COMMITS
October 25, 2025
B
qwen3moe on gh200
bhaktatejas922 committed
M
[Attention] Add missing kv cache scale setup (#27490)
Matthew Bonanni committed
C
[Misc] Simplify max tokens in multimodal registry (#27500)
Cyrus Leung committed
K
Z
Revert "[Misc] Remove use of CUDA_VISIBLE_DEVICES for device selectio… (#27502)
Zhuohan Li committed
J
[CI] Add tests for cudagraph (#27391)
Jiangyun Zhu committed
Y
[KVConnector] Migrate the LMCache integration code to be vLLM native (#25542)
Yihua Cheng committed
October 24, 2025
V
[Misc][DP] Guard mxfp4 implementation selection (#27484)
Varun Sundar Rabindranath committed
W
[Log] Optimize Startup Log (#26740)
Wentao Ye committed
P
[Distributed] Basic set of configuration for large EP deployment on GB200 (#27328)
Pengchao Wang committed
L
[Perf][Async Scheduling] Remove CPU->GPU sync in dummy_run (#27455)
Lehua Ding committed
J
[Document] Add ms-swift library to rlhf.md (#27469)
jinghanhu committed
Z
[CI/Build] Fix test_torch_utils in AMD CI (#27317)
Zhewen Li committed
I
[Bugfix] Fix interns1-vit qk norm code path (#27480)
Isotr0py committed
M
[Attention] Add MLA prefill backend: trtllm_ragged_attention_deepseek (#26397)
Ming Yang committed
K
[Bugfix] Fix MultiConnector stats reconstruction across process boundaries (#27366)
kourosh hakhamaneshi committed
C
[NIXL][BUGFIX] delay done_recving queue cleanup to bottom of get_finished (#27297)
Chendi.Xue committed
R
[compile] Turn standalone_compile back on (#27460)
Richard Zou committed
F
L
[Doc] Fix minor issues in docs/design/metrics.md (#27436)
Lifans committed
C
Fix test named tool use (#27458)
Chauncey committed
F
[MISC] `cudagraph_capture_sizes` related improvements (#26016)
fhl2000 committed
I
Fix AArch64 CPU Docker pipeline (#27331)
ioana ghiban committed
C
[Benchmark] Enable benchmark to run with `encoding_format="bytes"` (#27467)
Cyrus Leung committed
C
2
[BugFix] Fix torchrun DP with LLM class (#27395)
22quinn committed
I
[MM][Bugfix] Replace `PatchEmbed`'s conv3d to linear layer (#27418)
Isotr0py committed
Y
[Docs] remove v1 column for embedding models (#27446)
Yu Jiaqi committed
R