12月1月2月3月4月5月6月7月8月9月10月11月
周一周三周五
一年内 259282 次贡献
更少更多
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex/qwen3-tts-default-chunk-ramp 分支2c157fb670 [Perf][Qwen3-TTS] Enable chunk ramp by default
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 feat/qwen3-tts-deadline-scheduler 分支90c890ad6a [Perf][TTS] Add deadline-aware audio scheduling
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex/dlo-hwr-registered-h2d 分支f3255b8daf Merge branch 'main' into codex/dlo-hwr-registered-h2d
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex-flux2-klein-bf16-hwr 分支44f7a6b7f7 [Diffusion] Add FLUX.2-klein BF16 HWR contract
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex-fix-omni-sleep-wake-resume 分支c6286c9acf [CI] Resume AR admission after sleep-mode wake
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex-fix-env-var-ratchet-6217 分支9a7ea9b003 [Bugfix] Reconcile env-var inventory with post-#6217 main drift
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex-dlo-cache-budget-config 分支比较 8 提交 »e1c4b6153a [Model] Make the dummy-run recipe feature-neutral33297bdb9d [Model] Fold the DLO warmup recipe into _make_dummy_requestf4394f1d3d [Model] Prime DLO cache retention with a startup warmupc9bb65902e [Model] Learn the retention budget directly from observed peaksb78db6036a [Model] Fix adaptation test fixture so the capacity cap does not bind
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 minicpm-challenge 分支ecd9d99da0 [CI][MiniCPM-o] Add NPU Video-MME accuracy nightly as a standalone job (#6596)
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 main 分支比较 44 提交 »5ec576ab4c [Bugfix] Fix missing Realtime audio completion (#6564)030fccb2bf [Test] Add MiniMax-H3 DLO DP2 T2VA smoke (#6555)816335cb46 [Doc] Add MiMo-Audio recipe for RTX 5090/5090D 32GB (#6559)59cf3add40 [Feature] Support MiniMax-H3 Turbo LoRA with DLO (#6550)7439ed41c2 Revert "[Bugfix] Restore name-based model detection for HF cache snapshot paths" (#6642)
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex/dlo-hwr-loader-integration 分支比较 29 提交 »e009c2cd06 [Refactor] Restrict PR2 to HWR DLO integrationd7e2833ea7 Merge remote-tracking branch 'origin/main' into codex/dlo-hwr-loader-integrationc8e2b98fb6 [Tests] Consolidate Host Weight Runtime coverage9974c0b26e [Refactor] Rename Host Weight Runtime loader module678aa25864 Revert "Revert "[CI]Fix Wan DMD pipeline test alignment"" (#6578)
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 dev/vllm-align 分支比较 122 提交 »0d582c2da0 tests: restore Qwen-Image PSNR gate to main parity (27.0)ad4f1b5c88 fix: preserve ec_transfer_params through omni request-output construction8191ddf72a chore: SPDX headers to the vLLM-Omni convention on touched files42a8a86080 tests: drop the branch-local scenarios-deferral copy superseded by main204a967090 rebase: align vllm-omni with vLLM v0.28.0 final (2cf0a6915ce5)
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex-magi2-preview 分支比较 111 提交 »894ab2b665 Address new MAGI-2 review feedback9d90200b3d Merge latest origin/main into codex-magi2-preview71cc362c34 Merge origin/main into codex-magi2-previewf0808c9793 [Bugfix] Fix remote replica membership lifecycle races (#5277)702a409664 [Feature][MiniCPM-o] Support configurable concurrent duplex sessions (#6021)
- frostbyte_neo 推送了仓库 github-featured/vllm-project--vllm-omni 的 codex-dlo-lifecycle-optimization 分支比较 22 提交 »1843f05ced [Model] Optimize MiniMax-H3 DLO component lifecycle29e3165bb9 [Test] add marker for stability and add Wan2.2 npu test case (#6343)51b7565f33 [perf] Add mask-free TRTLLM packed-padding path (#6542)b454e34966 [Refactor][2/N Scheduler]Replace _FULL_PAYLOAD_INPUT_STAGES with a resolved stage transport capability. (#6149)bae58f84b8 [minor, fix] Remove cuda sync on wake_up in AsyncOmni (#4092)
