vllm-project/vllm

Pull Requests Commits

Make KV-offload eval CI steps non-optional

tlrmchlsmth committed 10 hours ago

Verified bfe9fe85

Fix ruff format: collapse assert expression to single line

tlrmchlsmth committed 1 day ago

Verified a9a689c8

Run GSM8K eval twice with prefix caching disabled

tlrmchlsmth committed 1 day ago

Verified 64c9d191

Add SimpleCPUOffloadConnector variant for DeepSeek-V4-Flash

tlrmchlsmth committed 3 days ago

Verified e79afe8f

Add Qwen 3.5 coverage and fix CI labels

tlrmchlsmth committed 3 days ago

Verified 32a49608

Abstract connector type; test both OffloadingConnector and SimpleCPUOffloadConnector

tlrmchlsmth committed 6 days ago

Verified 872edfc4

Classify offloading CI steps by size, not architecture

tlrmchlsmth committed 6 days ago

Verified 09168959

Use [tag] labels and short keys for offloading CI steps

tlrmchlsmth committed 6 days ago

Verified b10906e9

Apply ruff formatting

tlrmchlsmth committed 6 days ago

Verified e8a20abe

Add Buildkite CI steps for KV offloading eval

tlrmchlsmth committed 6 days ago

Verified 3512a46c

Add Gemma-4-E4B-it and DeepSeek-V4-Flash to offloading eval

tlrmchlsmth committed 6 days ago

Verified 3a7b454b

[Eval] Add GSM8K correctness test for SSM models with KV offloading

tlrmchlsmth committed 6 days ago

Verified 2cfdd033

[ROCm][CI] Fine-tuning queues and test names (#39238)

AndreasKaratzas committed 7 days ago

Verified 2a6f8f0c

Fix P/D with DP Supervisor (#46628)

robertgshaw2-redhat committed 7 days ago

Verified c5e3c408

[Perf] Remove redundant clone for GLM, Deepseek etc (#46651)

yewentao256 committed 7 days ago

Verified 8b4d93ba

[Kernel][MoE] Tune block-FP8 fused MoE for low-batch decode (#46642)

mgoin committed 7 days ago

Verified e8e7b592

[ROCm]: Bump aiter to 0.1.16.post2 (#46692)

Rohan138 committed 7 days ago

Verified e53a1723

[CI] Re-enable skipped glm and seedoss parser tests (#46671)

sfeng33 committed 7 days ago

Verified 96eb8ddc

[Bugfix] FLASHINFER_MLA_SPARSE_SM120 compatibility with GLM-5 NVFP4 (#46506)

lucifer1004 committed 7 days ago

Verified 8fa36fbb

[Bugfix] Fix NVFP4+MTP crash: force unquantized mtp.fc for Qwen3Next (#46316)

WindChimeRan committed 7 days ago

Verified e45b2799

[Core] Avoid mixed length specdec batches via padding (#45237)

qianlihuang committed 7 days ago

Verified d490b981

[ROCM] [Communication] Add INT3 quantization method for quickreduce (#45666)

haoyangli0109 committed 7 days ago

Verified 1744adc2

[ROCm][CI] rm duplicate Distributed Torchrun ci test (#46729)

divakar-amd committed 7 days ago

Verified cdfa2fd7

[Pooling] Fix Cohere embed billed image token accounting for mixed-content inputs (#46093)

taneem-ibrahim committed 7 days ago

Verified 6f3da461

[CI] Pin GitHub Actions to commit hashes in macos-smoke-test.yml (#38290)

russellb committed 7 days ago

Verified d3130d87

[MoE] [MoE Refactor] Add moe kernel oracle abc 37753 (#43461)

qyYue1389 committed 7 days ago

Verified 9bfd878a

[Hardware][AMD][CI] Mirror Basic Models (Others) and Weight Loading Multiple GPU test groups (#46668)

mawong-amd committed 7 days ago

Verified 2365b7a8

[NIXL][Mamba] Add Mamba1 support to NIXL P/D disaggregation (#45019)

Josephasafg committed 7 days ago

Verified 15be7873

[CPU][CI/Build] Allow more CPU CI agents (#46702)

bigPYJ1151 committed 7 days ago

Verified 92221485

[XPU][CI]Refine .buildkite/ci_config_intel.yaml for Intel GPU CI (#46674)

zxd1997066 committed 7 days ago

Verified a6f41ab6

Older