Skip to content

[Bugfix][Core] Clean up failed HWR atomic metadata writes - #6956

Merged
hsliuustc0106 merged 6 commits into
vllm-project:mainfrom
BANANASJIM:bugfix/hwr-atomic-json-cleanup
Sep 11, 2026
Merged

[Bugfix][Core] Clean up failed HWR atomic metadata writes#6956
hsliuustc0106 merged 6 commits into
vllm-project:mainfrom
BANANASJIM:bugfix/hwr-atomic-json-cleanup

Conversation

@BANANASJIM

@BANANASJIM BANANASJIM commented Sep 2, 2026

Copy link
Copy Markdown
Contributor

Purpose

Host Weight Runtime publishes domain.json and domain-policy.json through exclusively created temporary files followed by an atomic rename. A close or replace failure could leave the temporary metadata file behind, while secondary close or unlink failures during write-error handling could obscure the original write error.

Clean up the owned temporary file after write, close, or replace failures on a best-effort basis while preserving the primary exception. Successful publication and parent-directory synchronization remain unchanged.

Test Plan

Add CPU fault-injection coverage for:

  • cleanup after a replace failure while preserving existing metadata;
  • preservation of the replace error when temporary-file cleanup also fails;
  • preservation of the write error when close and cleanup also fail;
  • cleanup after a standalone close failure.
python -m pytest -n 0 -q \
  tests/host_weight_runtime/test_filesystem_store.py \
  -k atomic_json

python -m pytest -n 0 -q \
  tests/host_weight_runtime/test_filesystem_store.py

pre-commit run --files \
  vllm_omni/host_weight_runtime/filesystem/store.py \
  tests/host_weight_runtime/test_filesystem_store.py

vLLM Version: 0.28.0

vLLM-Omni Commit: b412ce2

Test Result

  • Focused atomic metadata tests: 4 passed, 80 deselected
  • Filesystem store tests: 83 passed, 1 skipped
  • All applicable pre-commit hooks for the touched files passed.

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Sep 2, 2026

Copy link
Copy Markdown

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

@vllm-omni-review-bot

vllm-omni-review-bot commented Sep 2, 2026

Copy link
Copy Markdown

This PR appears to belong to: docs/design/module/host_weight_runtime.md.

Module owners: @hsliuustc0106

@BANANASJIM, please review your own changes and leave a short self-review comment describing what you checked. PRs without author self-review may not be assigned a reviewer.

Please take a look when you have a chance. If you would like an automated review, mention @vllm-omni-review-bot in a comment.

@hsliuustc0106 hsliuustc0106 added the ready label to trigger buildkite CI label Sep 2, 2026
hsliuustc0106
hsliuustc0106 approved these changes Sep 3, 2026
View reviewed changes

@hsliuustc0106 hsliuustc0106 left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed the atomic metadata write failure paths and regression coverage; no actionable findings.

@BANANASJIM
BANANASJIM force-pushed the bugfix/hwr-atomic-json-cleanup branch from 0ee6098 to 2340e8b Compare September 3, 2026 03:07
@hsliuustc0106 hsliuustc0106 added ready label to trigger buildkite CI and removed ready label to trigger buildkite CI labels Sep 3, 2026
NickCao
NickCao reviewed Sep 3, 2026
View reviewed changes
Comment thread vllm_omni/host_weight_runtime/filesystem/store.py Outdated
@@ -301,22 +301,27 @@ def _write_atomic_json(path: Path, value: object, *, mode: int = 0o444) -> None:
temporary = path.parent / f".{path.name}.{os.getpid()}.{uuid.uuid4().hex}.tmp"
fd = os.open(temporary, os.O_CREAT | os.O_EXCL | os.O_WRONLY | os.O_CLOEXEC, 0o600)

@NickCao NickCao Sep 3, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we use a NamedTemporaryFile context manager, that's simpler.

@BANANASJIM
BANANASJIM force-pushed the bugfix/hwr-atomic-json-cleanup branch 4 times, most recently from cea0a6a to 2efce32 Compare September 4, 2026 07:21
@hsliuustc0106 hsliuustc0106 added high priority high priority issue, needs to be done asap and removed ready label to trigger buildkite CI labels Sep 4, 2026
@BANANASJIM
BANANASJIM force-pushed the bugfix/hwr-atomic-json-cleanup branch from bb72a01 to c6d051c Compare September 5, 2026 07:45
@hsliuustc0106 hsliuustc0106 added the ready label to trigger buildkite CI label Sep 5, 2026
@baonudesifeizhai baonudesifeizhai mentioned this pull request Sep 5, 2026
Open
1 task done
@BANANASJIM
[Bugfix][Core] Clean up failed atomic metadata writes
af9ac94 
Signed-off-by: BANANASJIM <bananasjim1@gmail.com>
@BANANASJIM
[Bugfix][Core] Simplify atomic metadata writes with NamedTemporaryFile
df54b77 
Replace the manual open/write-loop/close bookkeeping in
_write_atomic_json with a tempfile.NamedTemporaryFile(delete=False)
context manager, per review feedback. The temporary file keeps the same
hidden ".<name>.*.tmp" shape in the destination directory, is still
created O_CREAT|O_EXCL with 0o600, and is still fsync'd, chmod'd to the
publication mode, fsync'd again and closed before os.replace. The single
outer except still unlinks the owned temporary on a best-effort basis
while preserving the primary exception.

Behavior change: when the write body fails *and* closing the handle also
fails, the close error now propagates with the write error attached as
its __context__, instead of the close error being swallowed. The
regression tests are updated to inject faults at the temporary-file
handle instead of at os.open/os.write/os.close.

Signed-off-by: BANANASJIM <bananasjim1@gmail.com>
@BANANASJIM
[Bugfix][Core] Preserve metadata write errors during close failures
ea04700 
Signed-off-by: BANANASJIM <bananasjim1@gmail.com>
@BANANASJIM
BANANASJIM force-pushed the bugfix/hwr-atomic-json-cleanup branch from 3b7a57a to ea04700 Compare September 6, 2026 08:13
@hsliuustc0106 hsliuustc0106 removed the ready label to trigger buildkite CI label Sep 9, 2026
@hsliuustc0106 hsliuustc0106 added the ready label to trigger buildkite CI label Sep 9, 2026
@heyuanliu-intel heyuanliu-intel mentioned this pull request Sep 10, 2026
Merged
@hsliuustc0106
hsliuustc0106 merged commit 44fed17 into vllm-project:main Sep 11, 2026
8 of 9 checks passed
JoseCarlosGarcia95 added a commit to valendra-tech/vllm-omni that referenced this pull request Sep 16, 2026
@JoseCarlosGarcia95 @Asthenia0412
Asthenia @hsliuustc0106 @andyluo7 @liangmenghuang @claude @hmellor @KrystalRay KrystalRay @twu3202 @specture724 @smartDream-chao @eval-dev @sphinxkkkbc @NancyFyong @ZJLi2013 @cursoragent @rk9595 @ZenAlexa @BANANASJIM @EchoHayate TRAE CLI @XuTianle0101 @yenuo26 @natureofnature @mglyn @Sparks-M @chi030303 @bobboli @NumberWan @zyz111222 @ZhengWG @amy-why-3459 @kunkunblueberry @Shaun-Walsh @NickCao @CarrotSwordsman mershi @yuweih205 @bojiang-li @Sy0307 @Joshna-Medisetty @Hiro208 hsliu_ustc @Gaohan123 @FayeSpica @tzhouam @herotai214 @LHXuuu @SamitHuang @wkutak @rahul-steiger-nv @tlysanhuo @psv666 @MaciejBalaNV @zhang-keliang @linyueqian @david6666666 @yancaocn yancaocn @suyanli220 suyan.li @wtz2333 @gxxx-hum @kuafou QI JIA @codex @MrlixiangWE @abinggo @junpengw67-max @congw729 @gerayking @Holworth
chore: resolve current upstream conflicts for Qwen3 duplex (#7)
7702a6f 
* [Bugfix][Examples] Use --profiler-config flag in offline TTS examples (vllm-project#6763)

Signed-off-by: Asthenia <asthenia0412@gmail.com>
Co-authored-by: Asthenia <asthenia0412@gmail.com>

* [Bugfix] Skip HWR store-size scans when no limit is configured (vllm-project#7131)

Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [CI][ROCm] Route LTX2 Ulysses parity to two-GPU lane (vllm-project#7234)

Signed-off-by: andyluo7 <andy.luo@amd.com>

* [Bugfix][Model] GR00T-N1.7: honor the per-request seed for flow-matching noise (vllm-project#7253)

Signed-off-by: liangmengh <liangmengh@nvidia.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>

* Add vLLM-Omni library info to Hugging Face Hub requests (vllm-project#5381)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

* [Bugfix][NPU] Limit MiniMax H3 modulation grid size (vllm-project#6794)

Signed-off-by: KrystalRay <keeleiray@gmail.com>
Co-authored-by: KrystalRay <keeleiray@gmail.com>

* [Bugfix] Build the forced-aligner prompt without a chat template (word timestamps one bin late) (vllm-project#7240)

Signed-off-by: Tianyao Wu <rayroy31@gmail.com>

* [Refactor][Diffusion] Resolve offload topology through one plan resolver (vllm-project#7209)

Signed-off-by: specture724 <specture724@gmail.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

* [Doc] Add AI usage policy for contributions (vllm-project#7305)

Signed-off-by: hsliuustc0106 <liuhongsheng4@huawei.com>

* [Bugfix][MiMo-Audio] Align code2wav decode with tokenizer device (vllm-project#6539)

Signed-off-by: chaosansui <zzc15560846421@163.com>
Signed-off-by: Zhichao Zhang <60429419+smartDream-chao@users.noreply.github.com>

* [Bugfix][MiniCPM-o] Fix the audio_embeds input path (vllm-project#5730)

Signed-off-by: eval-dev <0xe5bca0@gmail.com>
Signed-off-by: eval <74645252+eval-dev@users.noreply.github.com>

* [Feat][OmniVoice]Support Varlen Attn,  Request-Batch and Step-Execution (vllm-project#6408)

Signed-off-by: boatman <109857087+sphinxkkkbc@users.noreply.github.com>

* [Model] Add Audio8 TTS Preview 0.6B (DualAR, 44.1 kHz codec) (vllm-project#6157)

Signed-off-by: NancyFyong <NancyFyong@users.noreply.github.com>
Co-authored-by: NancyFyong <NancyFyong@users.noreply.github.com>

* [Bugfix][Frontend] Accept the msgpack-numpy package's numpy markers on the OpenPI endpoint (vllm-project#6051)

Signed-off-by: zjli2013 <leezhengjiang@126.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

* [Frontend] Opt-in WebSocket TTS split_granularity and session seed (vllm-project#7046)

Signed-off-by: rk9595 <rakesh.kariya@somaiya.edu>
Signed-off-by: Rakesh Kariya <rakesh.kariya@somaiya.edu>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

* [Bugfix][Frontend] Clear the P0 multimodal cache through the renderer (vllm-project#7003)

Signed-off-by: ZenAlexa <zimingwang945@gmail.com>

* [Bugfix][Frontend] Enforce image pixel limits for video input references (vllm-project#6963)

Signed-off-by: BANANASJIM <bananasjim1@gmail.com>

* [Bugfix][TTS] Isolate shared Higgs v3 reference encode from request cancellation (vllm-project#7076)

Signed-off-by: Allen Wu <allenwu2795@gmail.com>
Co-authored-by: TRAE CLI <traecli@bytedance.com>

* [Bugfix][CosyVoice3] Resolve hash snapshot pipeline (vllm-project#6896)

Signed-off-by: xutianle <xutianle@fudan.edu.cn>

* [CI] Skip Qwen3-Omni Server VAD multi-turn realtime test (vllm-project#7279) (vllm-project#7314)

Signed-off-by: wangyu <410167048@qq.com>

* [Bugfix][Magi2] Allow import without an active Triton driver (vllm-project#7239)

Signed-off-by: andyluo7 <andy.luo@amd.com>

* [Core] Split Omni connector model runner mixin (vllm-project#6903)

Signed-off-by: natureofnature <wzliu@connect.hku.hk>

* [Bugfix] Make LTX vocoder decoding deterministic (vllm-project#7231)

Signed-off-by: mglyn <1203789601@qq.com>

* [Doc] [Recipe] Add FLUX.1-schnell recipe for RTX 5090 32GB (vllm-project#7299)

Signed-off-by: Sparks-M <41097544+Sparks-M@users.noreply.github.com>

* [Doc] Qwen3-TTS: add 0.6B on 1x A100 40GB (vllm-project#7289)

Signed-off-by: chi030303 <106855944+chi030303@users.noreply.github.com>

* [Perf][Model] Add optimized LTX-2.5 DiffVAE operators (vllm-project#7308)

Signed-off-by: mglyn <1203789601@qq.com>

* [2/N] Add a minimal temporal chunk callback for MiniMax-H3 (vllm-project#7017)

Signed-off-by: specture724 <specture724@gmail.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>

* [Feature][Diffusion] Expose detailed pipeline timings (vllm-project#6822)

Signed-off-by: Bo Li <22713281+bobboli@users.noreply.github.com>

* [Bugfix] Resolve vllm-project#6931 hub FA3 on torch 2.13 via kernels 0.16.1 (vllm-project#7185)

Signed-off-by: NumberWan <wantszkin2003@gmail.com>

* [Bugfix][Ascend] fix npu 310/a5 bugs (vllm-project#6685)

Signed-off-by: zouyizhou <zouyizhou@huawei.com>

* [Bugfix][Engine] Group overlapping device stages into one sequential init component (vllm-project#7328)

Signed-off-by: ZhengWG <zwg0606@gmail.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>

* fix: reserve Qwen3-Omni NVFP4 backend fix (vllm-project#7200)

Signed-off-by: kunkunblueberry <1833921874@qq.com>

* [BugFix] Add field validators for /v1/audio/generate request (vllm-project#4741)

Signed-off-by: Shaun Walsh <shaunwalsh24@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Nick Cao <ncao@redhat.com>

* [CI][ROCm] Match CUDA/NPU L2/L3 label routing (vllm-project#6966)

Signed-off-by: andyluo7 <andy.luo@amd.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [CI/Build] Avoid duplicate stage CLI deploy config (vllm-project#7007)

Signed-off-by: mershi <mershi@tencent.com>
Co-authored-by: mershi <mershi@tencent.com>

* [CI/Build][ROCm] Normalize SenseNova paged-decode hardware markers (vllm-project#6935)

Signed-off-by: andyluo7 <andy.luo@amd.com>

* [Model] Skip unused frame packing in Wan2.2 S2V (vllm-project#7155)

Signed-off-by: hyw <yuweih205@gmail.com>

* [Doc] Add dual DGX Spark MiniMax-H3 results (vllm-project#7343)

Signed-off-by: bojiang-li <327132355+bojiang-li@users.noreply.github.com>

* [Model] Optimize MOSS-TTS Local batched execution and streaming codec (vllm-project#7202)

Signed-off-by: Sy03 <1370724210@qq.com>

* [Bugfix][XPU] Restore N-D output shape for W8A16 FP8 linear (vllm-project#7301)

Signed-off-by: Joshna Medisetty <joshna.medisetty@intel.com>
Signed-off-by: Joshna-Medisetty <joshna.medisetty@intel.com>

* [Doc] Document num_outputs_per_prompt for /v1/videos (vllm-project#7341)

Signed-off-by: Guangjian <hiro20833@gmail.com>

* [Skills] Add perf-evidence isolation, stage-attribution, and realtime-contract requirements (vllm-project#6820)

Signed-off-by: hsliu_ustc <hsliu_ustc@noreply.gitcode.com>
Co-authored-by: hsliu_ustc <hsliu_ustc@noreply.gitcode.com>

* [Bugfix] Allow LLM replicas on different GPUs to initialize concurrently (vllm-project#7292)

Signed-off-by: Gao Han <hgaoaf@connect.ust.hk>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>

* [CI/Build] Stabilize LTX2 vocoder autocast test on ROCm (vllm-project#7336)

Signed-off-by: andyluo7 <andy.luo@amd.com>

* [NPU][CI] Add A5 and 310P CI support (vllm-project#6875)

Signed-off-by: Weiming Liao <liaowm5@gmail.com>
Co-authored-by: wangyu <53896905+yenuo26@users.noreply.github.com>

* [Kernel] Enable LTX DiffVAE fusions on SM100 and SM103 (vllm-project#7350)

Signed-off-by: mglyn <1203789601@qq.com>

* [Bugfix][MiniCPM-o] Align structured chat content with native omni rendering (vllm-project#7344)

Signed-off-by: Sy03 <1370724210@qq.com>

* [Rebase] Rebase to vLLM 0.29.0 (vllm-project#7230)

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
Signed-off-by: Zhou Taichang <tzhouam@connect.ust.hk>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

* [Refactor] P0.2: Migrate API server helpers out of api_server (vllm-project#5453)

Signed-off-by: herotai214 <herotai214@gmail.com>

* [CI] Stabilize Qwen3-Omni Server VAD E2E (vllm-project#7356)

Signed-off-by: LHXuuu <xulianhao.xlh@antgroup.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>

* [CI/Build] Diff-aware source_file_dependencies for CUDA/NPU pipelines (vllm-project#6597)

Signed-off-by: wangyu <410167048@qq.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

* [Core][Diffusion] Add a typed pre-D2H video media contract (vllm-project#6615)

Signed-off-by: NancyFyong <NancyFyong@users.noreply.github.com>
Signed-off-by: Samit <285365963@qq.com>
Co-authored-by: NancyFyong <NancyFyong@users.noreply.github.com>
Co-authored-by: Samit <285365963@qq.com>

* [Bugfix] Bound HWR domain initialization lock waits (vllm-project#7128)

Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Bugfix] Escalate diffusion worker shutdown and retain survivors (vllm-project#7126)

Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Misc] Add standalone safetensors retention diagnostic (vllm-project#7145)

Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [CI] Isolate layerwise offload memory measurements (vllm-project#6938)

Signed-off-by: andyluo7 <andy.luo@amd.com>

* [Model] Add Cosmos3 mixed W8A8/W8A16 and W4A4/W4A16 denoising (vllm-project#6560)

Signed-off-by: Rahul Steiger <rsteiger@aws-cmh-slurm-1-vscode-04.cm.cluster>
Signed-off-by: Wojciech Kutak <wkutak@nvidia.com>
Co-authored-by: Rahul Steiger <rsteiger@nvidia.com>

* [Test] Use public render_jinja_template in MiniCPM-o native template test (vllm-project#7362)

Signed-off-by: tly <2200895168@qq.com>

* [Bugfix] Fix video prewarm cache retention and cancel-restart delay (vllm-project#7363)

Signed-off-by: psv666 <2693925048@qq.com>

* Cosmos3 action policy improvements (vllm-project#6460)

Signed-off-by: Maciej Bala <mbala@nvidia.com>
Signed-off-by: MaciejBalaNV <mbala@nvidia.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [BugFix][CI] Restore diff-aware source filtering for post-merge L3 (vllm-project#7371)

Signed-off-by: wangyu <410167048@qq.com>

* [Bugfix] Fail when a diffusion LoRA adapter binds no layer (vllm-project#7349)

Signed-off-by: Guangjian <hiro20833@gmail.com>

* [Bugfix] Fix host-memory leak on aborted /v1/images/generations (vllm-project#6462) (vllm-project#6561)

Signed-off-by: summer <128961079+zhang-keliang@users.noreply.github.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Refactor] Declare model-local KV held outside the paged manager (vllm-project#6171)

Signed-off-by: Yueqian Lin <linyueqian@outlook.com>

* [Realtime] Emit current (non-beta) OpenAI audio/transcript event names (vllm-project#7339)

Signed-off-by: Nick Cao <ncao@redhat.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>

* [Bugfix][Core] Clean up failed HWR atomic metadata writes (vllm-project#6956)

Signed-off-by: BANANASJIM <bananasjim1@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Bugfix] Keep MiniMax-H3 reference audio budgets separate (vllm-project#7281)

Signed-off-by: david6666666 <530634352@qq.com>

* [Bugfix] Fix Helios USP: per-component split for correct sequence parallelism (vllm-project#6930)

Signed-off-by: yancaocn <yancaochn@163.com>
Co-authored-by: yancaocn <yancaochn@163.com>

* [Perf][Diffusion] Optimize HSDP startup via Rank-0 shared weight loading and accelerated LoRA delta computation (vllm-project#7005)

Signed-off-by: samithuang <285365963@qq.com>

* [Example] Migrate HunyuanImage-3.0 to model_extras + shared task examples (vllm-project#5559)

Signed-off-by: suyanli220 <suyanli220@gmail.com>
Signed-off-by: suyan.li <suyan.li@bytedance.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: suyan.li <suyan.li@bytedance.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Model] Avoid scalar synchronizations in GLM-Image preparation (vllm-project#7172)

Signed-off-by: hyw <yuweih205@gmail.com>

* [Model][ERNIE-Image] Delay AdaLN modulation broadcast (vllm-project#7171)

Signed-off-by: hyw <yuweih205@gmail.com>

* [Kernel][MiniMax-H3] Run Q/K RMSNorm-RoPE in one launch (vllm-project#7167)

Signed-off-by: hyw <yuweih205@gmail.com>

* [CI][ROCm] Align AMD image with vLLM 0.29 (vllm-project#7395)

Signed-off-by: andyluo7 <andy.luo@amd.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Bugfix] Add embed_multimodal to MiniCPM-o 4.5 omni LLM class (vllm-project#7384)

Signed-off-by: Guangjian <hiro20833@gmail.com>

* [Model] Add LingBot World Ulysses sequence parallelism (vllm-project#6841)

Signed-off-by: wtz2333 <2955110911@qq.com>
Co-authored-by: Zhou Taichang <tzhouam@connect.ust.hk>

* [Feature][TTS] Add Speech API streaming metrics (vllm-project#6853)

Signed-off-by: XIN GAO <1037396230@qq.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Bugfix][Model] Fix FLUX.2 Klein multi-image edit metadata (vllm-project#7430)

Signed-off-by: QI JIA <qi.jia@shengshu.ai>
Co-authored-by: QI JIA <qi.jia@shengshu.ai>
Co-authored-by: Cursor <cursoragent@cursor.com>

* [BugFix] Fix leftovers of the legacy OpenAI realtime API event names (vllm-project#7426)

Signed-off-by: Nick Cao <ncao@redhat.com>
Co-authored-by: Codex <noreply@openai.com>

* [Model] Add Tencent AuK speech generation and editing (encoder + diffusion pipeline) (vllm-project#7385)

Signed-off-by: Yueqian Lin <linyueqian@outlook.com>
Co-authored-by: Sy03 <1370724210@qq.com>

* [XPU][Docker] Align XPU image and CI with vLLM v0.29.0 (vllm-project#7441)

Signed-off-by: Joshna-Medisetty <joshna.medisetty@intel.com>

* [Bugfix] Add explicit error when using CFGP with distilled Cosmos3 models (vllm-project#7427)

Signed-off-by: Maciej Bala <mbala@nvidia.com>

* [Perf][Diffusion] Run MammothModa2 DiT attention through the shared attention layer (vllm-project#7094)

Signed-off-by: MrlixiangWE <mrdanaer@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Bugfix] Give model CLI flags typed owners in the Omni config (vllm-project#7390)

Signed-off-by: Guangjian <hiro20833@gmail.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>

* [Bugfix] Require a model for `vllm serve --omni` (fixes vllm-project#4158) (vllm-project#4167)

Signed-off-by: abinggo <107740309+abinggo@users.noreply.github.com>

* [Bugfix] Send a downstream terminal chunk when a parked stage ends (vllm-project#6889)

Signed-off-by: psv666 <2693925048@qq.com>

* [NPU] upgrade to v0.29.0 (vllm-project#7433)

Signed-off-by: Weiming Liao <liaowm5@gmail.com>

* [Bugfix][Model][Lance] Support decoded video frames in video editing (vllm-project#5128)

Signed-off-by: 吴俊鹏 <248679769+junpengw67-max@users.noreply.github.com>
Co-authored-by: 吴俊鹏 <248679769+junpengw67-max@users.noreply.github.com>

* [Refactor][Diffusion] Remove model-specific names from LoRA and ModelOpt loader defaults (vllm-project#5907)

Signed-off-by: Alicia <115451386+congw729@users.noreply.github.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* Optimize CosyVoice3 Stage1 flow batching (vllm-project#4876)

Signed-off-by: gerayking <399geray@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [3/N] Encode streamed video on the worker with bounded batching (vllm-project#7018)

Signed-off-by: specture724 <specture724@gmail.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

* [Kernel][Boogu-Image] Fuse Q/K RMSNorm + interleaved RoPE via fused_qk_norm_rope (vllm-project#6982)

Signed-off-by: Qihan Kang <rollykanggg@gmail.com>

* [Bugfix][Frontend] Honor output_compression on the image generations route (vllm-project#7447)

Signed-off-by: hsliu_ustc <hsliu_ustc@noreply.gitcode.com>
Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>

---------

Signed-off-by: Asthenia <asthenia0412@gmail.com>
Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Signed-off-by: andyluo7 <andy.luo@amd.com>
Signed-off-by: liangmengh <liangmengh@nvidia.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: KrystalRay <keeleiray@gmail.com>
Signed-off-by: Tianyao Wu <rayroy31@gmail.com>
Signed-off-by: specture724 <specture724@gmail.com>
Signed-off-by: hsliuustc0106 <liuhongsheng4@huawei.com>
Signed-off-by: chaosansui <zzc15560846421@163.com>
Signed-off-by: Zhichao Zhang <60429419+smartDream-chao@users.noreply.github.com>
Signed-off-by: eval-dev <0xe5bca0@gmail.com>
Signed-off-by: eval <74645252+eval-dev@users.noreply.github.com>
Signed-off-by: boatman <109857087+sphinxkkkbc@users.noreply.github.com>
Signed-off-by: NancyFyong <NancyFyong@users.noreply.github.com>
Signed-off-by: zjli2013 <leezhengjiang@126.com>
Signed-off-by: rk9595 <rakesh.kariya@somaiya.edu>
Signed-off-by: Rakesh Kariya <rakesh.kariya@somaiya.edu>
Signed-off-by: ZenAlexa <zimingwang945@gmail.com>
Signed-off-by: BANANASJIM <bananasjim1@gmail.com>
Signed-off-by: Allen Wu <allenwu2795@gmail.com>
Signed-off-by: xutianle <xutianle@fudan.edu.cn>
Signed-off-by: wangyu <410167048@qq.com>
Signed-off-by: natureofnature <wzliu@connect.hku.hk>
Signed-off-by: mglyn <1203789601@qq.com>
Signed-off-by: Sparks-M <41097544+Sparks-M@users.noreply.github.com>
Signed-off-by: chi030303 <106855944+chi030303@users.noreply.github.com>
Signed-off-by: Bo Li <22713281+bobboli@users.noreply.github.com>
Signed-off-by: NumberWan <wantszkin2003@gmail.com>
Signed-off-by: zouyizhou <zouyizhou@huawei.com>
Signed-off-by: ZhengWG <zwg0606@gmail.com>
Signed-off-by: kunkunblueberry <1833921874@qq.com>
Signed-off-by: Shaun Walsh <shaunwalsh24@gmail.com>
Signed-off-by: mershi <mershi@tencent.com>
Signed-off-by: hyw <yuweih205@gmail.com>
Signed-off-by: bojiang-li <327132355+bojiang-li@users.noreply.github.com>
Signed-off-by: Sy03 <1370724210@qq.com>
Signed-off-by: Joshna Medisetty <joshna.medisetty@intel.com>
Signed-off-by: Joshna-Medisetty <joshna.medisetty@intel.com>
Signed-off-by: Guangjian <hiro20833@gmail.com>
Signed-off-by: hsliu_ustc <hsliu_ustc@noreply.gitcode.com>
Signed-off-by: Gao Han <hgaoaf@connect.ust.hk>
Signed-off-by: Weiming Liao <liaowm5@gmail.com>
Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
Signed-off-by: Zhou Taichang <tzhouam@connect.ust.hk>
Signed-off-by: herotai214 <herotai214@gmail.com>
Signed-off-by: LHXuuu <xulianhao.xlh@antgroup.com>
Signed-off-by: Samit <285365963@qq.com>
Signed-off-by: Rahul Steiger <rsteiger@aws-cmh-slurm-1-vscode-04.cm.cluster>
Signed-off-by: Wojciech Kutak <wkutak@nvidia.com>
Signed-off-by: tly <2200895168@qq.com>
Signed-off-by: psv666 <2693925048@qq.com>
Signed-off-by: Maciej Bala <mbala@nvidia.com>
Signed-off-by: MaciejBalaNV <mbala@nvidia.com>
Signed-off-by: summer <128961079+zhang-keliang@users.noreply.github.com>
Signed-off-by: Yueqian Lin <linyueqian@outlook.com>
Signed-off-by: Nick Cao <ncao@redhat.com>
Signed-off-by: david6666666 <530634352@qq.com>
Signed-off-by: yancaocn <yancaochn@163.com>
Signed-off-by: samithuang <285365963@qq.com>
Signed-off-by: suyanli220 <suyanli220@gmail.com>
Signed-off-by: suyan.li <suyan.li@bytedance.com>
Signed-off-by: wtz2333 <2955110911@qq.com>
Signed-off-by: XIN GAO <1037396230@qq.com>
Signed-off-by: QI JIA <qi.jia@shengshu.ai>
Signed-off-by: MrlixiangWE <mrdanaer@gmail.com>
Signed-off-by: abinggo <107740309+abinggo@users.noreply.github.com>
Signed-off-by: 吴俊鹏 <248679769+junpengw67-max@users.noreply.github.com>
Signed-off-by: Alicia <115451386+congw729@users.noreply.github.com>
Signed-off-by: gerayking <399geray@gmail.com>
Signed-off-by: Qihan Kang <rollykanggg@gmail.com>
Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
Signed-off-by: José Carlos <jose@valendra.tech>
Co-authored-by: Yancy <138764723+Asthenia0412@users.noreply.github.com>
Co-authored-by: Asthenia <asthenia0412@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Co-authored-by: andyluo7 <43718156+andyluo7@users.noreply.github.com>
Co-authored-by: liangmenghuang <liangmengh@nvidia.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Lei Ke <1141466880@qq.com>
Co-authored-by: KrystalRay <keeleiray@gmail.com>
Co-authored-by: Tianyao Wu <54675599+twu3202@users.noreply.github.com>
Co-authored-by: Anjie Hou <149605198+specture724@users.noreply.github.com>
Co-authored-by: Zhichao Zhang <60429419+smartDream-chao@users.noreply.github.com>
Co-authored-by: eval <74645252+eval-dev@users.noreply.github.com>
Co-authored-by: boatman <1930807094@qq.com>
Co-authored-by: NancyFyong <88076188+NancyFyong@users.noreply.github.com>
Co-authored-by: NancyFyong <NancyFyong@users.noreply.github.com>
Co-authored-by: zhengjia <ZJLi2013@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Rakesh Kariya <83279947+rk9595@users.noreply.github.com>
Co-authored-by: Ziming Wang <125807850+ZenAlexa@users.noreply.github.com>
Co-authored-by: Jim Ban <77719403+BANANASJIM@users.noreply.github.com>
Co-authored-by: Allen Wu <85376543+EchoHayate@users.noreply.github.com>
Co-authored-by: TRAE CLI <traecli@bytedance.com>
Co-authored-by: xutianle <24210290017@m.fudan.edu.cn>
Co-authored-by: wangyu <53896905+yenuo26@users.noreply.github.com>
Co-authored-by: NATURE <wzliu@connect.hku.hk>
Co-authored-by: Mu GuanLin <1203789601@qq.com>
Co-authored-by: Sparks <41097544+Sparks-M@users.noreply.github.com>
Co-authored-by: chi030303 <106855944+chi030303@users.noreply.github.com>
Co-authored-by: Bo Li <22713281+bobboli@users.noreply.github.com>
Co-authored-by: NumberWan <wantszkin2003@gmail.com>
Co-authored-by: zyz111222 <zouyizhou@huawei.com>
Co-authored-by: Zheng Wengang <zwg0606@gmail.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>
Co-authored-by: kunkun <72174834+kunkunblueberry@users.noreply.github.com>
Co-authored-by: Shaun Walsh <153730091+Shaun-Walsh@users.noreply.github.com>
Co-authored-by: Nick Cao <ncao@redhat.com>
Co-authored-by: shiyichuan <93317314+CarrotSwordsman@users.noreply.github.com>
Co-authored-by: mershi <mershi@tencent.com>
Co-authored-by: hyw <109567717+yuweih205@users.noreply.github.com>
Co-authored-by: bojiang-li <327132355+bojiang-li@users.noreply.github.com>
Co-authored-by: Sy03 <1370724210@qq.com>
Co-authored-by: Joshna-Medisetty <joshna.medisetty@intel.com>
Co-authored-by: Guangjian Dong <163994576+Hiro208@users.noreply.github.com>
Co-authored-by: hsliu_ustc <hsliu_ustc@noreply.gitcode.com>
Co-authored-by: Gao Han <hgaoaf@connect.ust.hk>
Co-authored-by: Weiming Liao <liaowm5@gmail.com>
Co-authored-by: Zhou Taichang <tzhouam@connect.ust.hk>
Co-authored-by: herotai214 <68222888+herotai214@users.noreply.github.com>
Co-authored-by: LHXuuu <xulianhao.xlh@antgroup.com>
Co-authored-by: Samit <285365963@qq.com>
Co-authored-by: wkutak <wkutak@nvidia.com>
Co-authored-by: Rahul Steiger <rsteiger@nvidia.com>
Co-authored-by: tlysanhuo <166924864+tlysanhuo@users.noreply.github.com>
Co-authored-by: psv666 <150513104+psv666@users.noreply.github.com>
Co-authored-by: MaciejBalaNV <mbala@nvidia.com>
Co-authored-by: summer <128961079+zhang-keliang@users.noreply.github.com>
Co-authored-by: Yueqian Lin <70319226+linyueqian@users.noreply.github.com>
Co-authored-by: WeiQing Chen <40507679+david6666666@users.noreply.github.com>
Co-authored-by: Yan Cao <31481315+yancaocn@users.noreply.github.com>
Co-authored-by: yancaocn <yancaochn@163.com>
Co-authored-by: SuyanLi <126558907+suyanli220@users.noreply.github.com>
Co-authored-by: suyan.li <suyan.li@bytedance.com>
Co-authored-by: wtz2333 <2955110911@qq.com>
Co-authored-by: GXIN <37653830+gxxx-hum@users.noreply.github.com>
Co-authored-by: Qi Jia <kuafou@gmail.com>
Co-authored-by: QI JIA <qi.jia@shengshu.ai>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: DanaerLee <mrdanaer@gmail.com>
Co-authored-by: longguo <107740309+abinggo@users.noreply.github.com>
Co-authored-by: junpengw67-max <junpengw67@gmail.com>
Co-authored-by: 吴俊鹏 <248679769+junpengw67-max@users.noreply.github.com>
Co-authored-by: Alicia <115451386+congw729@users.noreply.github.com>
Co-authored-by: geray <48796550+gerayking@users.noreply.github.com>
Co-authored-by: KANG Qihan <3149604185@qq.com>
khairulkabir1661 pushed a commit to khairulkabir1661/vllm-omni that referenced this pull request Sep 25, 2026
@BANANASJIM @hsliuustc0106
[Bugfix][Core] Clean up failed HWR atomic metadata writes (vllm-proje…
6cd84e2 
…ct#6956)

Signed-off-by: BANANASJIM <bananasjim1@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@NickCao NickCao NickCao left review comments
@hsliuustc0106 hsliuustc0106 hsliuustc0106 approved these changes
@yenuo26 yenuo26 Awaiting requested review from yenuo26 yenuo26 is a code owner
@Gaohan123 Gaohan123 Awaiting requested review from Gaohan123 Gaohan123 is a code owner
@ywang96 ywang96 Awaiting requested review from ywang96 ywang96 is a code owner

Assignees

No one assigned

Labels

high priority high priority issue, needs to be done asap ready label to trigger buildkite CI

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@BANANASJIM @vllm-omni-review-bot @NickCao @hsliuustc0106