vllm-project/vllm

Pull Requests Commits

khluu committed 59 days ago

6d4a8e6d

[Tool] `adjust_request` to reasoning parser, and Gemma4 fixes (#39027)

bbrowning committed 59 days ago

0d784f46

[Bugfix] Fix Gemma4 streaming tool call corruption for split boolean/number values (#39114)

sfeng33 committed 59 days ago

a7b39273

[Bugfix][Frontend] Fix Gemma4 streaming HTML duplication after tool calls (#38909)

yoke233 committed 59 days ago

bae948a9

[Gemma4] Enable Fast Prefill Optimization (#38879)

LucasWilkinson committed 59 days ago

fc29ef13

[Bugfix] Fix invalid JSON in Gemma 4 streaming tool calls by stripping partial delimiters (#38992)

Gregory-Pereira committed 59 days ago

10a26d1d

[Bugfix]: Fix Gemma4ToolParser.__init__() missing `tools` parameter (#38847)

hospedales committed 66 days ago

2a69949b

feat(models): implement Google Gemma 4 architecture support (MoE, Multimodal, Reasoning, Tool-Use) (#38826)

lucianommartins committed 67 days ago

8adcf8c4

Revert "[Bugfix] Restrict TRTLLM attention to SM100, fixing GB300 (SM103) hang (#38730)"

khluu committed 67 days ago

cfad6a50

[Bugfix] Restrict TRTLLM attention to SM100, fixing GB300 (SM103) hang (#38730)

stecasta committed 68 days ago

c284a667

[Misc] Rename think_start_str/think_end_str to reasoning_start_str/reasoning_end_str (#38242)

chaunceyjiang committed 68 days ago

3a30a1a6

(security) Enforce frame limit in VideoMediaIO (#38636)

jperezdealgaba committed 68 days ago

29982d48

[Feat][v1] Simple yet General CPU KV Cache Offloading (#37160)

ivanium committed 68 days ago

1dbbafd3

[Bugfix][MLA] Add logits size budget to sparse indexer prefill chunking (#36178)

LucasWilkinson committed 68 days ago

0ee3b7fc

[Bugfix][Async] Fix async spec decoding with hybrid models (#38556)

MatthewBonanni committed 68 days ago

268bed9c

[CI] fix LM Eval Qwen3.5 Models (B200) (#38632)

ZJY0516 committed 68 days ago

bcc0fdd0

[CI Failure] pin colmodernvbert revision (#38612)

noooop committed 68 days ago

69b8bd4b

[Bugfix][CPU] Skip set_num_threads after thread binding (#38535)

bigPYJ1151 committed 69 days ago

12449f94

[CI] Fix SPLADE pooler test broken by #38139 (#38495)

haosdent committed 69 days ago

b92312df

[MoE] Add RoutingMethodType.Simulated to TRT-LLM FP8/NVFP4 kernel allowlists (#38329)

jaewonlee-fb committed 70 days ago

Verified d816834c

[Misc] Always use `forward_mulmat` for `Conv3d` on newer versions of torch. (#38487)

ywang96 committed 70 days ago

Verified 92f0db57

[CI] Add temperature=0.0, reduce max_tokens, and add debug prints to audio_in_video tests (#38492)

AndreasKaratzas committed 70 days ago

Verified bea23536

Add @ZJY0516 to CODEOWNERS (#38497)

ZJY0516 committed 70 days ago

Verified c133f337

[Bugfix] Support multi-type params parsing for DeepSeek v3.2 (#33703)

kizill committed 70 days ago

Verified a6db99ba

[ROCm][CI] Enable hybrid chunked prefill test (#38317)

AndreasKaratzas committed 70 days ago

Verified 4f2ed5fd

[QeRL] Fix online quantized reloading (#38442)

kylesayrs committed 71 days ago

Verified d28d86e8

[Perf] Remove redundant device copies for CPU-only pooling token IDs, 48.9% E2E throughput improvement (#38139)

yewentao256 committed 71 days ago

Verified 995dea13

[Transformers v5] fix missing pixtral/voxtral multimodal dispatch (#38410)

allgather committed 71 days ago

Verified 8c0b6267

[ROCm][CI] Fix cross-attention dispatch for encoder-decoder models (#38450)

AndreasKaratzas committed 71 days ago

Verified 43cc5138

[Spec Decode, BugFix] Propagate norm_before_fc from Eagle3 speculator (#38111)

shubhra committed 71 days ago

Verified 5b8c30d6

Older