| COMMIT |
1.00 |
Add Granite 4.1 Vision (granite4_vision) (#45597) |
|
Commit message contains explicit AI assi |
2026-05-05 |
| COMMIT |
1.00 |
Unwrap `text_config` in `AutoModelFor*.from_config` (#45770) |
|
Commit message contains explicit AI assi |
2026-05-05 |
| COMMIT |
1.00 |
Add EXAONE 4.5 implementations (#45471) |
|
Commit message contains explicit AI assi |
2026-05-04 |
| COMMIT |
1.00 |
Add DeepSeek V4 (#45643) |
|
Commit message contains explicit AI assi |
2026-05-02 |
| COMMIT |
1.00 |
Support for a new Granite-Speech-Plus model (#45695) |
|
Commit message contains explicit AI assi |
2026-04-29 |
| COMMIT |
1.00 |
fixing more typos (#45689) |
|
Commit message contains explicit AI assi |
2026-04-28 |
| COMMIT |
1.00 |
Fix configuration reading and error handling for kernels (#4 |
|
Commit message contains explicit AI assi |
2026-04-23 |
| COMMIT |
1.00 |
Fix `AttributeError` on `s_aux=None` in `flash_attention_for |
|
Commit message contains explicit AI assi |
2026-04-23 |
| COMMIT |
1.00 |
Add expert parallelism (EP) config support for Qwen3 MoE (# |
|
Commit message contains explicit AI assi |
2026-04-22 |
| COMMIT |
1.00 |
[`Privacy Filter`] Add model (#45580) |
|
Commit message contains explicit AI assi |
2026-04-22 |
| PR |
1.00 |
DeepGEMM BF16 + mixed FP8/FP4 + MegaMoE + refactor |
|
PR body explicitly mentions AI collabora |
2026-04-24 |
| PR |
1.00 |
Pass packed boundary metadata to Qwen3.5 linear-attention fa |
|
PR body explicitly mentions AI collabora |
2026-03-26 |
| PR |
1.00 |
fix: ModuleNotFoundError caused by distributed race conditio |
|
PR body explicitly mentions AI collabora |
2026-05-05 |
| PR |
1.00 |
Extract dynamic vision/audio tensors into standalone pure fu |
|
PR body explicitly mentions AI collabora |
2026-04-13 |
| PR |
1.00 |
Add Xiaomi MiMo-V2 |
|
PR body explicitly mentions AI collabora |
2026-03-31 |
| PR |
1.00 |
fix: correct spelling in continuous_api docstring |
|
PR body explicitly mentions AI collabora |
2026-05-03 |
| PR |
1.00 |
Fix link to modular transformers documentation |
|
PR body explicitly mentions AI collabora |
2026-05-02 |
| PR |
1.00 |
First model |
|
PR body explicitly mentions AI collabora |
2026-05-05 |
| PR |
1.00 |
docstring |
|
PR body explicitly mentions AI collabora |
2026-05-05 |
| PR |
1.00 |
fix: Added Mps support in float fallback backends list |
|
PR body explicitly mentions AI collabora |
2026-04-28 |
| PR |
1.00 |
Fix split batch size |
|
PR body explicitly mentions AI collabora |
2026-05-02 |
| PR |
1.00 |
.. |
|
PR body explicitly mentions AI collabora |
2026-05-05 |
| PR |
1.00 |
Exclude audio modules from conversion process |
|
PR body explicitly mentions AI collabora |
2026-04-28 |
| PR |
1.00 |
Add Conformer model |
|
PR body explicitly mentions AI collabora |
2026-05-03 |
| PR |
1.00 |
torch.backends.fp32_precision cascade conv/rnn so removing t |
|
PR body explicitly mentions AI collabora |
2026-05-04 |
| PR |
1.00 |
feat: Add GGUF loading support for Llama 4 (text) |
|
PR body explicitly mentions AI collabora |
2026-04-21 |
| PR |
1.00 |
Add EXAONE 4.5 implementations |
|
PR body explicitly mentions AI collabora |
2026-04-16 |
| PR |
1.00 |
Add ctsm model |
|
PR body explicitly mentions AI collabora |
2026-04-17 |
| PR |
1.00 |
Better Grouped GEMM + EP |
|
PR body explicitly mentions AI collabora |
2026-04-24 |
| PR |
1.00 |
test pull request |
|
PR body explicitly mentions AI collabora |
2026-05-04 |
| PR |
1.00 |
[MINISTRAL3] Fix conversion script yarn's apply_scale suppor |
|
PR body explicitly mentions AI collabora |
2026-05-02 |
| PR |
0.50 |
[Weight Converter] More fine-grained mappings on classes, sc |
|
Slight AI hallmark with 'This PR aims... |
2026-04-27 |
| PR |
0.20 |
[codex] save codebase deep dive audio progress |
|
Concise changelog, slight formality but |
2026-05-05 |
| PR |
0.20 |
Adding support for Nandi Models |
|
Polite, formal tone in free text; possib |
2026-03-29 |
| PR |
0.20 |
Fix gemma gguf tokenizer |
|
Technical root cause analysis; clear, no |
2025-10-15 |
| PR |
0.20 |
Docs add custom loss example |
|
Standard changelog language, focused and |
2025-10-15 |
| PR |
0.20 |
Add regression test for MusicgenMelody audio conditioning (G |
|
Domain-specific, clear regression test s |
2026-05-01 |
| PR |
0.15 |
Qwen3 ASR and Forced Aligner |
|
Simple, template-driven; minimal added t |
2026-02-08 |
| PR |
0.15 |
[SAM3] Enable single-scale input support in Mask Decoder |
|
Technical domain detail and concise summ |
2025-12-26 |
| COMMIT |
0.10 |
[nemotron_h] respect _no_reinit flag on dt_bias and out_proj |
|
Commit contains domain context and code |
2026-05-01 |
| COMMIT |
0.10 |
Doc translate to Persian(farsi) (#45664) |
|
Slightly formal in part, but domain-spec |
2026-04-30 |
| COMMIT |
0.10 |
docs(README_zh-hans): clarify conditions for not using Trans |
|
Somewhat formal wording but still within |
2026-04-28 |
| PR |
0.10 |
Extended n-to-1 kernel fusion via `KernelConfig` |
|
Technical detail and normal API discussi |
2026-04-27 |
| PR |
0.10 |
feat: add bf16_loss training argument for VRAM-efficient QLo |
|
Domain-specific jargon, minimal formalit |
2026-05-03 |
| PR |
0.10 |
Add Granite 4.1 Vision (granite4_vision) |
|
Product announcement tone, but domain-ri |
2026-04-23 |
| PR |
0.10 |
No agent PR descriptions |
|
Informal tone and incomplete sentences i |
2026-05-05 |
| PR |
0.10 |
fix: align attention_mask padding with appended eos token in |
|
Technical, terse summary and domain deta |
2026-05-03 |
| PR |
0.10 |
fix attribute access in PermuteForRope._apply |
|
Describes a specific bug with concise, d |
2026-05-03 |
| PR |
0.10 |
fix: return correct forward output in AriaTextForCausalLM |
|
Terse, technical, directly references mo |
2026-05-03 |
| PR |
0.10 |
Optimize Qwen3 RoPE: precompute cos/sin cache for static rop |
|
Technical content, direct reference to m |
2026-05-02 |
| PR |
0.10 |
fix(bitsandbytes): implement reverse_op for Bnb4bitDeseriali |
|
Shows domain context, uses common model |
2026-05-02 |
| PR |
0.10 |
Fix IndexError in sdpa_mask and flex_attention_mask for 0D t |
|
References a specific issue, clear techn |
2026-05-02 |
| PR |
0.10 |
fix: remove upper bound on tokenizers version constraint |
|
Specific dependency update with technica |
2026-05-05 |
| PR |
0.10 |
[CB] Fixes for SDPA and CPU offloading |
|
Uses domain jargon and technical detail; |
2026-05-01 |
| PR |
0.10 |
Fix "AttributeError: NewTokenizer has no attribute special_a |
|
Jargon and specific bug reference; not o |
2026-04-07 |
| PR |
0.10 |
DeepSeek OCR specifies an incorrect tokenizer class on the H |
|
Direct and technical; references code an |
2026-05-01 |
| PR |
0.10 |
feat: add crop() to StaticCache layers for assisted generati |
|
Informal, technical jargon and abbreviat |
2026-05-02 |
| PR |
0.10 |
Add DeepSeek V4 |
|
Short, informal, includes human reviewer |
2026-04-25 |
| PR |
0.10 |
Add V-JEPA 2.1 inference support |
|
Technical, context-specific detail—human |
2026-04-17 |
| PR |
0.10 |
[nemotron_h] respect _no_reinit flag on dt_bias and out_proj |
|
PR content uses technical terms and cont |
2026-04-23 |
| PR |
0.10 |
Add RF-DETR |
|
Direct, minimal explanation; no AI-style |
2025-03-21 |
| PR |
0.10 |
fix: AutoConfig reloads wrong class after save_pretrained + |
|
Domain-specific, explanatory, informal t |
2026-04-30 |
| PR |
0.10 |
🚨 Get rid of most Apex references |
|
Domain-specific context and informal, tr |
2026-04-30 |
| PR |
0.10 |
[CB] Better overall script and decode bucketting |
|
Technical summary, realistic usage, not |
2026-04-27 |
| PR |
0.10 |
[docs] model testing |
|
Brief, informal, domain-specific phrases |
2026-03-31 |
| COMMIT |
0.05 |
Add image processors refactor to v5 migration guide (#45556) |
|
Brief, changelog-focused; no AI hallmark |
2026-04-28 |
| COMMIT |
0.05 |
[docs] modular transformers (#45327) |
|
Standard PR commit log, minimal free tex |
2026-04-28 |
| COMMIT |
0.05 |
[docs] dtype (#45659) |
|
Short, lacks AI-generated phrasing or ex |
2026-04-28 |
| COMMIT |
0.05 |
[docs] cb memory management (#45587) |
|
Extremely brief chunked log, no AI phras |
2026-04-28 |
| COMMIT |
0.05 |
[docs] cpu offloading (#45660) |
|
Minimal, no formal language or AI stylis |
2026-04-28 |
| COMMIT |
0.05 |
Fix `x_clip`: 8 failed test cases (#45394) |
|
Test fix, highly specific, fully normal |
2026-04-28 |
| PR |
0.05 |
Fix UnboundLocalError for `is_updated` in encoder-decoder cr |
|
Contains domain detail and abrupt cutoff |
2026-05-04 |
| PR |
0.05 |
audio tester class |
|
Technical, concise, some cutoff, domain |
2026-04-13 |
| PR |
0.05 |
TP refactor for FSDP + TP integration |
|
Very terse, domain-specific notes; infor |
2026-03-26 |
| PR |
0.05 |
chore(mlinter): implement part of rule TRF003 |
|
Brief, domain-specific, terse phrasing; |
2026-04-30 |
| PR |
0.05 |
perf(qwen3_vl): replace Conv3d with F.linear in patch embed |
|
Technical explanation and cutoff; lacks |
2026-05-04 |
| PR |
0.05 |
[CB] Refactor any model-related code in a separate class |
|
Domain jargon and summary with incomplet |
2026-04-27 |
| PR |
0.05 |
Fix tf32 issue: set `torch.backends.cudnn.conv.fp32_precisio |
|
Direct, technical, informal, cutoff; hum |
2026-04-05 |
| PR |
0.05 |
[Flax] Fix eval and data_args usage in streaming example |
|
Concise, domain-specific explanation wit |
2021-10-23 |
| PR |
0.05 |
PythonBackend slow tokenizer convert_ids_to_tokens fix |
|
Brief, technical fix with specific conte |
2026-04-30 |
| PR |
0.05 |
fix(processing): Filter kwargs in ProcessorMixin call to pre |
|
Direct, technical description; lacks AI- |
2025-10-15 |
| PR |
0.05 |
fix(musicgen_melody): use DynamicCache instead of EncoderDec |
|
Domain-specific language, concise, clear |
2026-05-01 |
| PR |
0.05 |
fix(quantizers): make user-supplied skip_modules additive wi |
|
Concise, technical changes; domain langu |
2026-05-01 |
| PR |
0.05 |
fix(utils): Resolve backbone utils test regressions |
|
Uses changelog format, domain references |
2026-04-23 |
| PR |
0.01 |
Add new model: Kimi2-6 |
|
Minimal free-text, just a dot; clearly h |
2026-04-24 |
| PR |
0.01 |
Add Molmo2 |
|
No actual free-text content filled; huma |
2026-01-23 |
| COMMIT |
0.00 |
fix: forward use_cache kwarg to attention mixer in nemotron_ |
|
Brief technical commit messages with dom |
2026-05-05 |
| COMMIT |
0.00 |
fix: correct spelling in continuous_api docstring (#45749) |
|
Minimal, terse technical correction; typ |
2026-05-05 |
| COMMIT |
0.00 |
Fix link to modular transformers documentation (#45746) |
|
Direct update explanation; clear, concis |
2026-05-05 |
| COMMIT |
0.00 |
Gemma4: fix failed test cases (#45568) |
|
List of technical fixes and updates, sig |
2026-05-05 |
| COMMIT |
0.00 |
First model (#45788) |
|
Casual tone and specific jargon; reflect |
2026-05-05 |
| COMMIT |
0.00 |
Fix CI: Allow more artifacts to be download in CI (#45785) |
|
Debug note and domain-specific context, |
2026-05-05 |
| COMMIT |
0.00 |
Add `concurrency` to `PR CI` workflow file (`pr-ci-caller.ym |
|
Single word commit, terse; standard huma |
2026-05-05 |
| COMMIT |
0.00 |
Reorder decorators for autodoc and dataclass (#45702) |
|
Short, technical log phrases typical of |
2026-05-05 |
| COMMIT |
0.00 |
deepseek r1 distilled tokenizer fix for qwen2 mapping (#4574 |
|
Typo present, informal tone; clear human |
2026-05-05 |
| COMMIT |
0.00 |
fix: Added Mps support in float fallback backends list (#45 |
|
Structured code fixes, informal language |
2026-05-05 |
| COMMIT |
0.00 |
Github Actions PR CI (caller) (#45476) |
|
Commit message is terse and includes dom |
2026-05-04 |
| COMMIT |
0.00 |
make sure we call check_auto in CI (#45775) |
|
Informal tone; domain-specific and conci |
2026-05-04 |
| COMMIT |
0.00 |
Better Grouped GEMM + EP (#45621) |
|
Commit log is informal, terse, and techn |
2026-05-04 |
| COMMIT |
0.00 |
DeepSeek OCR specifies an incorrect tokenizer class on the H |
|
Human-written; issue title is concise an |
2026-05-04 |
| COMMIT |
0.00 |
Fix auto mapping script (#45774) |
|
Extremely terse and informal commit mess |
2026-05-04 |
| COMMIT |
0.00 |
PythonBackend slow tokenizer convert_ids_to_tokens fix (#457 |
|
Commit message is terse and lacks AI-sty |
2026-05-04 |
| COMMIT |
0.00 |
[MINISTRAL3] Fix conversion script yarn's apply_scale suppor |
|
Commit message is terse and technical; n |
2026-05-03 |
| COMMIT |
0.00 |
🚨 Get rid of most Apex references (#45723) |
|
Concise and informal message, domain-typ |
2026-05-01 |
| COMMIT |
0.00 |
fix(utils): Resolve backbone utils test regressions (#45594) |
|
Short, domain-specific commit message; t |
2026-05-01 |
| COMMIT |
0.00 |
[CB] Better overall script and decode bucketting (#45653) |
|
Informal, terse checklist style is human |
2026-05-01 |
| COMMIT |
0.00 |
[docs] model testing (#45152) |
|
Terese commit messages; human style; dom |
2026-04-30 |
| COMMIT |
0.00 |
update dev (#45726) |
|
Brief update message; lacks AI hallmarks |
2026-04-30 |
| COMMIT |
0.00 |
[`OAI Privacy Filter`] Add integration test (#45725) |
|
Short, action-based messages; human comm |
2026-04-30 |
| COMMIT |
0.00 |
Speedup Qwen2VLImageProcessor (#45719) |
|
Technical, terse commit messages; human |
2026-04-30 |
| COMMIT |
0.00 |
Remove dead beam-search dummies from dummy_pt_objects.py (#4 |
|
Single, precise technical commit message |
2026-04-30 |
| COMMIT |
0.00 |
[Model] Add PP-FormulaNet Model Support (#45626) |
|
Terse, domain-specific commits; includes |
2026-04-30 |
| COMMIT |
0.00 |
chore(typing): add ty type checking for 10 utility files (#4 |
|
Clear domain context and project-specifi |
2026-04-30 |
| COMMIT |
0.00 |
[serve] cb error (#45691) |
|
Brief commit log format, no AI signals. |
2026-04-29 |
| COMMIT |
0.00 |
Fix trust_remote_code local cache collisions for local model |
|
Technical commit log, concise, no AI hal |
2026-04-29 |
| COMMIT |
0.00 |
Llama3 video fix (#45040) |
|
Standard iterative commit log, domain te |
2026-04-29 |
| COMMIT |
0.00 |
[Fix Phi4 test] Fall back to model config for image processo |
|
Technical changelog, brief summary, no A |
2026-04-29 |
| COMMIT |
0.00 |
Fix custom-module copies inheriting read-only permissions (# |
|
Detailed technical context, not overly f |
2026-04-29 |
| COMMIT |
0.00 |
Python code in model docs (#45608) |
|
Informal, terse commit log, domain-speci |
2026-04-29 |
| COMMIT |
0.00 |
fix failed test cases for blt model (#45596) |
|
Technical commit log, signed off by huma |
2026-04-29 |
| COMMIT |
0.00 |
chore(typing): add ty type checking for 3 pipeline files (#4 |
|
Standard technical changelog and co-auth |
2026-04-29 |
| COMMIT |
0.00 |
change got reverted (#45680) |
|
Extremely terse, typical human revert me |
2026-04-28 |
| COMMIT |
0.00 |
fix padding side issue for fast_vlm tests (#45592) |
|
Terse, domain-specific, includes real na |
2026-04-28 |
| COMMIT |
0.00 |
zero_shot_object_detection ValueError fix for python 3.13 (# |
|
Very brief, precise, with clear domain r |
2026-04-28 |
| COMMIT |
0.00 |
Fix pageable H2D copies in Gated DeltaNet PyTorch fallback ( |
|
Concise, technical commit message with d |
2026-04-28 |
| COMMIT |
0.00 |
Fix UnboundLocalError in shard_and_distribute_module for rep |
|
Short, informal message; human co-author |
2026-04-28 |
| COMMIT |
0.00 |
No serving in quality docker image (#45677) |
|
Brief commit message, human co-author, n |
2026-04-28 |
| COMMIT |
0.00 |
Laguna XS.2 implementation (#45673) |
|
Minimal title-only commit, no sign of AI |
2026-04-28 |
| COMMIT |
0.00 |
[MistralCommonBackend] Soften validation mode and apply_chat |
|
Structured changelog, domain-specific co |
2026-04-28 |
| COMMIT |
0.00 |
Fix `NameError: PeftConfigLike` triggered by `PreTrainedMode |
|
Concise, code-centric, domain-specific c |
2026-04-27 |
| COMMIT |
0.00 |
Fix cross-attention cache layer type for T5Gemma2 long input |
|
Terse, technical, and human-typical with |
2026-04-27 |
| COMMIT |
0.00 |
chore(typing): added modeling_utils to ty (#45425) |
|
Informal, uses jargon and review summari |
2026-04-27 |
| COMMIT |
0.00 |
model: Add DEIMv2 to Transformers (#44339) |
|
Uses changelog format, dense with domain |
2026-04-27 |
| COMMIT |
0.00 |
[Qwen3.5] Fix GDN linear attention multi-token cached forwar |
|
Detailed description with human-like bug |
2026-04-27 |
| COMMIT |
0.00 |
[gemma4] infer from config instead of hardcoding (#45606) |
|
Informal, concise updates and normal cod |
2026-04-27 |
| COMMIT |
0.00 |
Update quants tests (#45480) |
|
Minimalist, lacks AI style, uses terse c |
2026-04-27 |
| COMMIT |
0.00 |
Fix GraniteMoeHybrid _update_mamba_mask crash on attention-o |
|
Technical summary with rationale, inform |
2026-04-27 |
| COMMIT |
0.00 |
🔴🔴🔴 fix: skip `clean_up_tokenization` for BPE tokenizers in |
|
Patch details, domain-specific explanati |
2026-04-27 |
| COMMIT |
0.00 |
Fix colmodernvbert tests (#45652) |
|
Very terse, informal, intentionally mini |
2026-04-27 |
| COMMIT |
0.00 |
[CB] [Major] Add CPU request offloading (#45184) |
|
Commit messages are terse, informal, and |
2026-04-27 |
| COMMIT |
0.00 |
Fix peft constructors (#45622) |
|
Very terse and non-formal commit, human |
2026-04-27 |
| COMMIT |
0.00 |
chore: speedup modular converter (~30%) (#45046) |
|
Highly technical, terse, and informal; n |
2026-04-27 |
| COMMIT |
0.00 |
Fix whisper return language (#42227) |
|
Technical commit log with co-author trai |
2026-04-27 |
| COMMIT |
0.00 |
Add `supports_gradient_checkpointing` to `NemotronHPreTraine |
|
Short, direct commit style typical of hu |
2026-04-27 |
| COMMIT |
0.00 |
Raise clear error for `problem_type="single_label_classifica |
|
Explanation is technical with domain det |
2026-04-24 |
| COMMIT |
0.00 |
CircleCI with torch 2.11 (#45633) |
|
Repetitive commit summary, highly typica |
2026-04-24 |
| COMMIT |
0.00 |
chore: bump doc-builder SHA for main doc build workflow (#45 |
|
Standard terse chore commit, no AI style |
2026-04-24 |
| COMMIT |
0.00 |
Allow more artifacts to be download in CI (#45629) |
|
Sparse, informal; lacks signature AI phr |
2026-04-24 |
| COMMIT |
0.00 |
chore(qa): split pipeline and add type checking (#45432) |
|
Contains abbreviations and minimal phras |
2026-04-24 |
| COMMIT |
0.00 |
Skip failing offloading tests (#45624) |
|
Commit messages are brief and telegraphi |
2026-04-24 |
| COMMIT |
0.00 |
generate: drop stale num_return_sequences warning on continu |
|
Technical justification, abbreviations, |
2026-04-24 |
| COMMIT |
0.00 |
Remove unnecessary generate warnings (#45619) |
|
Brief, imperative commit style, no AI si |
2026-04-24 |
| COMMIT |
0.00 |
fix: compute auxiliary losses when denoising is disabled in |
|
Commit uses terse, technical language an |
2026-04-23 |
| COMMIT |
0.00 |
qa: bumped mlinter and allow local override (#45585) |
|
Informal commit lines and explicit human |
2026-04-23 |
| COMMIT |
0.00 |
Processing Utils: continue when content is a string (#45605) |
|
Terse commit message; direct technical f |
2026-04-23 |
| COMMIT |
0.00 |
SonicMoe (#45433) |
|
Informal, iterative commit messages with |
2026-04-23 |
| COMMIT |
0.00 |
fix transformers + torchao nvfp4 serialization (#45573) |
|
Casual tone, human-specific comments, an |
2026-04-23 |
| COMMIT |
0.00 |
[AMD CI] Fix expectations for Gemma3n (#45602) |
|
Short, lower-case, informal change descr |
2026-04-23 |
| COMMIT |
0.00 |
[docs] multi-turn tool calling (#45554) |
|
Very terse, typical human commit without |
2026-04-23 |
| COMMIT |
0.00 |
Allow for registered experts from kernels hub (#45577) |
|
Iterative, collaborative commit history |
2026-04-23 |
| COMMIT |
0.00 |
[CB] Changes for long generation (#45530) |
|
Highly informal, domain-heavy and collab |
2026-04-23 |
| COMMIT |
0.00 |
Align latest model attention function dispatch (#45598) |
|
Extremely terse, lacks AI-writing hallma |
2026-04-23 |
| COMMIT |
0.00 |
Gemma3n and Gemma4 cannot use rotary kernel (#45564) |
|
Short, direct human style; contains tech |
2026-04-23 |
| COMMIT |
0.00 |
do not index past decoded chars with special tokens (#45435) |
|
Informal tone, domain-specific phrasing, |
2026-04-22 |
| COMMIT |
0.00 |
Update dev version (#45583) |
|
Brief, informal; lacks AI stylistic cues |
2026-04-22 |
| COMMIT |
0.00 |
Update torchao usage for XPU and CPU (#45560) |
|
Very terse, technical, humanlike style. |
2026-04-22 |
| COMMIT |
0.00 |
[docs] per-request sampling params (#45553) |
|
Minimal, informal; clear domain-specific |
2026-04-22 |
| PR |
0.00 |
[docs] adding audio/video processors |
|
Terse, informal, linked commentary—hallm |
2026-05-05 |
| PR |
0.00 |
Fix WeightConverter substring match on leaf-style source pat |
|
Technical explanation with in-line code, |
2026-05-05 |
| PR |
0.00 |
Gate enable_gqa=True on actual flash-attention eligibility |
|
PR uses template; detailed technical con |
2026-05-04 |
| PR |
0.00 |
fix: forward use_cache kwarg to attention mixer in nemotron_ |
|
Informal commit message, addresses typo, |
2026-05-05 |
| PR |
0.00 |
fix: validate special token ids against attribute values |
|
Technical bugfix explanation, no AI phra |
2026-05-05 |
| PR |
0.00 |
[docs] contributing |
|
Informal bullet list, typical human styl |
2026-04-15 |
| PR |
0.00 |
feat(llama): add has_weight parameter to LlamaRMSNorm for Fl |
|
Domain terms, technical context, and nat |
2026-05-02 |
| PR |
0.00 |
Fix xdist collisions for captured_info artifacts and preserv |
|
Bug context, references, and non-boilerp |
2026-04-25 |
| PR |
0.00 |
Unwrap `text_config` in `AutoModelFor*.from_config` |
|
PR uses template; filled content is dire |
2026-05-04 |
| PR |
0.00 |
Gemma4: fix failed test cases |
|
Bulleted list, informal typos, and techn |
2026-04-22 |
| PR |
0.00 |
Fix WeightConverter regex incorrectly matching shared_expert |
|
Well-structured, technical regex discuss |
2026-05-05 |
| PR |
0.00 |
[generation] Encode multimodal data only once |
|
Direct technical explanation and jargon; |
2026-05-05 |
| PR |
0.00 |
fix: per-instance cache in compile_compatible_method_lru_cac |
|
Terse summary and technical details, wit |
2026-05-05 |
| PR |
0.00 |
fix(testing): use worker-specific captured_info files for py |
|
Problem statement and technical domain c |
2026-05-05 |
| PR |
0.00 |
Fix CI: Allow more artifacts to be download in CI |
|
Informal language and natural error expl |
2026-05-05 |
| PR |
0.00 |
Modularize `ProcessorMixin` into smaller components |
|
PR uses template; content is technical a |
2026-04-17 |
| PR |
0.00 |
Add `concurrency` to `PR CI` workflow file (`pr-ci-caller.ym |
|
Technical file/path reference and concis |
2026-05-05 |
| PR |
0.00 |
Blockwise mask fn as opt arg in all masking functions |
|
Informal, to-the-point, abbreviations an |
2026-04-16 |
| PR |
0.00 |
Reorder decorators for autodoc and dataclass |
|
Informal, mentions 'smth', personal tone |
2026-04-29 |
| PR |
0.00 |
deepseek r1 distilled tokenizer fix for qwen2 mapping |
|
Extremely terse, informal, no AI or temp |
2026-05-02 |
| PR |
0.00 |
Github Actions PR CI (caller) |
|
Terse and direct; not AI-like. |
2026-04-16 |
| PR |
0.00 |
Fix mps device check for moe histogram routing |
|
Uses domain-specific wording; direct fix |
2026-05-03 |
| PR |
0.00 |
refector: renamed file glob to cache to make it clearer |
|
Informal and brief, with human-like revi |
2026-04-30 |
| PR |
0.00 |
make sure we call check_auto in CI |
|
Refers to issue/commit in direct and typ |
2026-05-04 |
| PR |
0.00 |
Add deepseek 3.2 exp |
|
Contains code and minimal description; h |
2025-10-01 |
| PR |
0.00 |
Fix auto mapping script |
|
Informal free-text and error context; cl |
2026-05-04 |
| PR |
0.00 |
End-to-end test of Gemma 3 + FA2 construction |
|
Domain-specific and concise; template se |
2026-05-03 |
| PR |
0.00 |
Fix unhandled exception noise from background safetensors co |
|
Technical description, concise, with dom |
2026-05-03 |
| PR |
0.00 |
Require input_ids for repetition penalty |
|
Domain jargon and warning explanation; h |
2026-04-13 |
| PR |
0.00 |
fix: restore vocabulary loading in CamembertTokenizer |
|
Concise, domain-specific; clear human st |
2026-04-30 |
| PR |
0.00 |
Added library_name and library_version to HfApi/Hub calls fo |
|
REPO PR TEMPLATE, but free-text is conci |
2026-04-30 |
| PR |
0.00 |
[skills] fine-tuning |
|
Casual tone, code snippet, clearly human |
2026-04-30 |
| PR |
0.00 |
[New Model] Add MiniCPM3 support |
|
REPO PR TEMPLATE; actual text is brief a |
2026-04-23 |
