Commit Graph
63 Commits
Author SHA1 Message Date
Liangsheng Yin f35ec18d8b [Test] Remove more unit tests that mirror implementation or never run in CI (#41297) 2026-09-25 19:22:14 -07:00
Liangsheng Yin 1f6ce4b068 [Test] Remove unit tests that only mirror implementation or never run in CI (#41286) 2026-09-25 16:15:02 -07:00
Jackey HuaandClaude Fable 5.1 5bb24e399d [Quant] ModelOpt mixed precision: dispatch block-FP8 MoE experts and derive the block size (#38726)
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-24 12:30:01 -04:00
Cheng Wan acac4dd9d9 [Refactor] Clean up parallel runtime comments (#40632) 2026-09-21 14:32:22 -07:00
Cheng Wan 73f071db52 Deprecate the parallel getters the context answers, and ratchet them shut (#40342) 2026-09-21 12:25:32 -07:00
Cheng Wan 2d0e94e3a3 Check the topology identities where the layout is written, and build at the published widths (#40340) 2026-09-21 12:22:59 -07:00
Cheng Wan afe71f4b9e Read process groups through the runtime context (#40068) 2026-09-18 17:40:32 -07:00
Yang Liuandyangliu991 276663a79d [model-loader] Split weight loading from postprocessing (#34981)
Co-authored-by: yangliu991 <yangliu991@fb.com>
2026-09-14 17:10:19 -07:00
cebca698e2 [Qwen3.8] Enable NVIDIA NVFP4 on DGX Spark with file-backed PLE and PDL router fix (#39126)
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: rdxa <rdxa@rdxa-int-spark-01.yvb.moe>
Co-authored-by: Yangmin Li <yangminl@nvidia.com>
Co-authored-by: Manrique <nanomlm@gmail.com>
Co-authored-by: yhyang201 <yhyang201@gmail.com>
2026-09-13 16:23:41 +08:00
Aurick Qiao c389dd0863 Fix RunAI object-storage checkpoint index filtering (#38988) 2026-09-12 12:25:57 +08:00
fae8cd84cb Support NemotronH_Omni_Reasoning_V3 in SGLang (#35599)
Signed-off-by: Ryan Stewart <rystewart@nvidia.com>
Signed-off-by: rystewart-nvidia <rystewart@nvidia.com>
Co-authored-by: elvischenv <219235043+elvischenv@users.noreply.github.com>
Co-authored-by: Po-Han Huang (NVIDIA) <53919306+nvpohanh@users.noreply.github.com>
2026-09-10 16:57:22 -07:00
sglang-botandsglang-bot 6252993afe chore: update CI test est_time values (#38238)
Co-authored-by: sglang-bot <sglang-bot@users.noreply.github.com>
2026-09-06 17:49:41 -07:00
Po-Han Huang (NVIDIA) 991368d880 Fix Muse Glimmer ModelOpt mixed weight mapping (#37510) 2026-09-04 20:53:55 -07:00
92a4d8b5ee Clean logging under --weight-loader-prefetch-checkpoints (#33930)
Co-authored-by: Brayden Zhong <brayden@radixark.ai>
Co-authored-by: Mohammad Miadh Angkad <176301910+mmangkad@users.noreply.github.com>
2026-09-04 20:05:53 -07:00
55bf3380e0 Support Hy4-preview (#36805)
Co-authored-by: BBuf <1182563586@qq.com>
Co-authored-by: alphabetc1 <2508695655@qq.com>
2026-09-04 18:03:49 -07:00
4349538c02 [model] add cosmos3 reasoner to llm only inference (#33572)
Signed-off-by: joeltg <joel@reflection.ai>
Signed-off-by: Joe Rowell <joe@poolside.ai>
Co-authored-by: Dawid Majchrowski <dmajchrowski@nvidia.com>
Co-authored-by: Kedi Wu <kediw@nvidia.com>
Co-authored-by: Kedi Wu <31940276+kediwu0331@users.noreply.github.com>
Co-authored-by: Joel Gustafson <joelgustafson@protonmail.com>
2026-09-04 22:11:58 +08:00
Alex NailsandAlison Shao 28262c20df [CI][RFC] Replace black-jupyter with ruff-format (#37210)
Co-authored-by: Alison Shao <a.shao@wustl.edu>
2026-09-02 19:46:08 -07:00
Mick 0dd66def7c [chore] harden checkpoint quantization metadata parsing (#36922) 2026-09-03 09:27:22 +08:00
Ziang Li 9a85473a89 [FlashInfer v0.6.18] add FlashInfer CuTe DSL NVFP4 W4A16 mode (#35120) 2026-08-31 18:47:30 -07:00
6580d5cd9a weight cache: key daemon paths by GPU UUID (#36101)
Co-authored-by: siyu <liusy58@linux.alibaba.com>
Co-authored-by: Alex Nails <alex.nails@radixark.ai>
2026-08-31 01:44:45 -07:00
Pavan Sivaram GirijalaandYanbingJiang 4dc7dc8518 [Fix] Transformers-fallback (GPT-NeoX) + KV pool config (DeepSeek-VL2) (#35244)
Co-authored-by: YanbingJiang <yanbing.jiang@intel.com>
2026-08-31 11:05:50 +08:00
Joel Gustafsonandmmangkad 0a585d5bb1 [BugFix] Allow model_loader_extra_config with remote_instance + modelexpress backend (#34639)
Signed-off-by: joeltg <joel@reflection.ai>
Co-authored-by: mmangkad <176301910+mmangkad@users.noreply.github.com>
2026-08-29 00:43:10 -07:00
Cheng WanandClaude Opus 5 ca1d7ed8e6 config: a parallel leaf with no live counterpart is read bare (#36620)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
2026-08-27 12:56:11 -07:00
Schwinn Saereesitthipitak 08315c56df [Fix] Account resident weight memory in KV sizing (#34053)
Signed-off-by: Schwinn Saereesitthipitak <schwinns@nvidia.com>
2026-08-27 02:43:04 -07:00
Xun Sun adcf73d7f7 [Weight Cache] Enhance test and support EPLB (#36198) 2026-08-26 23:11:53 -07:00
Cheng Wan 8005df61d3 config: spell the parallel config tier at the call site (#36250) 2026-08-26 03:00:28 -07:00
Xun Sunandsiyu a90d770c40 [Weight Cache] Support static DP/EP layouts (#33684)
Co-authored-by: siyu <liusy58@linux.alibaba.com>
2026-08-23 19:57:52 -07:00
siyuandliusy58 cb10ca16dd [FEAT] Weight Daemon abstraction (#33279)
Co-authored-by: liusy58 <liusy58@smail.nju.edu.cn>
2026-08-22 02:13:46 -07:00
YAMY 5f12839591 [Fix] Support Kimi-K3 ModelOpt mixed NVFP4/FP8 checkpoint (#35077) 2026-08-19 08:13:45 -07:00
Mick ef490853bb quant: extract shared checkpoint quant metadata resolver (#35172) 2026-08-19 08:26:41 +08:00
6b94d39f13 [Model Loading] Overlap checkpoint staging with CUDA graph capture during startup (#32017)
Co-authored-by: Wenhui Zhu <wzhu59@asu.edu>
Co-authored-by: Alex Nails <alex.nails@radixark.ai>
2026-08-13 12:26:25 -07:00
Mohammad Miadh Angkad 8c5d5f75bf Fix DSV4 DSpark shared expert loading (#33312) 2026-08-11 07:36:50 +08:00
Cheng Wan 63833f8034 config: business code no longer reads the published ServerArgs (#34081) 2026-08-09 14:44:08 -07:00
Ziang LiandBrayden Zhong 4ad990ba7d [ModelOpt FP4] Support online MoE weight quantization (#33115)
Co-authored-by: Brayden Zhong <b8zhong@uwaterloo.ca>
2026-08-06 11:01:55 -07:00
Mohammad Miadh Angkad 65d5a0ec25 Support ModelOpt MXFP8 checkpoints (#32538) 2026-08-05 17:44:32 -07:00
Cheng Wan 47d8b5b749 config: route parallel config-leaf reads through get_parallel() (#33170)
The parallel namespace joins the accessor migration: 106 config-leaf reads
(enable_dp_lm_head, enable_dp_attention, pp_async_batch_depth, dp_size,
ep_join_rank_offset, dwdp_size, ...) flip from get_server_args()/
self.server_args to get_parallel(), which serves config leaves from the
published parallel bag via __getattr__.

- ParallelContext.__getattr__ is restructured to stay dynamo-traceable
  (object.__getattribute__ graph-breaks): gate helpers such as
  enable_moe_dense_fully_dp() run inside compiled model forwards. A
  fullgraph regression test pins the pattern.
- The five live-shadowed topology sizes (tp/pp/dcp/attn_cp/moe_dp_size)
  keep their server_args reads: the live @property wins on the accessor,
  and conditionally-initialized groups would fail loud at unconditional
  call sites.
- Elastic-EP scale writers (ep_size/dp_size x4 in model_runner) reroute
  to get_context().override together with their remaining instance
  readers (expert_location gpus-per-node paths); the ServerArgs.override
  ratchet drops 39 -> 35.
- The expert placement helpers (compute_logical_to_rank_dispatch_
  physical_map, _compute_logical_to_all_physical_map,
  _prefer_same_node_experts) now read everything from the bags and drop
  their server_args parameter; their unit tests publish the config they
  need instead of stubbing it.
2026-08-01 08:58:39 -07:00
Cheng Wan 55b6769b0e config: read resolved config via namespace accessors (#33013) 2026-07-31 15:06:59 -07:00
Nan Jiang 89f4a80c1f Support fastsafetensors no-GDS loading and page-cache release (#31859) 2026-07-31 23:12:32 +08:00
Danila ShtanandDanila Shtan 5f9b0db18c Fix async loading of RunAI-streamed tensors (#32896)
Co-authored-by: Danila Shtan <dan@nebius.com>
2026-07-31 21:46:33 +08:00
Trevor Morris a6221d776f feat: Support nvidia/MiniMax-M3-NVFP4 (#31989) 2026-07-30 14:32:03 -07:00
Broduker b61cb5f9de Fix DeepSeek V4 loading with RunAI Model Streamer. (#30240) 2026-07-30 23:03:34 +08:00
Cheng WanandShu Wang 659d349b61 [core/loader] Add presharded load format (#24256)
Co-authored-by: Shu Wang <shuwanguc@google.com>
2026-07-25 13:03:39 -07:00
f9c14e6bd4 [FEAT] Support fast engine recovery through weight cache (#27139)
Signed-off-by: Michael Qiu <qiudayu.qdy@antgroup.com>
Co-authored-by: liusy58 <liusy58@linux.alibaba.com>
Co-authored-by: Alex Nails <alex.nails@radixark.ai>
2026-07-25 14:31:21 +08:00
Liangsheng Yin f4f15162bc [Fix] Fail fast when a safetensors index references missing shard files (#32279) 2026-07-24 02:18:56 -07:00
Fardin Hoque b54adced46 [Quant] Support NVFP4_AWQ checkpoints in ModelOpt FP4 path (#31825) 2026-07-22 08:46:10 +08:00
Liangsheng Yin a36c873147 [misc] Remove unit test cases that fail the admission criteria (round 2) (#30703) 2026-07-09 16:35:59 -07:00
Cheng Wan c02b032da0 [fix] Repoint the prefetch-dispatch test at the loader's current config binding (#30631) 2026-07-09 02:21:04 -07:00
Mohammad Miadh Angkad 666a09fe2a Disable multi-threaded load by default when prefetch is on (#30146) 2026-07-09 00:28:53 -07:00
Mohammad Miadh Angkad b1942fc3ea [Model] Support Qwen3.6 ModelOpt mixed NVFP4 (#27906) 2026-07-05 21:31:15 -07:00
Mohammad Miadh Angkad 6c018eb4d1 Fix bounded checkpoint prefetching and buffered drop-cache handling (#29156) 2026-06-29 21:49:16 +00:00