vllm-project/vllm

Pull Requests Commits

Fix config passed to deepseek_eagle

tlrmchlsmth committed 112 days ago

9ffb182b

[NIXL] Add support for MLA caches with different latent dim (#25902)

NickLucche committed 112 days ago

Verified 80608ba5

[perf] Use CPU tensor to reduce GPU->CPU sync (#25884)

lhtin committed 112 days ago

Verified e184c9c5

[Model] Move `vision_feature_select_strategy` into `resolve_visual_encoder_outputs` (#25938)

DarkLight1337 committed 112 days ago

Verified d7e34b42

[Bugfix][Model]fix ernie45 moe gate&bias dtype to float32 (#25936)

CSWYF3634076 committed 112 days ago

Verified ef6e0e71

Updated TRL integration docs (#25684)

sergiopaniego committed 113 days ago

Verified 1ad3aca6

[Doc] Add Cambricon MLU support (#25942)

a120092009 committed 113 days ago

Verified 8d0afa9b

[New Model] DeepSeek-V3.2 (Rebased to Main) (#25896)

zyongye committed 113 days ago

Verified fa7e254a

[Bugfix]: Clean up chunked prefill logging when using whisper (#25075)

simondanielsson committed 113 days ago

Verified e23cacda

[Model][Bugfix] Fix MiDashengLM audio encoder mask by removing incorrect `logical_not` (#25925)

zhoukezi committed 113 days ago

Verified 2e1b8bc2

[BugFix] Pass config_format via try_get_generation_config (#25912)

acisseJZhong committed 113 days ago

Verified e47433b3

[BugFix] Fix DP/EP hang (#25906)

LucasWilkinson committed 113 days ago

Verified 23194d83

Move`VllmConfig` from `config/__init__.py` to `config/vllm.py` (#25271)

hmellor committed 113 days ago

Verified 61aedb5f

[Benchmark] Support benchmark throughput for external launcher DP (#25913)

zhuohan123 committed 113 days ago

Verified d3bd1711

[Bug] Fix Weight Loading for Block FP8 Cutlass SM90 (#25909)

yewentao256 committed 113 days ago

Verified 89e4050a

Test Prompt Embeds/LoRA compatibility and Enable LoRA Support for OPT Models (#25717)

qthequartermasterman committed 113 days ago

Verified 78a47f87

[V0 Deprecation] Remove `vllm.worker` and update according imports (#25901)

aarnphm committed 113 days ago

Verified 6a113d9a

[NIXL] Increase default KV block eviction timeout on P (#25897)

NickLucche committed 113 days ago

Verified 2e4fe48c

[Doc] Polish example for torchrun dp (#25899)

zhuohan123 committed 113 days ago

Verified 8eb0a1d9

[Kernel] Chunk-aligned mamba2 (#24683)

tdoublep committed 113 days ago

Verified fea3e476

[Bugfix][ROCm] Fixing trying to import non-existent symbols from libnccl.so (#25605)

gshtras committed 113 days ago

Verified 61a34316

[Doc] Add documentation for vLLM continuous benchmarking and profiling (#25819)

namanlalitnyu committed 113 days ago

Verified 9bedac96

[BugFix][torch.compile] KV scale calculation issues with FP8 quantization (#25513)

adabeyta committed 113 days ago

Verified c42ff4f4

[Bugfix] Use correct key "ignore" for config.json non-quantized layers (#25706)

leejnau committed 113 days ago

Verified d5ab2851

[Model] Remove MotifForCausalLM (#25866)

jeejeelee committed 113 days ago

Verified e61eb5e0

[CI/Build] Include Transformers backend test in nightly transformers test (#25885)

Isotr0py committed 113 days ago

Verified 0899ba5b

[Bugfix][Speculative Decoding] Fix Eagle3 quantization config issue (#25883)

rahul-tuli committed 113 days ago

Verified 145ac733

[Nixl][P/D] Add cuda2cpu support (HD->DH transfer) (#24690)

chenxi-yang committed 113 days ago

Verified d0d138bc

[torch.compile] serialize cudagraph_mode as its enum name instead of value (#25868)

ZJY0516 committed 113 days ago

Verified 43227236

[Model][Bugfix] Fix issues in MiDashengLM implementation for quantized models (#25854)

zhoukezi committed 113 days ago

Verified 8616300a

Older