biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity
9,362 Commits 2 Branches 140 Tags
94e6b2d55f7fd539f01591c7521ec104b25f4e68
Commit Graph

327 Commits

Author SHA1 Message Date
Boyuan Feng
94e6b2d55f Allow users to specify kv cache memory size (#21489) 
Signed-off-by: Boyuan Feng <boyuan@meta.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
 2025-09-11 13:41:07 +00:00
Tao He
e93f4cc9e3 Add the support for the qwen3 next model (a hybrid attention model). (#24526) 
Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
 2025-09-11 15:32:09 +08:00
Nick Hill
e2d8c27f68 [BugFix] Fix pipeline parallel (#24621) 
Signed-off-by: Nick Hill <nhill@redhat.com>
 2025-09-10 23:05:30 -07:00
Michael Goin
fba7856581 [Perf] Warmup FlashInfer attention during startup (#23439) 
Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: Matthew Bonanni <mbonanni001@gmail.com>
 2025-09-10 15:03:17 -07:00
Russell Bryant
37e8182bfe [v1] Add Whisper model support (encoder-decoder) (#21088) 
Signed-off-by: Russell Bryant <rbryant@redhat.com>
Co-authored-by: NickLucche <nlucches@redhat.com>
 2025-09-10 13:53:35 -07:00
Nick Hill
f4f1a8df22 [BugFix] Ensure integrity of reused CPU tensors during async scheduling (#24527) 
Signed-off-by: Nick Hill <nhill@redhat.com>
Co-authored-by: guoze.lin <guozelin@tencent.com>
 2025-09-10 21:15:14 +08:00
Lucas Wilkinson
0ae43dbf8c [Attention] add DCP support for FLASH_ATTN_MLA backend (#24453) 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>
 2025-09-10 17:19:26 +08:00
Micah Williamson
1c63a16b65 [Core] Run garbage collector after CUDA graph capture to fix throughput regression (#24128) 
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Co-authored-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
 2025-09-09 10:38:10 -04:00
Woosuk Kwon
2e5d21378d Skip MM Encoder for non-first PP ranks (#24387) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-09-07 09:38:35 -07:00
youkaichao
558f0907dc [attention][DCP] use AttentionImpl.need_to_return_lse_for_decode (#24372) 
Signed-off-by: youkaichao <youkaichao@gmail.com>
 2025-09-07 01:18:59 +00:00
Bangsheng Tang
848562bd49 break execute_model in gpu_model_runner into sub-functions for custom scopes (#24265) 
Co-authored-by: Bangsheng Tang <bangsheng@meta.com>
 2025-09-06 14:02:47 -07:00
Andrew Sansom
305a1cc0d2 refactor: Turn GPUModelRunner.inputs_embeds to a CpuGpuBuffer (#24345) 
Signed-off-by: Andrew Sansom <andrew@protopia.ai>
 2025-09-05 23:01:23 -07:00
yzds
ac201a0eaf [Feature] Support Decode Context Parallel (DCP) for MLA (#23734) 
Signed-off-by: hongchao <hongchao@msh.team>
Signed-off-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: hongchao <hongchao@msh.team>
Co-authored-by: youkaichao <youkaichao@gmail.com>
 2025-09-06 13:24:05 +08:00
Didier Durand
35bf193864 [Doc]: fix typos in Python comments (#24294) 
Signed-off-by: Didier Durand <durand.didier@gmail.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
 2025-09-05 19:41:12 -07:00
Benjamin Chislett
cee182b297 [Perf][V1] Fully overlap model execution (#23569) 
Signed-off-by: Benjamin Chislett <benjamin.chislett@centml.ai>
 2025-09-05 18:20:17 -07:00
liuzhenwei
e599e2c65e [XPU][P/D] Add XPU support in NixlConnector (#22436) 
Signed-off-by: zhenwei <zhenwei.liu@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
 2025-09-04 21:03:12 -07:00
co63oc
1bd007f234 fix some typos (#24071) 
Signed-off-by: co63oc <co63oc@users.noreply.github.com>
 2025-09-02 20:44:50 -07:00
Russell Bryant
e32a0e8678 Upgrade xgrammar to 0.1.23 (#22988) 
Signed-off-by: Russell Bryant <rbryant@redhat.com>
 2025-09-03 02:32:59 +00:00
Maximilien de Bayser
d59c986444 Remove runtime checks based on pooling params (#24051) 
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
 2025-09-02 11:54:37 +08:00
Woosuk Kwon
b55713683c [Misc] Move fast prefill logic to separate method (#24013) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-09-01 05:40:38 +00:00
Woosuk Kwon
8c742a66d1 [Misc] Avoid redundant copy for encoder-only models (#24012) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-09-01 04:02:43 +00:00
Didier Durand
9701352e4b [Doc]: fix typos in Python comments (#24001) 
Signed-off-by: Didier Durand <durand.didier@gmail.com>
 2025-08-31 08:21:59 +00:00
Andy Lo
038e9be4eb [LoRA] Much faster startup when LoRA is enabled (#23777) 
Signed-off-by: Andy Lo <andy@mistral.ai>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
 2025-08-30 15:37:39 +00:00
wang.yuqi
d9e00dbd1f [Performance] V1 Classify Models E2E Performance Optimization (#23541) 
Signed-off-by: wang.yuqi <noooop@126.com>
 2025-08-29 03:12:32 -07:00
Yong Hoon Shin
cb293f6a79 [V1] Enable prefill optimization for Gemma3n (#22628) 
Signed-off-by: Yong Hoon Shin <yhshin@meta.com>
 2025-08-28 14:54:30 -07:00
Wentao Ye
321938e9ac [Feature] Add VLLM_DISABLE_PAD_FOR_CUDAGRAPH to Avoid Hang Issue (#23595) 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
 2025-08-27 21:52:24 +00:00
Hyogeun Oh (오효근)
4e4d017b6f [Docs] Fix warnings in mkdocs build (continued) (#23743) 
Signed-off-by: Zerohertz <ohg3417@gmail.com>
Signed-off-by: Hyogeun Oh (오효근) <ohg3417@gmail.com>
 2025-08-27 17:17:29 +00:00
Cyrus Leung
52883ed084 [Model] Merge SupportsMultiModalWithRawInput with SupportsMultiModal (#23749) 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
 2025-08-27 10:01:50 -07:00
Woosuk Kwon
04ff1e43fb [Misc] Move CpuGpuBuffer to vllm/v1/utils.py (#23728) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-08-27 03:25:00 -07:00
Cyrus Leung
69244e67e6 [Core] Use key-only cache for BaseMultiModalProcessor (#23018) 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
 2025-08-27 14:19:13 +08:00
Li, Jiang
9b0187003e [Bugfix] Fix cuda event usage with CPU model runner (#23643) 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
 2025-08-26 17:10:42 +00:00
Chen Zhang
2b4fc9bd9b Support FlashAttention Backend for Hybrid SSM Models (#23299) 
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
 2025-08-26 12:41:52 +00:00
Zijing Liu
b395b3b0a3 [Disagg][Perf] Use CUDA event sync instead of blocking tolist to avoid unintentional copy ops blocking across different CUDA streams, improving disagg TTIT/TTFT (#22760) 
Signed-off-by: Zijing Liu <liuzijing2014@gmail.com>
Signed-off-by: Zijing Liu <liuzijing2014@users.noreply.github.com>
 2025-08-25 21:06:00 -07:00
Woosuk Kwon
0ff902f3b4 [Refactor] Refactor persistent buffers with CpuGpuBuffer (#23515) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
 2025-08-25 08:44:48 -07:00
Ayush Satyam
5c4b6e66fe [Attention] Unify mamba and attention backend selection (#23171) 
Signed-off-by: Ayush Satyam <ayushsatyam146@gmail.com>
 2025-08-25 09:09:36 +00:00
Chenguang Zheng
d765cf01fe [Core][Multimodal] Track encode cache entries by mm_hash and enable embedding sharing between requests (#22711) 
Signed-off-by: knlnguyen1802 <knlnguyen1802@gmail.com>
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: knlnguyen1802 <knlnguyen1802@gmail.com>
Co-authored-by: Roger Wang <hey@rogerw.io>
 2025-08-25 00:41:17 -07:00
Woosuk Kwon
ad78868450 [Misc] Remove unused slot_mapping buffer (#23502) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-08-24 14:03:36 -07:00
Nick Hill
c80c53a30f [BugFix] Fix batch updates for pooling models (#23398) 
Signed-off-by: Nick Hill <nhill@redhat.com>
 2025-08-23 08:20:41 +08:00
Woosuk Kwon
808d2e9aa0 [Misc] Move M-RoPE init logic to _init_mrope_positions (#23422) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-08-22 03:07:22 -07:00
Chen Zhang
17373dcd93 [Attention] Refactor AttentionMetadata Preparation for Encoder-only Models (#23154) 
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
 2025-08-22 05:05:59 +00:00
Nick Hill
603fbbbce0 [Misc] Misc code cleanup/simplification (#23304) 
Signed-off-by: Nick Hill <nhill@redhat.com>
 2025-08-21 17:22:55 +00:00
wang.yuqi
d70a16625d [Performance] V1 Pooling Models E2E Performance Optimization (#23162) 
Signed-off-by: wang.yuqi <noooop@126.com>
 2025-08-21 13:26:09 +00:00
Cyrus Leung
0c6e40bbaa [Refactor] Simplify code for MM budget (#23310) 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
 2025-08-21 08:00:16 +00:00
Woosuk Kwon
b029de9902 [Optimization] Make new_block_ids None if empty (#23262) 
Signed-off-by: Woosuk Kwon <woosuk@thinkingmachines.ai>
 2025-08-20 18:25:56 -07:00
rongfu.leng
4fbda0b20c [Feature] use --eplb_config to set eplb param (#20562) 
Signed-off-by: rongfu.leng <rongfu.leng@daocloud.io>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: rongfu.leng <lenronfu@gmail.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
 2025-08-20 14:07:28 -07:00
Woosuk Kwon
d6d13bd49e [Misc] Add max_seq_len to CommonAttentionMetadata (#23216) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-08-20 09:05:29 -07:00
Woosuk Kwon
40f26734b9 [Misc] Fix seq_lens for graph capture (#23175) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-08-19 03:58:16 -07:00
Woosuk Kwon
21bcc8263f [Misc] Avoid accessing req_ids inside a loop (#23159) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-08-19 09:39:38 +00:00
Woosuk Kwon
c9b38be8aa [Spec Decode] Make propose_draft_token_ids non-blocking for lower TTFT (#23041) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-08-18 17:20:38 -07:00
Woosuk Kwon
0dd3f4f5ab [Misc] Minor refactoring for prepare_inputs (#23116) 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
 2025-08-18 16:58:05 -07:00