vllm/tests at worktree-cpu_gdn_spec_conv_fix - vllm - Gitea: Git with a cup of tea

mirrors/vllm

0 Fork 0

mirror of https://github.com/vllm-project/vllm.git synced 2026-10-28 09:45:26 +08:00

Files

T

History

jiang1.li 82135ca498 [Bugfix][CPU] Fix GDN spec conv-state handling for short query lengths

A decode row without drafts stays a spec row (q_i=1) and keeps the
num_accepted offset left by its previous step. The fallback conv path
rolled the buffer assuming q_i == num_spec + 1, so B.copy_ hit a size
mismatch (6 vs 5); the native kernel makes the same assumption and
rejects num_accepted > seqlen. Roll as history+new tokens shifted left by
one (as the Triton kernel does) and only take the native path for
full-length rows. Exposed by the MRV2 warmup (#62888).

2026-10-23 13:41:31 +00:00

..

basic_correctness

[Core][Sleep Mode] Use torch.accelerator in CuMemAllocator free/sleep (#62701 )

2026-10-22 11:56:18 +08:00

[Bugfix][Benchmark] Support boolean sweep plot filters (#60774 )

2026-10-22 22:21:12 +08:00

[CI][Test][1/N] Enable unit tests to run on Rubin (#60298 )

2026-10-22 15:25:45 +00:00

[Bugfix] Don't enable MM embedding inputs on KV consumers in language-model-only mode (#62880 )

2026-10-22 22:18:09 -07:00

Add pydocstyle to the ruff rules (#52136 )

2026-09-16 16:44:29 +00:00

[Bugfix][Spec Decode] Trim token_ids/logprobs left past a stop string under speculative decoding (#47616 )

2026-10-09 14:22:31 -07:00

device_allocator

[ROCm] Capture CUDA graphs with expandable segments off for AITER all-reduce (#57481 )

2026-10-22 09:34:55 +08:00

[Bugfix] shm_broadcast: poll the notify socket from busy readers (#63165 )

2026-10-23 04:58:34 +00:00

[Frontend] Expose --sleep-mode-backend on the CLI (#63138 )

2026-10-22 23:27:19 +00:00

[Frontend] Accept images on /v1/systemone for multimodal next-token models (#62181 )

2026-10-23 07:30:02 +00:00

[Docs] Point users at --disable-flashinfer-autotune (#61489 )

2026-10-21 10:55:26 -04:00

[Fusion] Support 3D activation quantization fusion and enable it for Pixtral (#59612 )

2026-10-09 15:12:03 +08:00

Add pydocstyle to the ruff rules (#52136 )

2026-09-16 16:44:29 +00:00

[Perf] JIT monitor: hook TileLang JITImpl.compile instead of __call__ (#58685 )

2026-10-08 00:11:59 +00:00

[Bugfix][CPU] Fix GDN spec conv-state handling for short query lengths

2026-10-23 13:41:31 +00:00

[Bugfix][LoRA] Skip no-LoRA tokens (index -1) in CPU torch bgmv ops (#48421 )

2026-10-22 10:59:53 +08:00

[Fast Start] Fix weight cache daemon GPU memory and warm-cache warmup hang (#62583 )

2026-10-23 04:51:57 +00:00

[ROCm][Kimi-K3] Support MoRIIO consumers with specialized decode (#63031 )

2026-10-23 18:35:14 +08:00

[Bugfix][Multimodal] Return 400 for structurally-corrupted PNG images (#62704 )

2026-10-22 06:57:08 +00:00

[Model][ROCm] Add Muse Spark Open support on ROCm (#62622 )

2026-10-22 17:00:09 +08:00

[Attention] Remove cascade attention (#61126 )

2026-10-12 10:41:07 -07:00

[Bugfix][MRV2] Keep input_ids on later PP stages for raw-token models (#61016 )

2026-10-15 12:54:21 -07:00

[Feature][Pooling] Serve late chunking through /v1/embeddings (#61772 )

2026-10-22 04:00:51 +00:00

…

[Quantization][Quark] Support MXFP8 (W8A8) routed experts (#63125 )

2026-10-22 13:48:48 -04:00

[Bugfix][Frontend][Kimi K3] Fix Kimi K3 XTML marker leakage in responses API (#62507 )

2026-10-22 12:45:26 -07:00

[Frontend] Use VLLMValidationError for client input errors in renderers base (#62738 )

2026-10-22 18:14:12 +00:00

[ROCm] Add _maybe_pad_weight guard-condition and RL-reload stability tests (#62977 )

2026-10-22 01:20:28 +00:00

[Bugfix][Frontend] Normalize beam search scores by generated length (#62928 )

2026-10-22 07:26:05 +00:00

Add pydocstyle to the ruff rules (#52136 )

2026-09-16 16:44:29 +00:00

standalone_tests

[CI] Add nightly Python 3.14 smoke test (#60456 )

2026-10-12 10:11:26 -07:00

[Model][ROCm] Add Muse Spark Open support on ROCm (#62622 )

2026-10-22 17:00:09 +08:00

[Model][ROCm] Add Muse Spark Open support on ROCm (#62622 )

2026-10-22 17:00:09 +08:00

[Bugfix][Frontend][Kimi K3] Fix Kimi K3 XTML marker leakage in responses API (#62507 )

2026-10-22 12:45:26 -07:00

[Security][CI/Build] Add pre-commit check for hardcoded trust_remote_code=True (#62547 )

2026-10-19 21:19:35 +00:00

[Refactor] Remove dead tests code (#58916 )

2026-09-30 04:39:03 +00:00

transformers_utils

[Model] Add Recursive Flint (#62938 )

2026-10-22 14:56:35 -07:00

[Bugfix][CPU] Fix DP NUMA affinity regression (#62867 )

2026-10-23 13:18:13 +08:00

[Bugfix][CPU][CI] Accept assigned nodes in the no-Triton worker test (#63299 )

2026-10-23 17:42:35 +08:00

vllm_test_utils

Add pydocstyle to the ruff rules (#52136 )

2026-09-16 16:44:29 +00:00

[XPU] Enable kernel/sampler unit tests on Intel XPU (#62419 )

2026-10-23 16:23:12 +08:00

__init__.py

…

AGENTS.md

[Doc] Raise the bar for agent-written tests (#62601 )

2026-10-21 10:02:57 +00:00

ci_envs.py

Add pydocstyle to the ruff rules (#52136 )

2026-09-16 16:44:29 +00:00

conftest.py

[CI] Keep vLLM log propagation on through the first LLM init in caplog_vllm (#61535 )

2026-10-14 15:10:35 +08:00

test_attention_backend_registry.py

[Platform] Allow BackendEnum dynamic register (#41091 )

2026-10-08 09:56:19 +08:00

test_audio_media_size_precheck.py

[Security] Bound decoded and downloaded media per request (#61226 )

2026-10-13 10:17:14 +00:00

test_cmake_utils.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212 )

2026-09-16 10:57:24 -07:00

test_config.py

[Bugfix] Defer MRV2 Triton validation to worker startup (#62576 )

2026-10-22 17:51:11 -07:00

test_connections.py

[Multimodal][Docs] State that retried media downloads count toward the request budget (#62204 )

2026-10-19 02:36:29 +00:00

test_embedded_commit.py

…

test_envs.py

[Bugfix][Rust Frontend] Fix HTTP 500 on long conversations: fuel-limit only request templates (#62374 )

2026-10-21 16:55:58 +08:00

test_force_first_config.py

[Perf] Add VLLM_TRITON_FORCE_FIRST_CONFIG to skip Triton autotuning (#42425 )

2026-06-16 15:16:45 +02:00

test_fxgraphcache_pickle_patch.py

[ZenCPU] Make PT Backport Patch Accessible to vLLM (#38205 )

2026-04-10 08:29:35 +00:00

test_logger.py

[Core][Logging] Add built-in JSON formatter (#58739 )

2026-10-01 09:11:46 +01:00

test_logprobs.py

[Bugfix] Handle empty FlatLogprobs slices and delta output (#53704 )

2026-08-26 21:00:54 +08:00

test_outputs.py

[Bugfix][Core] Preserve speculative metrics when coalescing outputs (#55708 )

2026-10-07 08:35:14 +08:00

test_pcp_dp.py

[PCP] Support prefill context parallelism with data parallelism (#57075 )

2026-09-23 22:32:36 -04:00

test_pooling_params.py

[Bugfix][Pooling] Validate late-interaction mode and use count in the Python API (#63112 )

2026-10-23 05:45:15 +00:00

test_quantization_revision_pin.py

Preserve revision pins in secondary artifact loaders (#51446 )

2026-08-11 04:33:10 +00:00

test_ray_env_utils.py

[Bugfix] Set breakable graph env before Ray actor import (#53293 )

2026-08-28 23:11:50 +00:00

test_ray_env.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212 )

2026-09-16 10:57:24 -07:00

test_regression.py

[CI/Build] Use modelscope's international site for regression test (#42176 )

2026-05-09 19:47:09 -07:00

test_request_input_bounds.py

[Model][ROCm] Add Muse Spark Open support on ROCm (#62622 )

2026-10-22 17:00:09 +08:00

test_sampling_params.py

[watermark] compatibility validation (#56801 )

2026-10-05 14:37:13 -04:00

test_scalartype.py

…

test_sequence.py

…

test_tensorizer_revision_pin.py

Preserve revision pins in secondary artifact loaders (#51446 )

2026-08-11 04:33:10 +00:00

test_triton_utils.py

[CI] Allowlist-shrink batch 1: wire 17 root-level tests + drop 5 stale watermarking entries into misc.yaml (#58055 )

2026-09-27 22:23:46 -07:00

test_uvicorn_logging.py

[Core][Logging] Add built-in JSON formatter (#58739 )

2026-10-01 09:11:46 +01:00

test_version.py

[CI] Allowlist-shrink batch 1: wire 17 root-level tests + drop 5 stale watermarking entries into misc.yaml (#58055 )

2026-09-27 22:23:46 -07:00

test_vllm_port.py

…

test_zen_cpu_platform_detection.py

[CPU][Zen] Update zentorch to 2.14.0.1 & enable float16 on Venice and later (#61411 )

2026-10-15 00:23:50 -07:00

test_zentorch_version_pin.py

[ZenCPU][Test] Add zentorch/torch version pin consistency test (#52888 )

2026-10-06 13:12:19 +08:00

utils.py

[MRV2] Switch CI tests still running on MRV1 unnecessarily (#61014 )

2026-10-11 09:57:07 -07:00