Commit Graph
62 Commits
Author SHA1 Message Date
Ilia Yastrebov 41848caa63 [NIXL] Use int32 array for indices to avoid intermediate conversion (#51952)
Signed-off-by: Ilia Yastrebov <iyastrebov@nvidia.com>
2026-09-02 16:30:16 +02:00
Itay Alroy a2257f95b7 [Elastic EP] Reduce eager-mode reconfiguration downtime (#51885)
Signed-off-by: Itay Alroy <ialroy@nvidia.com>
2026-08-19 02:13:06 +00:00
Nick HillandClaude Opus 4.8 615d4cfade [Core] Check for GPU<->CPU syncs during CI (#43107)
Signed-off-by: Nick Hill <nickhill123@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-08-14 17:42:05 -07:00
Julien Debache 6e502b62cb fix and test EPLB balancedness calculation (#51813)
Signed-off-by: jdebache <jdebache@nvidia.com>
2026-08-12 20:27:22 -07:00
2279575cd9 [AMD][Bugfix][EPLB] Fix elastic EP scaling accuracy on ROCm (#47206)
Signed-off-by: okorzh <okorzh@amd.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>
2026-07-24 13:35:27 -07:00
Andreas Karatzas 16aca639b7 [ROCm] Upgrade NIXL and UCX (#49251)
Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
2026-07-21 22:30:02 -05:00
Kunshang JiandHarry Mellor 07f9baf756 Revert "[Platform] Replace torch.cuda.Event with torch.Event (#47140)" (#47668)
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
2026-07-06 14:18:33 +01:00
Kunshang JiandHarry Mellor e840f0d3f5 [Platform] Replace torch.cuda.Event with torch.Event (#47140)
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
2026-06-30 08:39:59 -07:00
Ilya MarkovandMarkov Ilya a2abce646f [EPLB] Mask padding in EPLB load recording (#38128)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Markov Ilya <markovilya19@gmail.com>
Co-authored-by: Markov Ilya <markovilya19@gmail.com>
2026-06-28 19:43:58 -07:00
Ilya MarkovandMarkov Ilya ac614587f5 [EPLB] Enable nixl eplb communicator for elastic ep (#45013)
Signed-off-by: Markov Ilya <markovilya197@gmail.com>
Signed-off-by: Markov Ilya <markovilya19@gmail.com>
Co-authored-by: Markov Ilya <markovilya19@gmail.com>
2026-06-22 10:54:08 -07:00
Ilya Markov 6471ec75bd [EPLB] Reject NCCL-based EPLB communicators with async EPLB (#44978)
Signed-off-by: Markov Ilya <markovilya197@gmail.com>
2026-06-10 19:51:27 +00:00
bnellnm dc68bd8c41 [MoE Refactor] FusedMoE/MoERunner inversion refactor (#41184)
Signed-off-by: Bill Nell <bnell@redhat.com>
2026-06-08 10:42:58 -04:00
4f423bd5bc [EPLB] Nixl communicator optimization. Zero-copy transfers (#41633)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Markov Ilya <markovilya19@gmail.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Markov Ilya <markovilya19@gmail.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
2026-06-04 03:40:34 +00:00
Wei ZhaoandWei Zhao 2427094152 [Feature] Support EPLB for DeepSeek v4 Mega Moe (#43339)
Signed-off-by: wzhao18 <wzhao18.sz@gmail.com>
Co-authored-by: Wei Zhao (Engrg-Hardware 1) <weizha@login-lyris01.lyris.clusters.nvidia.com>
2026-06-02 10:56:44 -07:00
SandishKumarHNandSandishKumarHN 977703aa94 [RFC][EPLB][#32028] Remove dead torch.accelerator.synchronize() from sync path (#40733)
Signed-off-by: SandishKumarHN <3078999+SandishKumarHN@users.noreply.github.com>
Co-authored-by: SandishKumarHN <3078999+SandishKumarHN@users.noreply.github.com>
2026-05-22 15:19:24 -04:00
bnellnmandRobert Shaw d9b4990783 [MoE Refactor] EPLB refactoring for FusedMoE (#41055)
Signed-off-by: Bill Nell <bnell@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
2026-05-12 14:16:31 -04:00
Wentao Ye f80aa53c9d [Refactor] Nixl util using lazy init (#41392)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
2026-05-09 17:46:52 -04:00
Nicolò Lucchesi 9d6500b89d [Misc] Delay EPLB Nixl import until needed (#41805)
Signed-off-by: NickLucche <nlucches@redhat.com>
2026-05-07 09:43:07 -07:00
a7fb008510 [EPLB] Optimize memory overhead in Nixl communicator (#40013)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Markov Ilya <markovilya19@gmail.com>
Co-authored-by: Markov Ilya <markovilya19@gmail.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
2026-04-30 07:46:49 -07:00
Sage Moore 62b1bbe470 [EPLB] Remove asyncio infrastructure from Async EPLB (#40730)
Signed-off-by: Sage Moore <sage@neuralmagic.com>
2026-04-24 00:21:15 +00:00
Sage Moore 3173441b0f [EPLB] Consolidate is_unchanged/is_received_locally into TransferMetadata (#37341)
Signed-off-by: Sage Moore <sage@neuralmagic.com>
2026-04-20 21:12:42 +00:00
Sage MooreandTyler Michael Smith 3461c8b027 [EPLB] Refactor Async EPLB synchronization logic (#37601)
Signed-off-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
2026-04-20 17:05:41 +00:00
Ilya Markov 50dd4cb427 [EPLB] Add nixl-based eplb communicator (#36276)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Markov Ilya <markovilya19@gmail.com>
2026-04-20 10:24:23 +00:00
Ilya MarkovandMarkov Ilya abdbb68386 [EPLB] Add alternative communication for EPLB weight exchange (#33176)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Markov Ilya <markovilya19@gmail.com>
Co-authored-by: Markov Ilya <markovilya19@gmail.com>
2026-03-31 08:17:12 -04:00
Ilya Markov 12701e8af2 [EPLB] Optmize eplb mapping and record in router for prefill (#36261)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
2026-03-30 19:48:33 +00:00
Sage Moore 497e234d38 [EPLB] Cleanup the transfer logic for the various eplb maps (#34520)
Signed-off-by: Sage Moore <sagmoore@redhat.com>
Signed-off-by: Sage Moore <sage@neuralmagic.com>
2026-03-27 10:18:46 +01:00
Ilya Markov 14acf429ac [EPLB] Remove main waits in case of slow EPLB (#36271)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
2026-03-24 11:50:44 +00:00
Sage MooreandTyler Michael Smith c32a58cc2a [EPLB] Simplify EPLB rearrange by only returning one map (#36267)
Signed-off-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
2026-03-18 20:34:00 -04:00
Kunshang Ji 53ec16a705 [Hardware] Replace torch.cuda.device_count/current_device/set_device API (#36145)
Signed-off-by: Kunshang Ji <jikunshang95@gmail.com>
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
2026-03-12 07:57:47 -07:00
Jiayi YanandHarry Mellor 6a895197fa [Bugfix][CI] fix typos (#34934)
Signed-off-by: 1195343015 <1195343015@qq.com>
Signed-off-by: Jiayi Yan <66017932+1195343015@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
2026-03-05 17:05:46 +00:00
Kunshang Ji 66a2209645 [Hardware] Replace torch.cuda.synchronize() api with torch.accelerator.synchronize (#36085)
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
2026-03-05 10:36:39 +00:00
dea268336f [1/N] Elastic EP Milestone 2 (#34861)
Signed-off-by: Yongji Wu <wuyongji317@gmail.com>
Signed-off-by: Itay Alroy <ialroy@nvidia.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Signed-off-by: Ron Tourgeman <rtourgeman@nvidia.com>
Co-authored-by: Yongji Wu <wuyongji317@gmail.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Ron Tourgeman <rtourgeman@nvidia.com>
2026-02-28 04:46:42 +00:00
Tyler Michael SmithandClaude Opus 4.6 eb19955c37 [WideEP] Remove pplx all2all backend (#33724)
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-26 14:30:10 -08:00
Ilya Markov bb2fc8b5e7 [BugFix] Fix async EPLB hang with DeepEP LL all2all backend (#32860)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
2026-02-10 22:34:47 +00:00
67132945bb [Perf] Move eplb rebalance algo to async thread (#30888)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
2026-02-10 22:19:10 +00:00
Sage MooreandRobert Shaw ce498a6d61 Change the type signature of MixtureOfExperts.expert_weights to MutableSequence[Sequence[Tensor]] (#33573)
Signed-off-by: Sage Moore <sagmoore@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
2026-02-04 17:02:46 -05:00
Ilya Markov d09135fbd0 [BugFix] Async Eplb fix potential race condition (#32881)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
2026-01-29 10:31:40 +00:00
bnellnm 327a02d8db [MoE Refactor] Separate Router into OO Classes (#30623)
Signed-off-by: Bill Nell <bnell@redhat.com>
2026-01-18 11:40:49 -05:00
Ilya MarkovandTyler Michael Smith c9a533079c [EPLB][BugFix]Possible deadlock fix (#32418)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
2026-01-16 09:11:01 -05:00
Sage Moore 6beef12b9b [EPLB][Cleanup] Remove is_async_enabled from EplbModelState (#32050)
Signed-off-by: Sage Moore <sage@neuralmagic.com>
2026-01-13 18:19:03 +00:00
Ilya Markov 1eb61ab34b [Refactor] EPLB rebalance algo to NumPy (#30697)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
2026-01-12 18:13:23 +00:00
Ilya MarkovandTyler Michael Smith 6170d47d22 [EPLB] Optimize EPLB with numpy (#29499)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
2026-01-07 15:21:35 -05:00
Tyler Michael Smith fe8a9fbd2e [Bugfix] Fix EPLB state logging error (#31455)
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
2026-01-05 04:06:28 +00:00
97f2f160fd [ROCm][CI] Add "Qwen3-Next-80B-A3B-Instruct MTP Async EPLB Accuracy Test" Back Into AMD CI (#30590)
Signed-off-by: David Chen <530634352@qq.com>
Signed-off-by: WeiQing Chen <40507679+david6666666@users.noreply.github.com>
Signed-off-by: Micah Williamson <micah.williamson@amd.com>
Co-authored-by: WeiQing Chen <40507679+david6666666@users.noreply.github.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
2025-12-14 06:56:26 +00:00
Sage Moore b4054c8ab4 Revert "[CI] Add Async Eplb nightly CI tests (#29385)" (#30431) 2025-12-11 00:48:35 +00:00
1119f6e47a Abstract eplb algo (#26471)
Signed-off-by: Che Ruan <cr623@ic.ac.uk>
Signed-off-by: mengxingkongzhouhan <117415539+mengxingkongzhouhan@users.noreply.github.com>
Signed-off-by: Mercykid-bash <ruanche0218@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Che Ruan <cr623@ic.ac.uk>
Co-authored-by: mengxingkongzhouhan <117415539+mengxingkongzhouhan@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
2025-12-04 19:09:09 +00:00
WeiQing ChenandCyrus Leung 7fe9c1a223 [CI] Add Async Eplb nightly CI tests (#29385)
Signed-off-by: David Chen <530634352@qq.com>
Signed-off-by: WeiQing Chen <40507679+david6666666@users.noreply.github.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
2025-12-03 09:51:08 +00:00
WeiQing ChenandSunChenxiang123 2601f18a82 [EPLB] Optimize EPLB for Async Rearrange Experts (#22179)
Signed-off-by: David Chen <530634352@qq.com>
Co-authored-by: SunChenxiang123 <1291824390@qq.com>
2025-11-24 09:08:29 -05:00
Sage Moore 798c7bebca [EPLB] Refactor balance_packing to use numpy and optimize GPU-CPU transfers in EPLB (#28369)
Signed-off-by: Sage Moore <sage@neuralmagic.com>
2025-11-11 00:19:51 -08:00
e50c454672 [BugFix] Support EP/DP + EPLB with MTP (#25311)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
2025-11-05 15:22:17 +00:00