Commit Graph
281 Commits
Author SHA1 Message Date
Shu Wang b2c8ce57c6 Fix Flashinfer CUTLASS MOE Allgather (#21963)
Signed-off-by: Shu Wang <shuw@nvidia.com>
2025-08-07 19:18:25 -07:00
WeiQing Chenandycyaw66 4be02a3776 [Bugfix] EPLB load statistics problem (#22167)
Signed-off-by: ycyaw66 <497410282@qq.com>
Signed-off-by: David Chen <530634352@qq.com>
Co-authored-by: ycyaw66 <497410282@qq.com>
2025-08-07 04:07:54 +00:00
Ning Xie 74333ae2f6 [Misc] correct static type check for GroupCoordinator (#21946)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
2025-08-05 03:17:46 -07:00
Ning Xie bd3db7f469 [Misc] log more detailed message for ensure_model_parallel_initialized (#22144)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
2025-08-04 19:36:55 -07:00
Ning Xie 29b97c0995 [Doc] add backend to doc string of initialize_model_parallel (#22142)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
2025-08-04 19:36:20 -07:00
lkchen f4f4e7ef27 [V0 deprecation][P/D] Deprecate v0 KVConnectorBase code (1/2) (#21785)
Signed-off-by: Linkun Chen <github@lkchen.net>
2025-08-04 19:11:33 -07:00
Ning Xie c2e75b3c11 remove duplicate code within cleanup_dist_env_and_memory (#22147)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
2025-08-03 20:03:58 -07:00
David Ben-DavidandDavid Ben-David aefeea0fde [V1] [P/D] Refactor KV Connector Path (#21980)
Signed-off-by: David Ben-David <davidb@pliops.com>
Co-authored-by: David Ben-David <davidb@pliops.com>
2025-08-03 04:03:40 -07:00
Ning Xie 7de45db9a5 [Misc] update doc comment for send (#22026)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
2025-08-03 00:55:20 -07:00
Rui Qiao d331759488 Introduce RayPPCommunicator for ray-based PP (#21660)
Signed-off-by: Rui Qiao <ruisearch42@gmail.com>
2025-08-01 11:50:58 -07:00
wxsm f4135232b9 feat(distributed): add get_required_kvcache_layout class method to kv connector api (#20433)
Signed-off-by: wxsm <wxsms@foxmail.com>
2025-07-30 16:41:51 +00:00
Chenguang Zhengandherotai214 4904e53c32 [Bugfix] SharedStorage Connector for V1 PD multimodal (#21611)
Signed-off-by: fake0fan <645327136@qq.com>
Signed-off-by: herotai214 <herotai214@gmail.com>
Co-authored-by: herotai214 <herotai214@gmail.com>
2025-07-30 09:18:37 -07:00
Calvin Chen e18f085103 skip fusedmoe layer for start_load_kv (#21378)
Signed-off-by: calvin chen <wen.chen@dynamia.ai>
2025-07-28 18:59:44 -07:00
Kuntai Du b18b417fbf Revert "[V1] Exception Handling when Loading KV Cache from Remote Store" (#21778)
Signed-off-by: KuntaiDu <kuntai@uchicago.edu>
2025-07-28 20:15:18 +00:00
Nick Hill 7d44c691b0 [P/D] Log warnings related to prefill KV expiry (#21753)
Signed-off-by: Nick Hill <nhill@redhat.com>
2025-07-28 18:40:53 +00:00
Adelineandliuyumoye 15a72ac478 [V1] Exception Handling when Loading KV Cache from Remote Store (#21534)
Signed-off-by: liuyumoye <adeline_ly2023@outlook.com>
Co-authored-by: liuyumoye <adeline_ly2023@outlook.com>
2025-07-27 20:34:17 -07:00
WeiQing Chen 97d6c30cc9 [BugFix] Fix shared storage connector load kv only load attention layer (#21428)
Signed-off-by: David Chen <530634352@qq.com>
2025-07-26 07:07:40 -07:00
6066284914 [P/D] Support CPU Transfer in NixlConnector (#18293)
Signed-off-by: Juncheng Gu <juncgu@gmail.com>
Signed-off-by: Richard Liu <ricliu@google.com>
Co-authored-by: Richard Liu <39319471+richardsliu@users.noreply.github.com>
Co-authored-by: Richard Liu <ricliu@google.com>
2025-07-24 17:58:42 +01:00
Rui Qiao 1e9ea8e69d [P/D] Move FakeNixlWrapper to test dir (#21328)
Signed-off-by: Rui Qiao <ruisearch42@gmail.com>
2025-07-24 08:53:45 -07:00
Li, Jiang a15a50fc17 [CPU] Enable shared-memory based pipeline parallel for CPU backend (#21289)
Signed-off-by: jiang1.li <jiang1.li@intel.com>
2025-07-21 09:07:08 -07:00
kourosh hakhamaneshi 9f414a12ad [BugFix] Make PD work with Ray (#21072)
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
2025-07-19 08:46:50 -07:00
Rui Qiao 217937221b Elastic Expert Parallel Initial Support (#20775)
Signed-off-by: Rui Qiao <ruisearch42@gmail.com>
2025-07-18 17:46:09 -07:00
Woosuk Kwon 4de7146351 [V0 deprecation] Remove V0 HPU backend (#21131)
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
2025-07-17 16:37:36 -07:00
Zhonghua Deng 8a4e5c5f3c [V1][P/D]Enhance Performance and code readability for P2pNcclConnector (#20906)
Signed-off-by: Abatom <abzhonghua@gmail.com>
2025-07-16 22:13:00 -07:00
Trevor Morrisandmgoin a8593237c0 Add pynccl all-gatherv and reducescatterv (#20154)
Signed-off-by: Trevor Morris <tmorris@nvidia.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
2025-07-11 18:59:23 -07:00
Varun Sundar RabindranathandVarun Sundar Rabindranath 53fa457391 [Misc] Add unit tests for MoE ModularKernel combinations + Profiling utility (#20449)
Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
2025-07-11 07:51:46 -07:00
Nick HillandDavid Ben-David 574ad60db9 [KVConnector] Always call connector clear_metadata() at end of step (#20756)
Signed-off-by: Nick Hill <nhill@redhat.com>
Co-authored-by: David Ben-David <sdavidbd@gmail.com>
2025-07-10 22:37:27 +01:00
Or Ozeri cc876d0f29 [KVConnector] Aggregate finished requests on the scheduler (#19555)
Signed-off-by: Or Ozeri <oro@il.ibm.com>
2025-07-10 09:22:18 +01:00
cd587c93ef [BugFix]: Properly set engine_id when using multi connector (#19487)
Signed-off-by: Nick Hill <nhill@redhat.com>
Co-authored-by: leiyiming <leiyiming@kingsoft.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
2025-07-09 20:32:44 +00:00
Liangliang Maandzhenwei-intel a3e4e85ece [XPU][CI] enhance xpu test support (#20652)
Signed-off-by: Ma, Liangliang <liangliang.ma@intel.com>
Co-authored-by: zhenwei-intel <zhenweiliu@habana.ai>
2025-07-09 16:53:09 +00:00
Nicolò Lucchesi 71d1d75b7a [PD][Nixl] Remote consumer READ timeout for clearing request blocks (#20139)
Signed-off-by: NickLucche <nlucches@redhat.com>
2025-07-08 08:56:40 +01:00
Jee Jee Li 1caca5a589 [Misc] Add SPDX-FileCopyrightText (#20428)
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
2025-07-04 07:40:42 +00:00
Nicolò Lucchesi 8d775dd30a [Misc] Fix Unable to detect current VLLM config. Defaulting to NHD kv cache layout warning (#20400)
Signed-off-by: NickLucche <nlucches@redhat.com>
2025-07-03 14:56:09 -07:00
Ning Xie 1dba2c4ebe [Misc] adjust for ipv6 for mookcacke url parse (#20107)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
2025-07-03 20:27:17 +00:00
Woosuk Kwon 7f280d69c9 [Optimization] Cache sampled token ids in model runner (#20291)
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
2025-07-01 11:01:31 -07:00
Nicolò Lucchesi 650d5dbd04 [Misc] Minor refactor of NIXL background handshake (#20068)
Signed-off-by: NickLucche <nlucches@redhat.com>
2025-07-01 12:40:14 +01:00
Michael Goin be250bbc67 [V1] Only print cudagraph tqdm on rank 0 with is_global_first_rank (#19516)
Signed-off-by: mgoin <mgoin64@gmail.com>
2025-07-01 06:02:09 +00:00
Zhonghua Deng ded1fb635b [Bugfix][V1][P/D]Fix the issue of occasional garbled output for P2pNcclConnector (#20263)
Signed-off-by: Abatom <abzhonghua@gmail.com>
2025-06-30 16:45:14 -07:00
Woosuk Kwon 2863befce3 [Optimization] Use Shared CachedRequestData Instance Across All Requests (#20232)
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
2025-06-30 09:07:50 -07:00
Wentao Ye 4d36693687 [Refactor] Create a function util and cache the results for has_deepgemm, has_deepep, has_pplx (#20187)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
2025-06-28 22:06:38 +00:00
li haoyangandilmarkov 0740e29b66 [Feature] add quick all reduce (#19744)
Signed-off-by: ilmarkov <imarkov@redhat.com>
Signed-off-by: Haoyang Li <Haoyang.Li@amd.com>
Co-authored-by: ilmarkov <imarkov@redhat.com>
2025-06-26 20:54:24 -07:00
Bowen Wang e9fd658a73 [Feature] Expert Parallelism Load Balancer (EPLB) (#18343)
Signed-off-by: Bowen Wang <abmfy@icloud.com>
2025-06-26 15:30:21 -07:00
Nicolò Lucchesi 2582683566 [PD] Skip tp_size exchange with rank0 (#19413)
Signed-off-by: NickLucche <nlucches@redhat.com>
2025-06-25 20:04:39 -07:00
Nick Hill 55c65ab495 [P/D] Avoid stranding blocks in P when aborted in D's waiting queue (#19223)
Signed-off-by: Nick Hill <nhill@redhat.com>
2025-06-25 15:19:44 -07:00
Nick Hill c40692bf9a [Misc] Add parallel state node_count function (#20045)
Signed-off-by: Nick Hill <nhill@redhat.com>
2025-06-25 13:38:53 -07:00
lkchenandNick Hill 91f7d9d0b6 [P/D] Asynchronously do _nixl_handshake (#19836)
Signed-off-by: Linkun Chen <github@lkchen.net>
Signed-off-by: Nick Hill <nhill@redhat.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
2025-06-24 12:46:10 -07:00
lkchen d0132f025d [Misc] Add type alias ReqId and EngineId for better readability (#19880)
Signed-off-by: Linkun Chen <github@lkchen.net>
2025-06-23 12:57:57 -07:00
lkchen 1bcd15edc7 [BugFix][P/D] Fix for cases where _recving_transfers can be cleaned up when *all* transfer done (#19874)
Signed-off-by: Linkun Chen <github@lkchen.net>
2025-06-22 22:41:53 -07:00
Nicolò LucchesiandNick Hill 2ebff5b77c [P/D][NixlConnector] Support tp_size > num_kv_heads deployments (#19691)
Signed-off-by: NickLucche <nlucches@redhat.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
2025-06-22 22:41:50 -07:00
Zhonghua Deng eccdc8318c [V1][P/D] An native implementation of xPyD based on P2P NCCL (#18242)
Signed-off-by: Abatom <abzhonghua@gmail.com>
2025-06-18 06:32:36 +00:00