00 |
doc: fix TokenizersBackend.convert_to_native_format docstrin |
|
Minimal, template-style docstring fix; n |
2026-04-06 |
| COMMIT |
0.00 |
Fix unexpected TF32 being enabled in testing (#45252) |
|
Minimal, terse commit message; human typ |
2026-04-05 |
| COMMIT |
0.00 |
Fix tf32 issue: set `torch.backends.cudnn.conv.fp32_precisio |
|
Commit message is terse, with typical de |
2026-04-05 |
| COMMIT |
0.00 |
Nvidia CI with `torch 2.11` (#45243) |
|
Short, domain-specific, and informal com |
2026-04-04 |
| COMMIT |
0.00 |
Update `get_test_info.py` (related to tiny model creation) ( |
|
Brief, technical, and uses domain abbrev |
2026-04-04 |
| COMMIT |
0.00 |
More fix for tiny model creation (#45228) |
|
Commit message and bullet points are ter |
2026-04-03 |
| COMMIT |
0.00 |
remove unnecessary entries in some auto model mappings (#452 |
|
Extremely terse commit note; clear human |
2026-04-03 |
| COMMIT |
0.00 |
fix: hf-doc-builder insallation was failing (#45225) |
|
Brief and informal phrasing; no AI trait |
2026-04-03 |
| COMMIT |
0.00 |
[CB] Add per-request logits processors (#45026) |
|
Terse, minimal commit message; typical h |
2026-04-03 |
| COMMIT |
0.00 |
[docs] formatting (#45196) |
|
Sparse, technical note with nonstandard |
2026-04-03 |
| COMMIT |
0.00 |
fix `test_register_result_handler` (#45188) |
|
Very brief commit message, lacks AI hall |
2026-04-03 |
| COMMIT |
0.00 |
[CB] Tweaks to update and minor fixes (#45179) |
|
Bullet style, typos, and brevity point t |
2026-04-03 |
| COMMIT |
0.00 |
Fix pypi release (#45210) |
|
Minimal, direct style, some typos, clear |
2026-04-03 |
| COMMIT |
0.00 |
update to dev version 5.6.0-dev0 |
|
Simple version update note, extremely br |
2026-04-03 |
| COMMIT |
0.00 |
fix(docs): correct gemma4 docs and examples (#45197) |
|
Standard commit message with co-author t |
2026-04-02 |
| COMMIT |
0.00 |
Add Turkish (tr) translation for Get Started section (#45158 |
|
Human style, domain-specific, and inform |
2026-04-02 |
| COMMIT |
0.00 |
[docs] transformers serve (#45174) |
|
Very brief and informal commit messages, |
2026-04-02 |
| COMMIT |
0.00 |
casually dropping the most capable open weights on the plane |
|
Casual, informal tone; no AI signatures; |
2026-04-02 |
| COMMIT |
0.00 |
Internalise the NomicBERT model (#43067) |
|
Structured sequential work, domain-speci |
2026-04-02 |
| COMMIT |
0.00 |
Fix resized LM head weights being overwritten by post_init ( |
|
Technically detailed, with typos and dir |
2026-04-02 |
| COMMIT |
0.00 |
[Qwen3.5 MoE] Add _tp_plan to ForConditionalGeneration (#451 |
|
Technical, domain-specific, and succinct |
2026-04-02 |
| COMMIT |
0.00 |
Fix TypeError: 'NoneType' object is not iterable in Generati |
|
Very brief and specific to bug; human co |
2026-04-02 |
| COMMIT |
0.00 |
fix(models): Fix dtype mismatch in SwitchTransformers and Ti |
|
Uses human-like shorthand and changelog |
2026-04-02 |
| COMMIT |
0.00 |
Generalize gemma vision mask to videos (#45185) |
|
Short, informal, with direct responses t |
2026-04-02 |
| COMMIT |
0.00 |
[misc] fix qwen35 tests: correct the text model type and ski |
|
Concise commit message, domain-specific, |
2026-04-02 |
| COMMIT |
0.00 |
🔒 Pin GitHub Actions to commit SHAs (#45180) |
|
Explicit, terse commit titles; no AI hal |
2026-04-02 |
| COMMIT |
0.00 |
Use doc-builder runnable example for GLM-ASR (#44277) |
|
Informal language and typographical erro |
2026-04-02 |
| COMMIT |
0.00 |
CI] Small T5 expectations updated (#45138) |
|
Extremely terse commit message; clearly |
2026-04-02 |
| COMMIT |
0.00 |
fix: correct type annotations across config classes for @str |
|
Contains domain-specific detail, terse s |
2026-04-01 |
| COMMIT |
0.00 |
Fix explicit local code resolution for tokenizers and image |
|
Technical, uses signing trailer but no A |
2026-04-01 |
| COMMIT |
0.00 |
Fix T5Attention shape mismatch under Tensor Parallelism (#45 |
|
Technical explanation, review references |
2026-04-01 |
| COMMIT |
0.00 |
[refactor] Serving into proper modules (#44796) |
|
Commit history is terse, informal, and t |
2026-04-01 |
| COMMIT |
0.00 |
Re-add regex substitutions to the response parsing spec (#45 |
|
Informal changelog, abbreviations, and t |
2026-04-01 |
| COMMIT |
0.00 |
fix bug for janus model image generation (#45044) |
|
Informal, domain-specific, signed by hum |
2026-04-01 |
| COMMIT |
0.00 |
Fix incorrect TrainingArguments example in training.md (#451 |
|
Brief commit messages, domain-specific, |
2026-03-31 |
| COMMIT |
0.00 |
Add parse_response to Processor, make it a bit more official |
|
Direct, minimal phrasing; no AI indicato |
2026-03-31 |
| COMMIT |
0.00 |
DeepGEMM (#44832) |
|
Informal, technical jargon, no sign of A |
2026-03-31 |
| COMMIT |
0.00 |
🚨 [Cache] Native mamba & hybrid cache (#44950) |
|
Casual commit style, lots of terse fixes |
2026-03-31 |
| COMMIT |
0.00 |
[serving] Fix continuous batching JSON response serializatio |
|
Detailed but technical, regression test |
2026-03-31 |
| COMMIT |
0.00 |
refactoring: speedup static checks with disk cache (#44992) |
|
Terse, to-the-point, refactoring context |
2026-03-31 |
| COMMIT |
0.00 |
:rotating_light: [`LightGlue`] Remove remote code execution |
|
Minimal commit messages, no AI hallmarks |
2026-03-31 |
| COMMIT |
0.00 |
[CB] Add warmup feature (#45112) |
|
Brief, iterative commits, domain terms, |
2026-03-31 |
| COMMIT |
0.00 |
feature: added import complexity checker (#45013) |
|
Simple feature summary, short updates, h |
2026-03-31 |
| COMMIT |
0.00 |
Fix tests for `janus` model (#44739) |
|
Issue/PR style with signed-off trailer, |
2026-03-31 |
| COMMIT |
0.00 |
CB improvements for serving (#45063) |
|
All items are terse commit messages typi |
2026-03-30 |
| COMMIT |
0.00 |
Add Music Flamingo (#43538) |
|
Human-typical terse, technical commit me |
2026-03-30 |
| COMMIT |
0.00 |
[docs] continuous batching (#44896) |
|
Brief, informal, technical wording; no A |
2026-03-30 |
| COMMIT |
0.00 |
Fix few issues in Qwen_3_Omni_Moe (#44848) |
|
Commit messages are typical, concise dev |
2026-03-30 |
| COMMIT |
0.00 |
Fix PP test_ocr_queries (#45123) |
|
Short technical fix message, human style |
2026-03-30 |
| COMMIT |
0.00 |
Fix TypeError in rope validation when ignore_keys is a list |
|
Technical explanation with specific deta |
2026-03-30 |
| COMMIT |
0.00 |
refactor: added cache in check_repo (#45012) |
|
Follows conventional commit style; infor |
2026-03-30 |
| COMMIT |
0.00 |
Remove unused TensorFlow env var (#45065) |
|
Very terse, typical human commit style, |
2026-03-27 |
| COMMIT |
0.00 |
fix: add identity reverse_op to dequantize ops for save_pret |
|
Technical, domain-specific, informal, hu |
2026-03-27 |
| COMMIT |
0.00 |
Fix when RoPE params are in kwargs (#45049) |
|
Concise, technical phrasing, typical of |
2026-03-27 |
| COMMIT |
0.00 |
chore: update update_metdata.yml (#45054) |
|
— |
2026-03-27 |
| COMMIT |
0.00 |
[`FA`] Fix BC support for a few versions + add deprecation c |
|
Commit is terse, informal, and uses doma |
2026-03-27 |
| COMMIT |
0.00 |
fix(testing): Fix Parakeet, Evolla, Pi0, and Phi-3 test fail |
|
Commit and co-authored-by trailer show h |
2026-03-27 |
| COMMIT |
0.00 |
Allow advanced users to override `model_type` in `AutoConfig |
|
Brief technical summary; concise, domain |
2026-03-27 |
| COMMIT |
0.00 |
Fix llama4 bnb mode (#44588) |
|
Commit log style; terse, technical, no A |
2026-03-27 |
| COMMIT |
0.00 |
Fix failing `SmolLM3IntegrationTest` (#45048) |
|
Minimal content; repetitive test referen |
2026-03-27 |
| COMMIT |
0.00 |
fix tests/quantization/fp_quant_integration/test_fp_quant.py |
|
Technical jargon, short notes, domain-sp |
2026-03-27 |
| COMMIT |
0.00 |
chore: remove old extras (#45024) |
|
Concise, informal domain language; human |
2026-03-27 |
| COMMIT |
0.00 |
Avoid `Image.open` failure (#44645) |
|
Technical fixes, terse comments, human c |
2026-03-27 |
| COMMIT |
0.00 |
chore: Fix mlinter cache location (#45052) |
|
Brief summary; informal, domain-jargon, |
2026-03-27 |
| COMMIT |
0.00 |
Embedding VLMs don't need a head (#45000) |
|
Informal notes; terse commit style, doma |
2026-03-27 |
| COMMIT |
0.00 |
Fix GraniteConfig type hints to accept int for multiplier fi |
|
Standard technical explanation; concise, |
2026-03-27 |
| COMMIT |
0.00 |
fix: preserve rotary_pct across save/load cycle in GPTNeoX c |
|
Human commit style; technical details, p |
2026-03-27 |
| COMMIT |
0.00 |
refactor: speed up docstring checker (#45009) |
|
Commit messages are terse and domain-spe |
2026-03-27 |
| COMMIT |
0.00 |
ci: add anti-slop action (#44847) |
|
— |
2026-03-26 |
| COMMIT |
0.00 |
Add doc page for capturing outputs (#44947) |
|
— |
2026-03-26 |
| PR |
0.00 |
changes |
|
Template content only, no free-text from |
2026-04-08 |
| PR |
0.00 |
Fix AttributeError in AssistantToTargetTranslator.unmap_inpu |
|
Minimal free-text, template section, no |
2026-04-08 |
| PR |
0.00 |
tweak checkers output on errors |
|
Terse description and informal language; |
2026-04-01 |
| PR |
0.00 |
[docs] static model rules |
|
Informal tone, direct feedback, lacks AI |
2026-04-03 |
| PR |
0.00 |
fix(generation): beam sample when num_beams * vocab_size exc |
|
PR template structure; human content, in |
2026-04-05 |
| PR |
0.00 |
Fix FA2 inference equivalence failures for Whisper (closes # |
|
Direct, informal technical writing; snip |
2026-04-07 |
| PR |
0.00 |
Fix Nemotron-H: add mlp layer type support |
|
Domain-specific, technical, some truncat |
2026-04-07 |
| PR |
0.00 |
Add GGUF support to Gemma4 (31B & 26B-A4B) text |
|
PR template structure only, no AI hallma |
2026-04-07 |
| PR |
0.00 |
throw error when conversion required |
|
Very terse style, casual, and includes a |
2026-03-27 |
| PR |
0.00 |
fix(ernie4_5_vl_moe): resolve three config loading failures |
|
Technically detailed, concise, problem-f |
2026-04-07 |
| PR |
0.00 |
Fix export for gemma4 and add Integration tests |
|
Extremely terse; informal and typical hu |
2026-04-07 |
| PR |
0.00 |
Configuration insoncistencies |
|
Uses domain-specific terms and bullet li |
2026-04-02 |
| PR |
0.00 |
Generic Sequence Classifier works for multimodal models |
|
Technical shorthand and references; info |
2026-03-13 |
| PR |
0.00 |
Fix CB Accuracy Regression under FA2 |
|
Technical detail and casual error explan |
2026-04-07 |
| PR |
0.00 |
Add edge case tests for out-of-range token id decoding in Qw |
|
Concise, technical, no AI hallmarks; lik |
2026-04-02 |
| PR |
0.00 |
Fix KeyError in apply_chat_template when message has no cont |
|
Uses domain terms, error reference, and |
2026-04-08 |
| PR |
0.00 |
Fix vllm cis |
|
Extremely brief and informal; clear huma |
2026-03-31 |
| PR |
0.00 |
Fix NaN weights on non-rank-0 FSDP processes |
|
Direct, technical, domain-jargon; matche |
2026-03-27 |
| PR |
0.00 |
Fix Zamba2MambaMixer ignoring use_mamba_kernels=False |
|
Concise and technical fix description ty |
2026-03-19 |
| PR |
0.00 |
Fix `Qwen2IntegrationTest` |
|
Lists technical changes and references; |
2026-04-06 |
| PR |
0.00 |
cohere_asr: fix bug for model_parallel_beam_search test case |
|
Brief, technical, includes reference to |
2026-04-03 |
| PR |
0.00 |
fix(security): prevent untrusted users from triggering TRL C |
|
Clear, specific, technically focused and |
2026-04-07 |
| PR |
0.00 |
fix bug for videomt model device mismatch |
|
Direct request, informal language, abbre |
2026-04-03 |
| PR |
0.00 |
[AMD CI] Fix Qwen2 expectations |
|
Direct reference to related PR, brief, a |
2026-04-07 |
| PR |
0.00 |
fix(cohere_asr): auto-fix failing tests |
|
Single-line, terse, ticket-style; clearl |
2026-04-07 |
| PR |
0.00 |
fix(videomt): auto-fix failing tests |
|
Short, uses typical fix notation, not AI |
2026-04-07 |
| PR |
0.00 |
fix(nomic_bert): auto-fix failing tests |
|
Short, terse, domain-style commit label. |
2026-04-07 |
