3 people cc4210f671 [Perf] Improve BF16x3 router GEMM accuracy and make it default on sm100 (#55899) ...

Signed-off-by: Thien Tran <gau.nernst@yahoo.com.sg>
Co-authored-by: OpenAI Codex <noreply@openai.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>

2026-09-09 07:33:12 -07:00

..

basic_correctness

[CI][Test] Deflake test_mem.py sleep-mode asserts via allocator bookeeping (#54312)

2026-08-29 12:05:17 -07:00

benchmarks

[Bugfix] Restore multimodal support on the plain "vllm" throughput backend (#52168)

2026-08-28 10:28:25 +00:00

compile

[Kernel] Enable optimized FlashInfer add-RMSNorm NVFP4 fusion (#51925)

2026-09-08 17:20:32 -07:00

config

[New model][Multimodal] Add DeepSeek-V4-Flash-Vision-Exp support (#54566)

2026-09-02 11:09:09 -07:00

cuda

[Bugfix][Platform] Stop re-initializing NVML on every device-capability check (fixes #50381) (#50393)

2026-08-07 08:30:00 +00:00

detokenizer

[Bugfix][Spec Decode] Select earliest-completing stop string in check_stop_strings (#49391)

2026-07-23 18:14:06 +08:00

distributed

[Core][KV Events] Echo session_id on GPU BlockStored events (#51381)

2026-09-04 02:22:50 +00:00

engine

[Qwen4Exp] Support UVA PLE-offload and Engram tensor parallelism (#54371)

2026-09-09 07:32:34 -07:00

entrypoints

[Frontend] Migrate input-validation errors to VLLMValidationError (harmony/mistral/chat_utils/params) (#55735)

2026-09-08 06:45:40 +00:00

evals

[CI][ROCm] Restore Wikitext coverage for Qwen OCP-MX (#55410)

2026-09-05 10:50:54 +08:00

fusion

[Fusion] Manual ActivationQuantFusionPass initial application (#51415)

2026-09-03 13:37:31 -04:00

ir

[CI][IR] Speed up vLLM IR test group (#55965)

2026-09-08 21:28:06 -05:00

jit_monitor

[ROCm] Defer tilelang import through its import from vllm.tilelang_utils import tilelang and relaxed has_tilelang (#51159)

2026-08-13 13:12:40 -05:00

kernels

[Perf] Improve BF16x3 router GEMM accuracy and make it default on sm100 (#55899)

2026-09-09 07:33:12 -07:00

lora

[Quant][Kernel] Remove GPTQ Group/Dynamic Activation Ordering (#54809)

2026-09-08 12:45:12 -07:00

model_executor

[CI][Test][Spec Decode] Fix CI failure of Qwen3 Omni DSpark loader mock (#56011)

2026-09-08 23:06:15 -07:00

models

[Qwen4Exp] Support UVA PLE-offload and Engram tensor parallelism (#54371)

2026-09-09 07:32:34 -07:00

multimodal

[Bugfix][Audio] Restore soundfile-first automatic decoding (#55642)

2026-09-07 06:21:02 +00:00

parser

[Bugfix] Parse DSML tool calls when the model omits the tool_calls wrapper (#55954)

2026-09-08 17:38:04 -07:00

platforms

[CPU] Support FP16/BF16 persisted GDN state on AMX (#52191)

2026-08-31 16:32:20 +08:00

plugins

[Frontend] Move engine/protocol.py out openai folder (#54492)

2026-08-31 10:13:50 +00:00

plugins_tests

[Frontend] Move cli_args.py and dp_supervisor.py out openai folder (#53659)

2026-08-25 07:02:32 +00:00

prompts

…

quantization

Fix block FP8 MTP in ModelOpt mixed checkpoints (#55513)

2026-09-08 22:47:21 +00:00

reasoning

feat: Add support for reasoning_token_count to reasoning parser (#54982)

2026-09-03 05:02:49 -07:00

renderers

[Bugfix] Gracefully handle unsupported reasoning_effort in chat templates (#54022)

2026-09-06 21:11:35 -07:00

rocm

[Bugfix][MoE] Preserve unquantized weight storage on ROCm (#46009)

2026-09-03 11:06:09 +08:00

samplers

[ROCm] Remove stale SDPA and skinny GEMM workarounds (#50907)

2026-08-11 14:19:16 -04:00

spec_decode

[MyPy][1/N] Fix mypy errors in some tests/ directories and enforce follow-imports=silent (#49570)

2026-07-30 05:02:51 -07:00

standalone_tests

[ROCm][CI] Use the same-build wheel in Python-only CI (#49514)

2026-08-16 22:52:24 -05:00

system_messages

…

tokenizers_

[Frontend] Migrate input-validation errors to VLLMValidationError (harmony/mistral/chat_utils/params) (#55735)

2026-09-08 06:45:40 +00:00

tool_parsers

[Bugfix] Parse DSML tool calls when the model omits the tool_calls wrapper (#55954)

2026-09-08 17:38:04 -07:00

tool_use

[Frontend] Move engine/protocol.py out openai folder (#54492)

2026-08-31 10:13:50 +00:00

tools

[CI][ROCm] Avoid redundant image pulls during smoke validation (#54408)

2026-08-31 21:48:25 -05:00

tracing

…

transformers_utils

[Bugfix] Handle null RoPE parameters for NoPE layers (#55949)

2026-09-08 20:20:05 +00:00

utils_

[Core] Enhance cpu<->gpu sync checking to include paged async copies (#53491)

2026-09-08 17:26:21 -07:00

v1

[Bugfix] Fix Mooncake heterogeneous TP with replicated GQA heads (#52516)

2026-09-09 14:18:32 +08:00

vllm_test_utils

…

weight_loading

[Quant][Kernel] Remove GPTQ Group/Dynamic Activation Ordering (#54809)

2026-09-08 12:45:12 -07:00

__init__.py

…

ci_envs.py

…

conftest.py

[Bugfix][CPU] Fix several bugs (#54042)

2026-08-31 15:33:06 +08:00

test_access_log_filter.py

…

test_attention_backend_registry.py

[Attention] Require explicit DCP support from attention implementations (#55780)

2026-09-08 14:04:27 -07:00

test_audio_media_size_precheck.py

fix(security): enforce VLLM_MAX_AUDIO_CLIP_FILESIZE_MB on all audio paths (#53561)

2026-08-24 21:37:02 +00:00

test_cmake_utils.py

[CI/Build] Add warning for unsupported global PTX architecture requests in... (#51901)

2026-08-15 07:01:52 -07:00

test_config.py

[Qwen4Exp] Support UVA PLE-offload and Engram tensor parallelism (#54371)

2026-09-09 07:32:34 -07:00

test_connections.py

Reject oversized media before fully downloading it (#51896)

2026-08-24 01:15:09 +00:00

test_embedded_commit.py

…

test_envs.py

[Frontend] Gate scale-out endpoints behind opt-in flag (#54579)

2026-09-02 14:31:14 +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

[Frontend] Consolidate online serving utils. (#44479)

2026-06-04 06:49:31 +00:00

test_logprobs.py

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

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

test_outputs.py

…

test_pooling_params.py

[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665)

2026-07-29 04:20:40 +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

fix nixl side-channel host selection (#41806)

2026-05-11 07:40:37 +00:00

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

[Bugfix] Reject empty bad-word tokenizations (#53433)

2026-08-31 12:07:17 -07:00

test_sampling_params.py

[Bugfix] Validate extension integers before engine serialization (#55606)

2026-09-08 16:30:06 +08:00

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

Exclude the cpu backend from vLLM's active-Triton-driver count (#53530) (#53530)

2026-08-25 04:13:13 +00:00

test_version.py

…

test_vllm_port.py

…

test_zen_cpu_platform_detection.py

[Bugfix] Log platform plugin detection failures (#52285)

2026-09-01 19:11:52 +00:00

utils.py

[CI] Avoid logging test server environment values (#54379)

2026-09-03 17:00:29 -04:00