iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
13,776 Commits 3 Branches 0 Tags
94de871546e8da687c08ed8a7e0a26531500d4bd
Commit Graph
2189 Commits
Author SHA1 Message Date
Wentao YeandNick Hill e1060a71a1 [Perf] Optimize detokenizer python logic (#32975) 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
 2026-02-09 23:54:41 -08:00
Chen Zhang 97fa8f6590 [BugFix] Avoid prefix cache hit in the same schedule step for mamba layers (#29387) 
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
 2026-02-10 07:41:16 +00:00
Lucas WilkinsonandBenjamin Chislett 81e217fe6b [Bugfix] Fix DP Attention Padding in Dummy Run (#34187) 
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Benjamin Chislett <bchislett@nvidia.com>
 2026-02-10 05:29:39 +00:00
Roger Wang 8a5e0e2b2b [Bugfix][Core] Fix CPU memory leak from Request reference cycle in prefix caching (#34183) 
Signed-off-by: Roger Wang <hey@rogerw.io>
 2026-02-10 13:03:32 +08:00
Ning XieChaunceyRussell Bryant
 		13397841ab [structured output] validate unsupported json features first (#33233) 
Signed-off-by: Andy Xie <andy.xning@gmail.com>
Co-authored-by: Chauncey <chaunceyjiang@gmail.com>
Co-authored-by: Russell Bryant <rbryant@redhat.com>
 2026-02-09 23:49:09 +00:00
Nick Hill e7e52781ff [ModelRunner V2][BugFix] Fix max_query_len calculation (#34167) 
Signed-off-by: Nick Hill <nickhill123@gmail.com>
 2026-02-09 21:47:17 +00:00
ZhengHongming888andgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> cb62e86f83 Add NUMA Core binding in nixl_connector for CPU xPyD (#32365) 
Signed-off-by: Hongming Zheng <hongming.zheng@intel.com>
Signed-off-by: ZhengHongming888 <hongming.zheng@intel.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
 2026-02-09 15:39:12 +00:00
JJJYmmmwulipcywang96Isotr0pyIsotr0pyRoger Wang
 		9562912cea [MODEL] Adding Support for Qwen3.5 Models (#34110) 
Signed-off-by: JJJYmmm <1650675829@qq.com>
Signed-off-by: JJJYmmm <92386084+JJJYmmm@users.noreply.github.com>
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: wulipc <wulipc@users.noreply.github.com>
Co-authored-by: ywang96 <ywang96@users.noreply.github.com>
Co-authored-by: Isotr0py <Isotr0py@users.noreply.github.com>
Co-authored-by: Isotr0py <2037008807@qq.com>
Co-authored-by: Roger Wang <hey@rogerw.io>
 2026-02-09 21:12:58 +08:00
Reagan LeeandReagan Lee 7c233dbb36 [Tiny] Rename encoder budget file to more specific name (#34103) 
Signed-off-by: Reagan Lee <“reaganjlee@gmail.com”>
Co-authored-by: Reagan Lee <“reaganjlee@gmail.com”>
 2026-02-09 03:48:19 +00:00
aabbccddwasd 179ae7da8f [Revert] Fix performance regression for GLM-4.7-GPTQ decode and MTP acceptance rate (#33771) 
Signed-off-by: aabbccddwasd <aabbccddwasd@qq.com>
 2026-02-08 08:13:24 -08:00
Reagan LeeReagan Leegemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
 		c4df59ad43 Add embedding input functionality for disabled modalities [remake] (#32493) 
Signed-off-by: Reagan Lee <“reaganjlee@gmail.com”>
Signed-off-by: Reagan Lee <reaganjlee@gmail.com>
Signed-off-by: Reagan Lee <96998476+reaganjlee@users.noreply.github.com>
Co-authored-by: Reagan Lee <“reaganjlee@gmail.com”>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
 2026-02-08 04:57:16 -08:00
Nick Hill a96197f564 [Perf] Simplify DeepseekV32 tokenizer, ensure fast detokenization used (#33855) 
Signed-off-by: Nick Hill <nickhill123@gmail.com>
 2026-02-08 07:16:34 +00:00
Rohan Potdar de3869bb4d move checks out of unified_kv_cache_update custom op (#33943) 
Signed-off-by: Rohan138 <rohanpotdar138@gmail.com>
 2026-02-07 05:30:09 -08:00
lukec 15a0b9e570 Fix spelling errors (#33978) 2026-02-06 23:58:50 -08:00
Cyrus Leung 48312e579a [Misc] Make PlaceholderRange.get_num_embeds a method (#34035) 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
 2026-02-07 05:30:17 +00:00
Nick Hill 40218a82ba [ModelRunner V2] Revert token rank comparison difference for now (#34017) 
Signed-off-by: Nick Hill <nickhill123@gmail.com>
 2026-02-07 11:11:05 +08:00
Aaron HaoandRobert Shaw 89a385d79f [Feat][RL] Pause and Resume with keep requests for single engine (#32351) 
Signed-off-by: ahao-anyscale <ahao@anyscale.com>
Signed-off-by: Aaron Hao <ahao@anyscale.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
 2026-02-07 00:08:58 +00:00
zhrrr 16786da735 [Model Runner V2] support apply penalty for spec decode (#33251) 
Signed-off-by: zhuhaoran <zhuhaoran.zhr@alibaba-inc.com>
 2026-02-06 10:56:48 -08:00
Wentao Ye 67a746e87f [Log] Optimize duplicate startup log (#33944) 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
 2026-02-06 17:49:56 +00:00
Cyrus Leung cd8b405bd0 [Refactor] Consolidate sequence normalization and enc-dec parsing (#33928) 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
 2026-02-06 15:43:47 +00:00
Andreas Karatzas 350ca72c04 [ROCm][AITER] Fix AITER import regression for explicit backend selection (#33749) 
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
 2026-02-06 15:08:16 +00:00
FredericOdermatt 1fb0495a72 [FIX] guidance: use max(vocab_size, len(tokenizer)) for n_vocab (#33509) 
Signed-off-by: Frederic Odermatt <frederic.odermatt@44ai.ch>
 2026-02-06 14:23:03 +00:00
Xinyu Chen e969a169ef support view_from_cpu_tensor on XPU (#33868) 
Signed-off-by: Xinyu Chen <xinyu1.chen@intel.com>
 2026-02-06 08:34:20 +00:00
Gassan Salama 1363e3d6d5 [cpu][performance] CPU Paged Attention NEON BFMMLA BF16 Implementation (#32263) 
Signed-off-by: Gassan <gassan.salama@arm.com>
 2026-02-06 15:01:48 +08:00
Cyrus Leung 035a6cb09a [Misc] Update code for encoder-decoder models (#33900) 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
 2026-02-06 11:38:39 +08:00
emricksini-h 325ab6b0a8 [Feature] OTEL tracing during loading (#31162) 2026-02-05 16:59:28 -08:00
Benjamin Chislett af3162d3aa [Spec Decode] Unified Parallel Drafting (#32887) 
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
 2026-02-05 12:37:18 -05:00
Aaron HaoandSumanthRH c1858b7ec8 [Feat][RL][1/2] Native Weight Syncing API: NCCL (#31943) 
Signed-off-by: ahao-anyscale <ahao@anyscale.com>
Signed-off-by: Aaron Hao <ahao@anyscale.com>
Co-authored-by: SumanthRH <sumanthrh99@gmail.com>
 2026-02-05 12:13:23 -05:00
Cyrus Leungandwang.yuqi 038914b7c8 [Refactor] Move task outside of PoolingParams.verify (#33796) 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
Co-authored-by: wang.yuqi <yuqi.wang@daocloud.io>
 2026-02-05 09:33:11 +00:00
Mark McLoughlin 2abd97592f [KV Connector][Metrics] Do not count local prefix cache hits in connector queries (#30522) 
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
 2026-02-05 09:57:27 +02:00
Chauncey 6abb0454ad [Perf] Optimize the performance of structured output + reasoning (#33557) 
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
 2026-02-05 15:45:29 +08:00
Li, Jiang 07daee132b [CI/Build] Parallelize CPU CI tests (#33778) 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
 2026-02-05 13:53:48 +08:00
Nick Hill add9f1fbd9 [Minor] Include StreamingInput in inputs package (#33856) 
Signed-off-by: Nick Hill <nickhill123@gmail.com>
 2026-02-05 04:38:20 +00:00
Luka Govedič e3bf79ffa0 Revert "[Attention][FA3] Update FA3 to include new swizzle optimization" (#33841) 2026-02-04 19:54:27 -08:00
zhanqiuhu bbe0574d8e [Bugfix] Disable TRTLLM attention when KV transfer is enabled (#33192) 
Signed-off-by: Zhanqiu Hu <zh338@cornell.edu>
 2026-02-05 00:49:18 +00:00
Nick Hill fa4e0fb028 [Core] Don't schedule spec tokens with prefill chunks (#33652) 
Signed-off-by: Nick Hill <nickhill123@gmail.com>
 2026-02-04 23:40:22 +00:00
Taeksang Kim 6e98f6d8b6 Implement zero-copy GQA for multimodal and CPU (#33732) 
Signed-off-by: Taeksang Kim <ts.kim@hyperaccel.ai>
 2026-02-04 20:11:39 +00:00
kourosh hakhamaneshi 2f6d17cb2f [rocm][ray] Fix: Unify Ray device visibility handling across CUDA and ROCm (#33308) 
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
 2026-02-04 10:09:14 -08:00
Lucas Wilkinson 0e92298622 [Misc] Delay deprecation of CommonAttentionMetadata properties (#33801) 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
 2026-02-04 08:41:57 -08:00
jiangkuaixue123 87d9a26166 [Bugfix] Fix ubatch wrapper num_tokens calculate (#33694) 
Signed-off-by: jiangkuaixue123 <jiangxiaozhou111@163.com>
 2026-02-04 16:41:45 +00:00
Wentao YeandNick Hill 711edaf0d0 [Perf] Optimize spec decoding + async scheduling, 1.5% Throughput improvement (#33612) 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
 2026-02-04 09:34:32 -05:00
Or Ozeri 8e32690869 [KV Connector][BugFix] scheduler: Delay freeing blocks of aborted async loads (#32255) 
Fixes a not-yet-reported case where it was possible for blocks to be
freed by an abort before an async transfer completed, resulting
in corrupted KV data.

Signed-off-by: Or Ozeri <oro@il.ibm.com>
 2026-02-04 11:16:34 +00:00
zhanqiuhu 4403e3ed4c [Metrics] Add labeled prompt token metrics for P/D disaggregation (#33290) 
Add labeled Prometheus metrics to distinguish where prompt tokens come
from in P/D disaggregated deployments.

In P/D disaggregation, decode instances receive KV cache from prefill instances.
Currently, decode reports inflated prompt throughput because it counts all
prompt tokens as "computed", even though most were transferred.

This PR adds labeled metrics so users can understand actual compute work vs
transferred work:

vllm:prompt_tokens_by_source_total{source="local_compute"}        # Tokens prefilled locally
vllm:prompt_tokens_by_source_total{source="external_kv_transfer"} # Tokens received via KV transfer  
vllm:prompt_tokens_by_source_total{source="local_cache_hit"}      # Tokens from local prefix cache
vllm:prompt_tokens_cached_total                                    # Total cached (local + external, -1 when all 

Signed-off-by: Zhanqiu Hu <zh338@cornell.edu>
 2026-02-04 07:46:48 +00:00
Frank Wang 45f8fd6f97 [Feature] Enable TRITON_ATTN for Batch Invariance (#33688) 
Signed-off-by: frankwang28 <frank.wbb@hotmail.com>
 2026-02-04 13:27:34 +08:00
Isotr0py 02080179a3 [Bugfix] Fix torchrun PP broadcast deadlock with async scheduling (#33701) 
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
 2026-02-04 02:17:37 +00:00
Nick Hill 52ee21021a [BugFix][Spec Decoding] Fix negative accepted tokens metric crash (#33729) 
Signed-off-by: Nick Hill <nickhill123@gmail.com>
 2026-02-03 23:34:41 +00:00
Harry Mellor 61e632aea1 Turn @config into a dataclass_transform (#31541) 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
 2026-02-03 17:40:59 +00:00
Lucas Wilkinson 2267cb1cfd [Attention][FA3] Update FA3 to include new swizzle optimization (#23465) 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
 2026-02-03 08:08:47 -08:00
Kuntai Du fbb3cf6981 [Bugfix][Async][Connector] avoid vllm-side double free during async scheduling + request abort + async KV cache transfer (#33377) 
Signed-off-by: KuntaiDu <kuntai@uchicago.edu>
 2026-02-03 21:50:15 +08:00
Nick Hill 61397891ce [Minor] Some code simplification in scheduler.py (#33597) 
Signed-off-by: Nick Hill <nickhill123@gmail.com>
 2026-02-03 15:00:00 +08:00