anodyine/vllm-omni
1
0 Fork 0
Code Issues 4 Pull Requests 2 Actions Packages Projects Releases Wiki Activity
955 Commits 235 Branches 31 Tags
e6d4cdd71a46ab97f5682d0cff6cd3e76db73400
Commit Graph
161 Commits
Author SHA1 Message Date
tzhouam e6d4cdd71a Refactor model mode initialization and enhance error handling in AsyncOmni. Update worker type retrieval in omni_stage and add audio token handling in Qwen3OmniMoe classes for improved robustness. 
Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-03-09 06:17:08 +00:00
Taichang Zhou fd4607f215 rebase 2026/03/07 
Signed-off-by: Taichang Zhou <tzhouam@connect.ust.hk>
 2026-03-07 13:30:58 +08:00
Taichang Zhou 5511bc14c7 Merge origin/main into dev/rebase-v0170 
- Resolved conflict in vllm_omni/engine/output_processor.py: use tracing_enabled for vLLM v0.17.0 API
- Resolved conflict in vllm_omni/model_executor/models/qwen3_tts/qwen3_tts_code_predictor_vllm.py: keep correct _CodePredictorDecoderLayer.forward()
- Update Dockerfile.ci vLLM wheel commit to b31e9326a (releases/v0.17.0)

Made-with: Cursor

Signed-off-by: Taichang Zhou <tzhouam@connect.ust.hk>
 2026-03-07 13:18:33 +08:00
Yueqian Lin 3bc89de0c0 Add non-async chunk support for Qwen3-TTS (#1678) 
Signed-off-by: linyueqian <linyueqian@outlook.com>
 2026-03-06 14:45:32 +08:00
Baoyuan Qi 32fd084ae2 [MiMo-Audio] Bugfix tp lg than 1 (#1688) 
Signed-off-by: 齐保元 <qibaoyuan@xiaomi.com>
 2026-03-06 14:45:08 +08:00
Sy03 90ec499bd5 [Refactor][Perf] Qwen3-TTS: re-prefill Code Predictor with torch.compile + enable Code2Wav decoder CUDA Graph (#1617) 
Signed-off-by: Sy03 <1370724210@qq.com>
 2026-03-06 18:21:55 +13:00
Du BinlinyueqianYueqian Lin
 		fd51841230 perf: replace per-element .item() GPU syncs with batch .tolist() in TTS code predictor (#1614) 
Signed-off-by: dubin555 <dubin555@gmail.com>
Signed-off-by: linyueqian <linyueqian@outlook.com>
Co-authored-by: linyueqian <linyueqian@outlook.com>
Co-authored-by: Yueqian Lin <70319226+linyueqian@users.noreply.github.com>
 2026-03-06 12:22:01 +08:00
tzhouam 8c41492307 Merge remote-tracking branch 'remotes/origin/main' into dev/rebase-v0169 2026-03-06 03:02:50 +00:00
Jack Leegoogle-labs-jules[bot]JackLeeHal
 		c0b714a8ac [Feature] Support flexible task_type configuration for Qwen3-TTS models (#1197) 
Signed-off-by: Jack Lee <695697442@qq.com>
Signed-off-by: jackleehal <695697442@qq.com>
Co-authored-by: google-labs-jules[bot] <161369871+google-labs-jules[bot]@users.noreply.github.com>
Co-authored-by: JackLeeHal <15664755+JackLeeHal@users.noreply.github.com>
 2026-03-05 13:39:09 -05:00
zhanqiuhuandYueqian Lin e643aaec15 [Bugfix] Fix all-silence TTS output: use float32 for speech tokenizer decoder (#1664) 
Signed-off-by: Zhanqiu Hu <zh338@cornell.edu>
Co-authored-by: Yueqian Lin <70319226+linyueqian@users.noreply.github.com>
 2026-03-06 00:04:26 +08:00
Juan Pablo Zuluagaandpablo 6a45efb287 [Feat][Qwen3TTS] reduce TTFA with flexible initial phase (#1583) 
Signed-off-by: pablo <pablo@agigo.ai>
Co-authored-by: pablo <pablo@agigo.ai>
 2026-03-05 23:14:41 +08:00
Ziming Huang 070ea0ddc3 [Optimize][Qwen3-Omni] Reduce inter-packet latency in async chunk (#1656) 
Signed-off-by: ZeldaHuang <hzm414167@alibaba-inc.com>
 2026-03-05 22:26:30 +08:00
tzhouam 395eab8984 Merge remote-tracking branch 'remotes/origin/main' into dev/rebase-v0169 2026-03-05 02:29:14 +00:00
dsinghviandHongsheng Liu b63e7c7c9a [Model] Fun cosy voice3-0.5-b-2512 (#498) 
Signed-off-by: Divyansh Singhvi <divyanshsinghvi@gmail.com>
Signed-off-by: dsinghvi <divyanshsinghvi@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
 2026-03-04 19:54:59 -05:00
ShiCheng DengcatcatiwzbiHongsheng LiudsinghviCanlin GuoPeiqi Yin汪志鹏
 		1612948062 add support for MammothModa2 model (#336) 
Signed-off-by: HonestDeng <2958906959@qq.com>
Signed-off-by: iwzbi <iwzbi@zju.edu.cn>
Signed-off-by: iwzbi <wzbi@zju.edu.cn>
Signed-off-by: Divyansh Singhvi <divyanshsinghvi@gmail.com>
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
Signed-off-by: yinpeiqi <yinpeiqi809@gmail.com>
Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Signed-off-by: 汪志鹏 <wangzhipeng628@gmail.com>
Co-authored-by: catcat <108673086+iwzbi@users.noreply.github.com>
Co-authored-by: iwzbi <iwzbi@zju.edu.cn>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Co-authored-by: dsinghvi <divyanshsinghvi@gmail.com>
Co-authored-by: Canlin Guo <canlinguosdu@gmail.com>
Co-authored-by: Peiqi Yin <60515999+yinpeiqi@users.noreply.github.com>
Co-authored-by: 汪志鹏 <wangzhipeng628@gmail.com>
 2026-03-04 22:52:56 +08:00
usberkeleyandprincepride 84338494d2 [Model] Add Hunyuan Image3 AR Support (#759) 
Signed-off-by: Bradley <bradley.b.pitt@gmail.com>
Signed-off-by: princepride <wangzhipeng628@gmail.com>
Signed-off-by: 汪志鹏 <wangzhipeng628@gmail.com>
Co-authored-by: princepride <wangzhipeng628@gmail.com>
 2026-03-04 21:49:38 +08:00
tzhouam 36a507e071 Merge remote-tracking branch 'remotes/origin/main' into dev/rebase-v0170 
Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-03-04 08:26:56 +00:00
tzhouam ae818f4b45 Fix Qwen3Omni VisionTransformer compatibility with updated vLLM Qwen2_5_VisionAttention 
The upstream vLLM Qwen2_5_VisionAttention.forward() now requires a
`sequence_lengths` argument, but Qwen3_VisionBlock.forward() does not
pass it, causing a TypeError at init time.

Instead of patching the vLLM package directly, introduce a
Qwen3Omni_VisionTransformer subclass in vllm-omni that overrides
forward() to properly compute and propagate sequence_lengths via
MMEncoderAttention (following qwen3_vl.py's pattern). Both thinker
and talker now import the fixed subclass.

Also bump the vLLM wheel to e196028dbaa41b.

Made-with: Cursor

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-03-04 02:51:02 +00:00
Yueqian Lin a3240fd484 [Bugfix] Fix transformers 5.x compat issues in online TTS serving (#1536) 
Signed-off-by: linyueqian <linyueqian@outlook.com>
 2026-03-04 01:28:32 +08:00
tzhouam 2398a7c1d7 Merge origin/main into dev/rebase-v0170 
Resolve conflict in omni_generation_scheduler.py: keep both
num_cached_tokens tracking and additional_information caching.

Made-with: Cursor

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-03-03 14:31:05 +00:00
Junhong Liu 742d845b8e Make chunk_size and left_context_size configurable via YAML for async chunking (#1423) 
Signed-off-by: Junhong Liu <98734602+LJH-LBJ@users.noreply.github.com>
Signed-off-by: Junhong Liu <ljh_lbj@163.com>
 2026-03-03 22:06:52 +08:00
tzhouam 16824aa295 Add max_query_len parameter to _LocalPredictorKVCache for improved query handling 
Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-03-03 14:04:07 +00:00
zhanqiuhu 80c89e00cb [Bugfix] Fix Qwen3-TTS code predictor crash due to missing vLLM config context (#1619) 
Signed-off-by: Zhanqiu <zh338@cornell.edu>
 2026-03-03 11:02:13 +08:00
tzhouam 3e539fcb6a Merge remote-tracking branch 'origin/main' into dev/vllm-align 
Made-with: Cursor

# Conflicts:
#	vllm_omni/entrypoints/openai/serving_chat.py
#	vllm_omni/worker/gpu_model_runner.py

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-03-03 02:57:27 +00:00
Baoyuan Qi f371f7bf05 [bugfix] Fix unexpected argument 'is_finished' in function llm2code2wav_async_chunk of mimo-audio (#1570) 
Signed-off-by: Baoyuan Qi <qibaoyuan@126.com>
Signed-off-by: 齐保元 <qibaoyuan@xiaomi.com>
 2026-03-02 21:49:06 +08:00
rein yang f082f6eedb [Bugfix] fix kernel error for qwen3-omni (#1602) 
Signed-off-by: Rein Yang <ruiruyang2@gmail.com>
 2026-03-02 18:12:30 +08:00
汪志鹏andCursor e37a89ffcb [Feature]: Support cfg kv-cache transfer in multi-stage (#1422) 
Signed-off-by: princepride <wangzhipeng628@gmail.com>
Signed-off-by: 汪志鹏 <wangzhipeng628@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
 2026-03-02 05:06:28 +00:00
dsinghviandZhou Taichang 68764cc4f4 [Refactor]: Phase1 for rebasing_additional_info (#1394) 
Signed-off-by: dsinghvi <divyanshsinghvi@gmail.com>
Signed-off-by: Divyansh Singhvi <divyanshsinghvi@gmail.com>
Signed-off-by: Zhou Taichang <tzhouam@connect.ust.hk>
Co-authored-by: Zhou Taichang <tzhouam@connect.ust.hk>
 2026-03-02 10:19:41 +08:00
tzhouam 26744bcb35 Merge origin/main into dev/vllm-align 
Resolved conflicts:
- vllm_omni/inputs/preprocess.py: took origin/main's elif mm_processor_kwargs branch, fixed undefined mm_uuids reference
- vllm_omni/model_executor/models/qwen3_tts/qwen3_tts_code_predictor_vllm.py: took origin/main's NPU support

Made-with: Cursor

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-03-02 02:10:32 +00:00
汪志鹏 3d9fa8d53f [BugFix]: fix a lot of bug (#1565) 
Signed-off-by: princepride <wangzhipeng628@gmail.com>
 2026-02-28 16:26:56 +08:00
wangyu cd2234a1bf [CI] Modify some CI test cases to run on L4 environment to reduce H100 resource usage. (#1543) 
Signed-off-by: yenuo26 <410167048@qq.com>
Signed-off-by: wangyu <53896905+yenuo26@users.noreply.github.com>
 2026-02-28 14:41:51 +08:00
Canlin Guo c81266703a [NPU][Bugfix] Align GPU side and recover qwen3-tts (#1564) 
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
 2026-02-28 12:46:13 +08:00
Zhou Taichang fe0c02ee78 [Debug] Merge vllm pull 35368 (#1534) 
Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-02-28 10:39:25 +08:00
Junhong Liu 42d86af1f1 Fix no embed text spk tokens (#1540) 
Signed-off-by: Junhong Liu <98734602+LJH-LBJ@users.noreply.github.com>
 2026-02-28 09:33:58 +08:00
Sy03 0cc3481acf [Feat] : Support Async chunk cleanup (#1087) 
Signed-off-by: Sy03 <1370724210@qq.com>
 2026-02-27 14:39:33 +08:00
+13
Baoyuan Qiwangyuwangyu31577Zhang ShijinHongsheng LiuSihao LiXU MingshiCanlin GuoIsotr0pyJohnJanWeiQing Chendongbo910220Jiangyun ZhuJunhong LiuTJianshijin zhangZhou TaichangrootPrajwal AShijin Zhangdingningning dingCopilot
 		b19c227fb3 [New Model]: XiaomiMiMo/MiMo-Audio-7B-Instruct support (#750) 
Signed-off-by: wangyu31577 <wangyu31577@hundsun.com>
Signed-off-by: 齐保元 <qibaoyuan@xiaomi.com>
Signed-off-by: hsliu <liuhongsheng4@huawei.com>
Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Signed-off-by: GG-li <3226868735@qq.com>
Signed-off-by: Sihao Li <111170255+GG-li@users.noreply.github.com>
Signed-off-by: XU Mingshi <91017482+mxuax@users.noreply.github.com>
Signed-off-by: mxuax <mxuax@connect.ust.hk>
Signed-off-by: Baoyuan Qi <qibaoyuan@126.com>
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: wuzhongjian <wuzhongjian_yewu@cmss.chinamobile.com>
Signed-off-by: dongbo910220 <1275604947@qq.com>
Signed-off-by: dongbo910220 <32610838+dongbo910220@users.noreply.github.com>
Signed-off-by: Jiangyun Zhu <riverclouds.zhu@qq.com>
Signed-off-by: Junhong Liu <98734602+LJH-LBJ@users.noreply.github.com>
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
Signed-off-by: baoyuan qi <qibaoyuan@126.com>
Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
Signed-off-by: Prajwal A <prajwalanagani@gmail.com>
Signed-off-by: Shijin Zhang <75300765+Dovis01@users.noreply.github.com>
Signed-off-by: 丁宁 <nndding@gmail.com>
Signed-off-by: SHIJIN ZHANG <75300765+Dovis01@users.noreply.github.com>
Signed-off-by: dingning<dingning7@xiaomi.com>
Signed-off-by: dingning <dingning7@xiaomi.com>
Signed-off-by: dingning <dingning@xiaomi.com>
Co-authored-by: wangyu <53896905+yenuo26@users.noreply.github.com>
Co-authored-by: wangyu31577 <wangyu31577@hundsun.com>
Co-authored-by: Zhang Shijin <zhangshijin@xiaomi.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Co-authored-by: Sihao Li <111170255+GG-li@users.noreply.github.com>
Co-authored-by: XU Mingshi <91017482+mxuax@users.noreply.github.com>
Co-authored-by: Canlin Guo <canlinguosdu@gmail.com>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: JohnJan <wuzhongjian_yewu@cmss.chinamobile.com>
Co-authored-by: WeiQing Chen <40507679+david6666666@users.noreply.github.com>
Co-authored-by: dongbo910220 <32610838+dongbo910220@users.noreply.github.com>
Co-authored-by: Jiangyun Zhu <riverclouds.zhu@qq.com>
Co-authored-by: Junhong Liu <ljh_lbj@163.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
Co-authored-by: shijin zhang <zsj1364226740@gmail.com>
Co-authored-by: Zhou Taichang <tzhouam@connect.ust.hk>
Co-authored-by: root <root@hk01dgx028.cm.cluster>
Co-authored-by: Prajwal A <34590600+LawJarp-A@users.noreply.github.com>
Co-authored-by: Shijin Zhang <75300765+Dovis01@users.noreply.github.com>
Co-authored-by: dingning <dingning7@xiaomi.com>
Co-authored-by: ning ding <nndding@gmail.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
 2026-02-27 11:19:49 +08:00
Xu Luandxulusjb c9b31a4c1a feat(qwen3-tts): Add CUDA Graph support for speech tokenizer decoder (#1205) 
Signed-off-by: xulusjb <fdukeshik@gmail.com>
Co-authored-by: xulusjb <fdukeshik@gmail.com>
 2026-02-27 10:48:25 +08:00
Jared Wen汪志鹏Hongsheng Liu
 		4146095623 [Perf] GLM Image (#920) 
Signed-off-by: JaredforReal <w13431838023@gmail.com>
Signed-off-by: Jared Wen <w13431838023@gmail.com>
Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Co-authored-by: 汪志鹏 <wangzhipeng628@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
 2026-02-27 10:29:14 +08:00
amy-why-3459Hongsheng LiuCanlin Guo
 		e1337c89c7 [BugFix] Restore talker's config (#1524) 
Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
Co-authored-by: Canlin Guo <961750412@qq.com>
 2026-02-27 06:44:11 +08:00
amy-why-3459 66c3e04263 [Performance]Qwen3-Omni performance optimization (#1378) 
Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
 2026-02-26 21:21:36 +08:00
Juan Pablo ZuluagaGao HanHongsheng Liu
 		66457c39fa [Bugfix][Qwen3TTS] (#1289) 
Signed-off-by: pablo <juanz9312@gmail.com>
Co-authored-by: Gao Han <gaohan19@huawei.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
 2026-02-26 18:58:35 +08:00
tzhouam 8d0c46487e Merge origin/main into dev/vllm-align 
Resolve conflicts:
- docker/Dockerfile.ci: Keep wheel install section, update hash to 6042e66cd5
- vllm_omni/platforms/npu/worker/npu_model_runner.py: Accept main's refactoring

Made-with: Cursor

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-02-26 08:19:53 +00:00
Yuanheng Zhao 904b97d5bd [Fix][Chore] Qwen3-TTS Modeling Minor Code Sanity Improvements (#1482) 
Signed-off-by: yuanheng <jonathan.zhaoyh@gmail.com>
 2026-02-26 14:39:08 +08:00
Canlin Guo 22c8cde328 [NPU] Upgrade to v0.16.0 (#1375) 
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
 2026-02-26 14:26:50 +08:00
Sangchun Ha 315daeba8c [Bugfix] Fix case-sensitive task_type matching in Qwen3TTSModelForGeneration (#1455) 
Signed-off-by: Sangchun Ha <seomk9896@gmail.com>
 2026-02-26 10:27:13 +08:00
Zhou Taichang 8e950daf6a [Debug] Multi-Request for Qwen 3 Omni use_audio_in_video (#1433) 
Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
 2026-02-26 09:26:25 +08:00
Yanick Schraner 97f4c5b1e7 [Bugfix] Add TTS configuration options (#1177) 
Signed-off-by: Yanick Schraner <yanick.schraner@bs.ch>
 2026-02-26 00:14:44 +08:00
tzhouam 22e01c416b Merge remote-tracking branch 'origin/main' into dev/vllm-align 2026-02-25 08:15:55 +00:00
Juan Pablo Zuluagaandpablo 4de077ea62 [Qwen3TTS][Feat] Code2Wav batched decoding (#1426) 
Signed-off-by: pablo <pablo@agigo.ai>
Co-authored-by: pablo <pablo@agigo.ai>
 2026-02-25 11:16:46 +13:00
Canlin GuoandHongsheng Liu af11b02081 [Minor][Refactor] Pass seq_token_counts explicitly (#1425) 
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
 2026-02-24 14:10:00 +08:00