vllm-project/vllm

Pull Requests Commits

pre commit, typo fix

tlrmchlsmth committed 3 days ago

607615b7

Upgrade Numba to fix segfault on CUDA drivers version 580

tlrmchlsmth committed 3 days ago

5daeb8f6

[Bugfix] Disable CG for Whisper+FA2 (#33164)

NickLucche committed 3 days ago

Verified 1f3a2c29

[Metrics] [KVConnector] Add Offloading Connector metrics (#27942)

omerpaz95 committed 3 days ago

Verified 7227d061

Fix weight mapping test for Transfomers v5 (#33162)

hmellor committed 3 days ago

Verified 14385c80

[Frontend] Frontend will only attach supported tasks corresponding entrypoints. (#33139)

noooop committed 3 days ago

Verified 76139d08

[AMD][QWEN3-NEXT] FP8 Tunings (#32042)

draftbk committed 3 days ago

Verified da8d0c44

[AMD][Kernel][BugFix] Use correct scale in concat_and_cache_ds_mla_kernel when on gfx942 (#32976)

rasmith committed 3 days ago

Verified 58996f35

[Models] Kimi-K2.5 (#33131)

ywang96 committed 3 days ago

Verified b539f988

[CI][Pooling] Stabilize ModernBERT test (#32909)

AndreasKaratzas committed 3 days ago

Verified 6c006457

[code clean] remove duplicate code (#33135)

andyxning committed 3 days ago

Verified b781eeaa

[Frontend] Cleanup serving engine (#33103)

DarkLight1337 committed 3 days ago

Verified e0b005d9

[torch.compile] Stop assuming 32 bit indexing (#33113)

zou3519 committed 3 days ago

Verified 3b8f0fe5

[Frontend] Reduce mixin usage in serving pooling (#33101)

DarkLight1337 committed 3 days ago

Verified c831911b

[Perf] avoid duplicate mem_get_info() call in get_current_memory_usage (#33064)

pacoxu committed 3 days ago

Verified 157caf51

[DOC]: Add warning about max_num_batched_tokens and max_model_len when chunked prefill is disabled (#33109)

VincentG1234 committed 3 days ago

Verified 0b53bec6

Fix IndexError with encoder-decoder models when using Custom Paged Attention (#33112)

sstamenk committed 3 days ago

Verified c568581f

fix: preserve native tool call ID in multi-turn tool calling (#32768)

wangln19 committed 3 days ago

Verified 2d705343

[MoE Refactor] Integrate Naive Prepare Finalize into MK (#32567)

robertgshaw2-redhat committed 3 days ago

Verified 5a93b916

[Model Runner V2] Remove UvaBufferPool for cpu->gpu copy (#33055)

WoosukKwon committed 3 days ago

Verified 6d86fde0

[Bugfix][TPU] Return a Default fp8 MoE Backend (#32908)

vanbasten23 committed 3 days ago

Verified 510ed1e8

[Bugfix][MXFP4] Call `trtllm_fp4_block_scale_moe` with kwargs (#33104)

wpc committed 3 days ago

Verified 8caffd92

[fix] CPUDNNLGEMMHandler pointer baked into inductor artifact (#32913)

dolpm committed 4 days ago

Verified 58a05b0c

[Logging] add `--disable-access-log-for-endpoints` CLI option (#30011)

JaredforReal committed 4 days ago

Verified 6ee7f18f

[Refactor] Remove unused `_moe_permute` function (#33108)

yewentao256 committed 4 days ago

Verified 8f987883

[ci] Sync test areas with test-pipeline.yaml and enable new pipeline generator (#33080)

khluu committed 4 days ago

Verified ebe0ba91

[Bugfix] Fix Dtypes for Pynccl Wrapper (#33030)

robertgshaw2-redhat committed 4 days ago

Verified 43a013c3

[Model] Bump transformers version for test registry (#33100)

DarkLight1337 committed 4 days ago

Verified c25dbee4

[Bugfix] Fix Voxtral streaming slot_mapping (#33073)

NickLucche committed 4 days ago

Verified 19ab0f7c

[FIX] Always support TP > 4 for FP4 Gemm (#31099)

danielafrimi committed 4 days ago

Verified 67fe677c

Older