| commit |
1.00 |
Synthetic test assets (#48589) |
|
Commit message contains explicit AI assi |
2026-09-10 |
| commit |
1.00 |
[MusicgenMelody] Fix conditioning silently dropped at genera |
|
Commit message contains explicit AI assi |
2026-09-10 |
| commit |
1.00 |
Assert cached decode matches recomputing without a cache (#4 |
|
Commit message contains explicit AI assi |
2026-09-10 |
| commit |
1.00 |
Mirror the registered pytree flatten when building automatic |
|
Commit message contains explicit AI assi |
2026-09-10 |
| commit |
1.00 |
[fix] Fix GlmOcr integration tests: wrong token IDs and imag |
|
Commit message contains explicit AI assi |
2026-09-09 |
| commit |
1.00 |
[CI] Deduplicate Nvidia/AMD CI reply comments and add header |
|
Commit message contains explicit AI assi |
2026-09-09 |
| commit |
1.00 |
[tests] Fix integration test golden values broken by fast im |
|
Commit message contains explicit AI assi |
2026-09-09 |
| commit |
1.00 |
Add Fun-ASR-Nano model (#46180) |
|
Commit message contains explicit AI assi |
2026-09-09 |
| commit |
1.00 |
[KimiLinear] Fix test_cpu_offload: set num_local_experts=4 i |
|
Commit message contains explicit AI assi |
2026-09-08 |
| commit |
1.00 |
Fix GPU memory teardown in CLI serve tests (#48618) |
|
Commit message contains explicit AI assi |
2026-09-08 |
| commit |
1.00 |
[fix] Fix how we read package versions - triggered by torch |
|
Commit message contains explicit AI assi |
2026-09-08 |
| commit |
1.00 |
esmfold2: keep `distogram_head` in fp32 as well (#48488) |
|
Commit message contains explicit AI assi |
2026-09-08 |
| commit |
1.00 |
[Fix] Fix A10 expectations for a test (#48454) |
|
Commit message contains explicit AI assi |
2026-09-07 |
| commit |
1.00 |
Add PR comment CI for AMD (MI300) (#48065) |
|
Commit message contains explicit AI assi |
2026-09-07 |
| commit |
1.00 |
Compress the agent conventions file and document two modular |
|
Commit message contains explicit AI assi |
2026-09-07 |
| commit |
1.00 |
Pass kwargs to the Mamba2 mixer in Nemotron-H, Falcon-H1 and |
|
Commit message contains explicit AI assi |
2026-09-04 |
| pr |
1.00 |
[Experimental] Qwen3.5-9B full fine-tuning with FSDP2 on B20 |
|
Explicitly states the PR was generated w |
2026-09-10 |
| pr |
1.00 |
[ImageGPT] Fix cross-attention crashing, dropping the encode |
|
PR body explicitly mentions AI collabora |
2026-09-10 |
| pr |
1.00 |
[Bug] Preserve MoE router auxiliary loss weighting in Traine |
|
PR body explicitly mentions AI collabora |
2026-09-10 |
| pr |
1.00 |
[Docs] Add MRoPE documentation to rope_utils.md |
|
PR body explicitly mentions AI collabora |
2026-09-10 |
| pr |
1.00 |
Add dots3-note Preview model support |
|
PR body explicitly mentions AI collabora |
2026-08-08 |
| pr |
1.00 |
Fix D-FINE / RT-DETR main loss being computed over the denoi |
|
The description explicitly states it was |
2026-09-04 |
| pr |
1.00 |
Fix `AutoModel.from_pretrained` not restoring `modules_to_sa |
|
PR body explicitly mentions AI collabora |
2026-09-07 |
| pr |
1.00 |
Add flash_attention_4 in attn_implementation AutoModel docst |
|
PR body explicitly mentions AI collabora |
2026-09-10 |
| pr |
1.00 |
DOC: Improve documentation for remap_legacy_layer_types func |
|
PR body explicitly mentions AI collabora |
2026-09-09 |
| pr |
1.00 |
Docs: Add text classification quickstart example to quicktou |
|
PR body explicitly mentions AI collabora |
2026-09-01 |
| pr |
1.00 |
[higgs_audio_v2] Fix: use config.num_codebooks in audio labe |
|
PR body explicitly mentions AI collabora |
2026-09-06 |
| pr |
1.00 |
Assert cached decode matches recomputing without a cache |
|
PR body explicitly mentions AI collabora |
2026-08-25 |
| pr |
1.00 |
Fix Qwen2.5-VL temporal RoPE for fractional video intervals |
|
Explicitly states the patch and descript |
2026-09-10 |
| pr |
1.00 |
Fix `AutoImageProcessor` requiring torchvision when only Pil |
|
The description explicitly mentions AI a |
2026-09-08 |
| pr |
1.00 |
Mirror the registered pytree flatten when building automatic |
|
PR body explicitly mentions AI collabora |
2026-09-06 |
| pr |
1.00 |
DO NOT LAND Test hfh 1.28.0.rc0 |
|
PR body explicitly mentions AI collabora |
2026-08-18 |
| pr |
1.00 |
Support loading original InternVL2 checkpoints natively |
|
The description explicitly states that A |
2026-07-03 |
| pr |
1.00 |
[fix] Fix GlmOcr integration tests: wrong token IDs and imag |
|
Explicitly states that the contribution |
2026-09-09 |
| pr |
1.00 |
[tests] Fix NougatModelIntegrationTest: pin artifact revisio |
|
Explicitly states it was generated with |
2026-09-09 |
| pr |
1.00 |
[CI] Deduplicate Nvidia/AMD CI reply comments and add header |
|
Explicitly states it was generated with |
2026-09-09 |
| pr |
0.88 |
Add files via uploAdd SCQOS governed model-transition proof |
|
Formal, abstract governance prose reads |
2026-09-10 |
| pr |
0.62 |
[DO NOT MERGE!] [fix] Update stale golden values in Musicgen |
|
Highly polished and unusually exhaustive |
2026-09-09 |
| pr |
0.48 |
[Gemma Quant] Fused Triton low-bit kernels (INT2/4/8) for Ge |
|
Polished, promotional structure has some |
2026-09-09 |
| pr |
0.20 |
Update tokenizer gguf support |
|
Brief generic wording is somewhat boiler |
2026-09-09 |
| pr |
0.16 |
fix: respect partial_rotary_factor in GPTNeoXJapaneseRotaryE |
|
Structured summary is polished but conta |
2026-09-10 |
| pr |
0.12 |
Fallback for unsupported grouped_mm tensor layouts |
|
Highly structured technical summary, but |
2026-09-09 |
| pr |
0.12 |
Add qwen3 tts |
|
Somewhat generic feature-summary languag |
2026-03-07 |
| pr |
0.10 |
[Kernels] Enable rotary_pos_emb hub kernel dispatch for part |
|
Specific kernel and model references sup |
2026-09-09 |
| pr |
0.08 |
Add validation: warn when pad_token_id is in eos_token_id li |
|
Polished structure, but the debugging de |
2026-08-17 |
| pr |
0.08 |
[MusicgenMelody] Fix conditioning silently dropped at genera |
|
Detailed, technically specific debugging |
2026-09-10 |
| pr |
0.08 |
Validate DeepSeek-V3 grouped router topology |
|
Structured summary, but routing-specific |
2026-09-09 |
| pr |
0.08 |
Module Fusion API |
|
Specific implementation details and doma |
2026-03-24 |
| pr |
0.08 |
n-to-1 kernel fusion via `KernelConfig` |
|
Technical, issue-specific explanation wi |
2026-04-10 |
| pr |
0.08 |
Extended n-to-1 kernel fusion via `KernelConfig` |
|
Detailed domain-specific design discussi |
2026-04-27 |
| pr |
0.08 |
Port ESMC and ESMFold2 to Transformers |
|
Casual tone, typos, and reviewer context |
2026-06-04 |
| pr |
0.06 |
Add Ovis2.5 model support |
|
Mostly specific project details, with on |
2026-07-30 |
| commit |
0.05 |
[nit] use requires_backends (#47576) |
|
Terse commit wording and a normal human |
2026-09-07 |
| pr |
0.05 |
fix: respect partial_rotary_factor in GPTNeoXJapaneseRotaryE |
|
Concise, domain-specific explanation and |
2026-09-10 |
| pr |
0.05 |
Add WeatherNext 2 weather forecasting model |
|
Detailed domain-specific validation and |
2026-08-10 |
| commit |
0.04 |
Retire test_multi_gpu_data_parallel_forward (#48508) |
|
Clear technical rationale and repository |
2026-09-04 |
| commit |
0.04 |
Fix VoxtralRealtime rejecting non-static cache implementatio |
|
Detailed domain-specific rationale and r |
2026-09-02 |
| pr |
0.04 |
add HyperClovaX Vision |
|
Specific project context and informal ph |
2026-02-27 |
| pr |
0.04 |
Support converting the unified Parakeet checkpoints (nvidia/ |
|
Highly specific conversion details and r |
2026-09-10 |
| pr |
0.04 |
Fix deepspeed ci |
|
Typos, informal wording, and concrete CI |
2026-09-09 |
| pr |
0.04 |
Add an FP8 backend to `QuantizedCache` |
|
Specific benchmark data and implementati |
2026-09-10 |
| commit |
0.03 |
Skip the expert-parallel sentinel masking when expert parall |
|
Detailed domain-specific changelog with |
2026-09-10 |
| commit |
0.03 |
Fix deepspeed ci (#48640) |
|
Highly specific CI diagnosis and fixes i |
2026-09-10 |
| commit |
0.03 |
Fix YOLOS device mismatch with device_map="auto" (#46886) |
|
Concrete device-map diagnosis and reposi |
2026-09-09 |
| commit |
0.03 |
[generate] stop synchronizing the accelerator on every decod |
|
Highly domain-specific performance discu |
2026-09-04 |
| commit |
0.03 |
Add support for NeuCodec (#47143) |
|
Detailed iterative commit history with h |
2026-09-02 |
| commit |
0.03 |
Allow capturing only necessary hidden_states with capture_ou |
|
Technical changelog with commit trailers |
2026-09-02 |
| pr |
0.03 |
[Umbrella] Enable FSDP2 + expert parallelism via a 2-D (fsdp |
|
Polished structure is technical and evid |
2026-08-22 |
| pr |
0.03 |
Give each tensor parallel model in a process its own device |
|
Deep PyTorch internals and a focused rep |
2026-08-28 |
| pr |
0.03 |
Register kimi_k2 as a native deepseek_v3 alias |
|
Short, repository-specific explanation w |
2026-08-27 |
| pr |
0.03 |
[Model] Add PP-DocLayoutV4 Model Support |
|
Detailed model-specific design changes a |
2026-08-28 |
| pr |
0.03 |
Adding Omnilingual ASR models |
|
Terse, unfinished notes and project-spec |
2026-01-13 |
| pr |
0.03 |
[generation] Encode multimodal data only once |
|
Informal domain-specific explanation wit |
2026-05-05 |
| pr |
0.03 |
Synthetic test assets |
|
Concise, specific description with natur |
2026-09-07 |
| pr |
0.03 |
Fix backslash handling in chat generate flags |
|
Concrete bug details, tests, and environ |
2026-09-10 |
| pr |
0.03 |
[docs] Context parallelism during eval |
|
Brief, specific documentation note with |
2026-09-09 |
| pr |
0.03 |
Qwen3.8 GGUF |
|
Casual wording, typo, and concise implem |
2026-09-09 |
| pr |
0.03 |
[fix] Remove vision-encoder-decoder from MODELS_WITH_INCORRE |
|
Specific regression analysis and checkpo |
2026-09-08 |
| commit |
0.02 |
[Fix] yolos offload issue (#48688) |
|
Terse, informal commit wording with a sp |
2026-09-10 |
| commit |
0.02 |
[higgs_audio_v2] Fix: use config.num_codebooks in audio labe |
|
Brief, concrete fix description and test |
2026-09-10 |
| commit |
0.02 |
Fix `AutoImageProcessor` requiring torchvision when only Pil |
|
Technical test details and a normal huma |
2026-09-10 |
| commit |
0.02 |
Register activation kernel layers on XPU (#47858) |
|
Specific kernel and XPU terminology with |
2026-09-10 |
| commit |
0.02 |
[GLM 5.3 Flash] Preserve original names when saving text che |
|
Concise model-specific change descriptio |
2026-09-10 |
| commit |
0.02 |
[tests] Fix NougatModelIntegrationTest: pin artifact revisio |
|
Specific test and artifact details with |
2026-09-09 |
| commit |
0.02 |
[fix] Update stale expected strings in HunYuanVL integration |
|
Specific regression details and conventi |
2026-09-09 |
| commit |
0.02 |
[fix] Update stale golden values and fix expected_logits sha |
|
Highly specific test behavior and shape |
2026-09-09 |
| commit |
0.02 |
Another day fixing CI (#48591) |
|
Brief informal commit bullets show no no |
2026-09-08 |
| commit |
0.02 |
[docs] Add a LiteRT page under community integrations (#4854 |
|
Specific tooling and documentation detai |
2026-09-07 |
| commit |
0.02 |
Kimi linear (#48250) |
|
Terse, typo-filled commit notes strongly |
2026-09-05 |
| commit |
0.02 |
CI: Point test fixtures at hf-internal-testing copies we alr |
|
Specific fixture paths and incremental i |
2026-09-04 |
| commit |
0.02 |
Fix sliding-window mask `layer_idx` in Gemma3/Gemma4 `create |
|
Specific debugging details and regressio |
2026-09-04 |
| commit |
0.02 |
Support nested FLA kernel imports for fla-core (#48221) |
|
Terse, specific technical wording with i |
2026-09-03 |
| commit |
0.02 |
Fix Inkling inputs_embeds and add more tests (#47827) |
|
Terse, domain-specific commit bullets in |
2026-09-03 |
| commit |
0.02 |
Simplify and fix qwen4 tests (#48340) |
|
Short informal commit bullets lack typic |
2026-09-03 |
| commit |
0.02 |
[docs] Partial checkpointing and group_by_length (#48463) |
|
Terse commit summaries with repository-s |
2026-09-02 |
| commit |
0.02 |
[docs] Kernel updates (#48465) |
|
Short, informal commit notes and domain- |
2026-09-02 |
| commit |
0.02 |
[`Kernels`] Enable functions into kernels registry and allow |
|
Terse implementation notes with specific |
2026-09-02 |
| commit |
0.02 |
[`Qwen4 Exp`] Use partial to avoid skipping mask more easily |
|
Brief technical notes with model-specifi |
2026-09-02 |
| commit |
0.02 |
[`DSA`] Fix vanillas SDPA with padding (#48363) |
|
Short informal commits describing specif |
2026-09-02 |
| commit |
0.02 |
Fix rotary embedding regression (#48477) |
|
Terse debugging notes with specific mode |
2026-09-02 |
| commit |
0.02 |
[MTP] Save memory by only capturing the last layer's hidden_ |
|
Terse, domain-specific title and commit- |
2026-09-02 |
| commit |
0.02 |
Support per-layer MTP configuration (#48264) |
|
Short, informal commit messages and spec |
2026-09-02 |
| commit |
0.02 |
[docs] Fix failing audio doctest (#48380) |
|
Terse commit-style wording and specific |
2026-09-01 |
| commit |
0.02 |
[docs] Fix code snippets (#47772) |
|
Brief, informal commit bullets with proj |
2026-09-01 |
| pr |
0.02 |
QA: Added a `MemoryCleanupMixin` class for tests |
|
Specific test cleanup details and typos |
2026-09-10 |
| pr |
0.02 |
Expert-parallel token dispatch: each rank trains on its own |
|
Specific benchmarks, limitations, and in |
2026-09-04 |
| pr |
0.02 |
Warm the page cache before sharded loading (opt-in via `HF_S |
|
Highly specific benchmarks, filesystem d |
2026-08-23 |
| pr |
0.02 |
Enable FSDP2 + expert parallelism via a 2-D (fsdp, tp) devic |
|
Detailed distributed-training measuremen |
2026-09-03 |
| pr |
0.02 |
[CB] Varlen path: copy only the used prefix of the index sto |
|
Includes concrete profiling findings, mo |
2026-09-06 |
| pr |
0.02 |
Tensor parallelism: let a split model be one replica among s |
|
Technical terminology, failure reproduct |
2026-09-06 |
| pr |
0.02 |
Trainer: make checkpoints work under tensor parallelism |
|
Detailed distributed checkpoint failure |
2026-09-06 |
| pr |
0.02 |
[DON'T MERGE BEFORE Sep 12 Saturday] Use huggingface_hub htt |
|
Concise, technically specific migration |
2026-09-10 |
| pr |
0.02 |
Kernel api doc, part 2 |
|
Informal review notes and specific issue |
2026-06-25 |
| pr |
0.02 |
PEFT x Dtensor-based TP integration |
|
Informal grammar and brief repository-sp |
2026-09-02 |
| pr |
0.02 |
[`MoE`] Fix eager EP |
|
Terse, domain-specific debugging details |
2026-09-09 |
| pr |
0.02 |
Register activation kernel layers on XPU |
|
Highly specific technical details and be |
2026-08-10 |
| pr |
0.02 |
[GLM 5.3 Flash] Preserve original names when saving text che |
|
Concise, repository-specific root-cause |
2026-09-10 |
| pr |
0.02 |
Ignore causal masks a bit earlier for sdpa |
|
Informal phrasing and tentative performa |
2026-08-13 |
| pr |
0.02 |
[docs] distributed training |
|
Specific documentation changes and file |
2026-03-16 |
| pr |
0.02 |
fix incorrect hub tokenizer class |
|
Lowercase style, typos, and specific tok |
2026-09-09 |
| commit |
0.01 |
[`MoE`] Fix eager EP (#48653) |
|
Extremely terse commit message with no A |
2026-09-10 |
| commit |
0.01 |
MRoPE continued (#48594) |
|
Informal phrasing, typos, and iterative |
2026-09-09 |
| commit |
0.01 |
[Quantizaiton]support 5/6/7 bits in AutoRound (#48481) |
|
Technical shorthand, iterative commits, |
2026-09-09 |
| commit |
0.01 |
[`Generate`] Avoid unconditionally downloading remote hub fi |
|
Terse, informal bullets strongly suggest |
2026-09-08 |
| commit |
0.01 |
Honor `shift_labels` in decoder-only LLM/VLM losses (#48493) |
|
Detailed domain-specific debugging and t |
2026-09-08 |
| commit |
0.01 |
[docs] Per-layer config (#48601) |
|
Extremely terse wording is characteristi |
2026-09-08 |
| commit |
0.01 |
Fix expert-parallel training: NaN gradients and missing grad |
|
Highly specific implementation details a |
2026-09-08 |
| commit |
0.01 |
Fix `generate_flags` parsing in `transformers chat` (#48597) |
|
Technical bug analysis uses precise code |
2026-09-08 |
| commit |
0.01 |
[Docker] Upgrade CPU torch to <=2.14.0, torchcodec to <=0.16 |
|
Concise dependency updates and verificat |
2026-09-08 |
| commit |
0.01 |
Fix AttributeError in gradient_checkpointing_enable(offload= |
|
Specific traceback behavior and reposito |
2026-09-08 |
| commit |
0.01 |
Add `supports_context_parallel` to `PreTrainedModel` (#48442 |
|
Terse, issue-specific wording with no AI |
2026-09-07 |
| commit |
0.01 |
[docs] mlinter reference (#48460) |
|
Minimal commit-style text, likely human- |
2026-09-07 |
| commit |
0.01 |
[GLM 5.3 Flash] Fix NaN gradients in chunked KDA (#48455) |
|
Highly domain-specific debugging evidenc |
2026-09-07 |
| commit |
0.01 |
[docs] Fix [[autodoc]] directives in ALBERT model documentat |
|
Brief, repository-specific documentation |
2026-09-07 |
| commit |
0.01 |
docs: fix docstring parameter names that do not match signat |
|
Precise parameter references and AST-bas |
2026-09-07 |
| commit |
0.01 |
docs: remove phantom parameters from docstrings (#48576) |
|
Concrete docstring discrepancies are pre |
2026-09-07 |
| commit |
0.01 |
Add h4 (#48473) |
|
Extensive model-specific implementation |
2026-09-07 |
| commit |
0.01 |
Guard against a None video processor class when the backend |
|
Specific failure path and analogous file |
2026-09-07 |
| commit |
0.01 |
Refactor GGUF to speed-up inference (#47779) |
|
Terse, fragmented commit notes and speci |
2026-09-04 |
| commit |
0.01 |
Fix failing tests for cohere_compass (#48005) |
|
Repetitive updates, sign-offs, and expli |
2026-09-04 |
| commit |
0.01 |
[Fix] Use dedicated helpers for DeepGEMM and SonicMoE tests |
|
Extremely terse commit message with no A |
2026-09-04 |
| commit |
0.01 |
fix failed test cases for glm5_next (#48497) |
|
Brief technical updates and repeated dev |
2026-09-04 |
| commit |
0.01 |
vibevoice: fix bug for quant cache (#48487) |
|
Short, informal bug-fix message with a n |
2026-09-04 |
| commit |
0.01 |
Processing tests [part 2] (#47922) |
|
Casual fragments, typos, and self-correc |
2026-09-04 |
| commit |
0.01 |
QA: Add noisy comment checker (#48484) |
|
Concise implementation notes and informa |
2026-09-04 |
| commit |
0.01 |
:rotating_light: Vision (2d/3d) rotary embeddings (#48105) |
|
Chaotic terse commit notes and typos str |
2026-09-04 |
| commit |
0.01 |
add xpu expectations for hunyuan_vl model tests (#48504) |
|
Brief technical change with a standard S |
2026-09-03 |
| commit |
0.01 |
[`Qwen 3.5 Moe`] Fix decorators (#48436) |
|
Very terse commit-style wording, includi |
2026-09-03 |
| commit |
0.01 |
Infinite loop in dependency search (#48393) |
|
Specific bug fix with a normal human co- |
2026-09-03 |
| commit |
0.01 |
Fix some tests by removing the deprecation cycle (#48503) |
|
Terse commit-style summary with specific |
2026-09-03 |
| commit |
0.01 |
Remove deprecation (#48500) |
|
Extremely terse, informal commit message |
2026-09-03 |
| commit |
0.01 |
[AfMoE] Standardize past_key_values argument naming across f |
|
Specific model terminology and issue ref |
2026-09-03 |
| commit |
0.01 |
Fix Pix2StructTextAttention init using hidden_size instead o |
|
Concise, domain-specific fix description |
2026-09-03 |
| commit |
0.01 |
Fix pre patch release utility (#48499) |
|
Terse commit message naming a specific u |
2026-09-03 |
| commit |
0.01 |
Update dev version (#48498) |
|
Brief, informal version-update message. |
2026-09-03 |
| commit |
0.01 |
Warn once when a hub-kernel function falls back to its refer |
|
Technical commit history and co-author t |
2026-09-03 |
| commit |
0.01 |
doc: fix syntax error and typos in VibeVoice documentation ( |
|
Brief title explicitly describing a synt |
2026-09-02 |
| commit |
0.01 |
Remove deprecated mask functions (#48476) |
|
Minimal commit message directly stating |
2026-09-02 |
| pr |
0.01 |
[Fix] yolos offload issue |
|
Terse, domain-specific explanation with |
2026-09-10 |
| pr |
0.01 |
Keep the expert-parallel sentinel slots out of the router gr |
|
Highly specific debugging evidence and c |
2026-09-10 |
| pr |
0.01 |
Add a unit test for BartModel to compare eager, sdpa on one |
|
Brief issue-specific description with na |
2025-07-15 |
| pr |
0.01 |
Fix Whisper beam_indices so compute_transition_scores does n |
|
Detailed implementation reasoning and ma |
2026-09-08 |
| pr |
0.01 |
Loading: convert transposed and RoPE-permuted weights before |
|
Concrete tensor-sharding examples and pr |
2026-09-04 |
| pr |
0.01 |
handle transpose during model loading with TP (Qwen3-VL-MoE) |
|
Awkward grammar and precise tensor-shape |
2026-08-27 |
| pr |
0.01 |
Handle empty FSDP shards in sharded loading: crash and hang |
|
Detailed rank-level reproducers and impl |
2026-08-23 |
| pr |
0.01 |
Prefetch only the byte spans each rank will read |
|
Precise performance data and informal te |
2026-09-07 |
| pr |
0.01 |
[CB] Size the KV pool with the per-rank head count under TP |
|
Concise, code-specific explanation with |
2026-09-06 |
| pr |
0.01 |
Skip the expert-parallel sentinel masking when expert parall |
|
Highly specific performance and implemen |
2026-08-21 |
| pr |
0.01 |
:rotating_light: Infer config class default fields without i |
|
Informal wording, typos, and exploratory |
2026-08-25 |
| pr |
0.01 |
fix videomae load error |
|
The terse checkpoint loading report and |
2026-09-10 |
| pr |
0.01 |
Fix slow integration tests on XPU |
|
Highly specific device, kernel, and quan |
2026-09-08 |
| pr |
0.01 |
Add MOSS-Transcribe-Diarize model |
|
No substantive free-text description bey |
2026-08-28 |
| pr |
0.01 |
Add Molmo2 |
|
Description is effectively empty aside f |
2026-01-23 |
| pr |
0.01 |
Reduce eager's computation |
|
Terse wording, typos, and informal phras |
2026-09-08 |
| pr |
0.01 |
Video processors - general maintenance |
|
Very terse informal wording with grammat |
2026-08-24 |
| pr |
0.01 |
🚨🚧 FeatureExtractor → AudioProcessor |
|
Minimal checklist content and automated |
2026-03-02 |
| pr |
0.01 |
🚨 🚨Bring some dinos to modern standards |
|
Informal wording, typo, and personal rev |
2026-05-28 |
| pr |
0.01 |
Fix `StaticCache` for Mllama and enable `torch.compile` |
|
Specific implementation details and an i |
2026-08-20 |
| pr |
0.01 |
fix: use sp_model.decode in SentencePieceBackend.convert_tok |
|
Terse technical explanation with concret |
2026-09-10 |
| pr |
0.01 |
fix: preserve backslashes in generate flag values |
|
Specific bug reproduction cases and impl |
2026-09-10 |
| pr |
0.01 |
fix: check GPU architecture before enabling GQA in SDPA on C |
|
Brief conventional fix description with |
2026-09-09 |
| commit |
0.00 |
[serge] Fix 2 integration tests for model `minimax` failing |
|
— |
2026-09-10 |
| commit |
0.00 |
Update dev (#48654) |
|
Extremely terse, informal commit message |
2026-09-09 |
| commit |
0.00 |
fix incorrect hub tokenizer class (#48641) |
|
Terse commit series with a concrete toke |
2026-09-09 |
| commit |
0.00 |
extend some case to xpu as well (#48502) |
|
Informal wording, signed-off commits, an |
2026-09-07 |
| commit |
0.00 |
[serge] Fix 2 integration tests for model `glm4_moe` failing |
|
— |
2026-09-07 |
| commit |
0.00 |
[serge] Fix 2 integration tests for model `nemotron` failing |
|
— |
2026-09-07 |
| commit |
0.00 |
[serge] Fix 2 integration tests regressed by commit 83d46aa2 |
|
— |
2026-09-07 |
| commit |
0.00 |
[serge] Fix 2 integration tests for model `kosmos2` failing |
|
— |
2026-09-07 |
| commit |
0.00 |
Allow nested rope params for tiny models (#48435) |
|
A lone period provides no evidence of AI |
2026-09-04 |
| commit |
0.00 |
[serge] Fix 2 integration tests for model `fsmt` failing wit |
|
— |
2026-09-03 |
| commit |
0.00 |
[serge] Fix 1 integration tests regressed by commit bd950935 |
|
— |
2026-09-02 |
| pr |
0.00 |
[serge] Fix 3 integration tests for model `pegasus` failing |
|
— |
2026-09-10 |
| pr |
0.00 |
[serge] Fix 2 integration tests for model `minimax` failing |
|
— |
2026-09-03 |
| pr |
0.00 |
Add support for NeuCodec |
|
Brief human-style description with concr |
2026-07-07 |
| pr |
0.00 |
Closed |
|
Single-word closure is plainly human and |
2026-09-10 |
| pr |
0.00 |
[serge] Fix 1 integration test for model `flex_olmo` failing |
|
— |
2026-09-09 |
| pr |
0.00 |
[serge] Fix 4 integration tests regressed by commit 774a675d |
|
— |
2026-09-09 |
| pr |
0.00 |
chore: update self-comment-ci-amd.yml |
|
— |
2026-09-09 |
| pr |
0.00 |
Update dev |
|
Extremely terse and informal commit-styl |
2026-09-09 |
