Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 4128f1ffe2 [SKILLS] Tiny upgrade diffusion skills (#24273) Xiaoyu Zhang 2026-05-02 22:04:05 +08:00
• b712dd48fe [codex] diffusion: enable group norm silu fuse by default (#23148) Xiaoyu Zhang 2026-05-02 20:55:51 +08:00
• 1360848ee1 Optimize large GroupNorm SiLU apply (#23938) Xiaoyu Zhang 2026-05-02 20:54:46 +08:00
• 83bf5d6869 [NPU]TP Communications compression For Qwen3 models for NPU (#20520) egvenediktovandronnie_zheng 2026-05-02 14:29:11 +03:00
• ebbaab5597 [NPU] Add GitHub test summary and deduplicate test code. Part 1 (#23835) 5 people 2026-05-02 14:18:18 +03:00
• 3259a2c789 Encode routed_experts in the detokenizer, off the tokenizer hot path (#24263) 3 people 2026-05-02 02:44:32 -07:00
• 589f90b368 [diffusion] chore: use lmsys as org for modelopt checkpoints (#23924) Xiaoyu Zhang 2026-05-02 17:18:58 +08:00
• f3dbadb82b fix: accept 0-indexed safetensors shard names in CI weight validator (#24237) Alison Shao 2026-05-02 00:58:15 -07:00
• 2e72a36420 [CI] Restore SMG e2e on 2-gpu-h100 / 4-gpu-h100 runners (#24222) Kangyan-ZhouandClaude Opus 4.7 2026-05-01 23:55:20 -07:00
• b939d5410f [AMD] enable sdma for moriep unittest (#24259) billishyahao 2026-05-02 14:34:53 +08:00
• 5ec3b26799 [diffusion] model: support JoyAI-Image-Edit (#22625) Clayandchengyusong1 2026-05-02 14:08:57 +08:00
• cd27baaffd [ci][cu13] Bump torch_memory_saver to 0.0.9.post1; restore manual tests (#23182) Kangyan-ZhouandClaude Opus 4.7 2026-05-01 22:50:38 -07:00
• 674a80d5d8 [CI] Add 1-gpu-h100-h200 to rerun-test runner_label choices (#24264) Kangyan-ZhouandClaude Opus 4.7 2026-05-01 22:45:48 -07:00
• b7d4647568 [diffusion] CI: change ground truth repo (#24219) Mick 2026-05-02 12:25:40 +08:00
• 63f225ca2e [session] fix mamba pool leak in StreamingSession.release_session + plumb idle leak check (#23496) Sam Shleifer 2026-05-01 23:38:08 -04:00
• d41e8c459d Support RunAI loading for quantized checkpoints (#23850) 3 people 2026-05-01 23:11:40 -04:00
• 321298da75 [SKILL] Upgrade sglang profile and auto_benchmark skills (#24250) Xiaoyu Zhang 2026-05-02 10:12:47 +08:00
• 4c2ed9a254 Flux2 nvfp4 quantization correctness on Blackwell (B200) (#23625) Johnsonms 2026-05-01 18:57:35 -07:00
• bfccc8e504 Allow configuring NIXL backend parameters from env (#24169) Aurick Qiao 2026-05-01 18:30:43 -07:00
• 193b977572 [diffusion] chore: clean scheduler (#24229) Mick 2026-05-02 09:30:06 +08:00
• cb8fbd53fc Reserve slot 0 as padding in all req pools (#24243) Liangsheng Yin 2026-05-01 16:41:36 -07:00
• b47fab6f5d [bugfix] Support MIXED forward mode in TBO splitter for DP attention (#24241) Cheng Wan 2026-05-01 16:01:23 -07:00
• 05de73efd1 [core/model] Use explicit model arch for Llama4 attention backend auto-selection (#24232) Lucia Fang 2026-05-01 15:49:30 -07:00
• 8a530468fd [Bug] Size mamba mappings from req pool, not mamba pool (#24244) Liangsheng Yin 2026-05-01 15:45:20 -07:00
• 79bc2505a5 [Bug Fix] Resolve EAGLE cuda graph IMA under PD + DP + MTP with GLM-5.1 (#23037) Yuxuan Zhang 2026-05-02 04:53:52 +08:00
• b58fa60a1f [core/attention] Add SGLANG_FLASHINFER_USE_PAGED env to force paged wrapper (#24165) Lucia Fang 2026-05-01 12:52:46 -07:00
• c7566a860b Add CI permissions for user luccafong, 842974287 (#24238) Lianmin Zheng 2026-05-01 12:50:36 -07:00
• ece8a1a788 Refactor device timer, clean up metrics collector, and add fwd occupancy metric (#24197) Lianmin Zheng 2026-05-01 10:25:25 -07:00
• 4a50cd781e [BugFix][HiMamba] Fix host-protected node deletion in HiMamba tombstone del (#23696) 3 people 2026-05-01 21:57:47 +08:00
• 5b7ce417d0 [P/D disagg] - support decode side radix cache (#19746) ishandhanani 2026-05-01 08:55:34 -05:00
• 4197c55968 [spec decoding] add tests for chain-style multi layer eagle + return_logprob (#24192) Qiaolin Yu 2026-05-01 16:48:48 +08:00
• 7e17a7a1e8 Upd: AITER->(#2879)a6bb499 (#24218) HAI 2026-05-01 01:26:08 -07:00
• d48095ba53 Bypass torch.cuda.use_mem_pool generator-CM in SymmetricMemoryContext (#24190) Cheng Wan 2026-05-01 01:25:49 -07:00
• d9e8a4a7f8 [SWA] Ensure we use pre-computed SWA cache location during prefill (#24138) 3 people 2026-05-01 00:01:49 -07:00
• 87dad74b33 Add benchmark/hicache/bench_warm_cache.py for exact warm-cache shared-prefix benchmarking (#24083) Clint 2026-05-01 01:48:32 -05:00
• a578bf814c [AMD] fix moriep unittest failure (#24205) billishyahao 2026-05-01 14:36:20 +08:00
• 6fa499cd6d [diffusion] CI: remove parametrized-only from diffusion PR test (#24202) Mick 2026-05-01 14:15:22 +08:00
• 8975479f87 [LoRA][MOE] Fix EP correctness in MoE LoRA slicing and virtual-experts kernels (#24171) Yanbin Jiang 2026-04-30 22:42:10 -07:00
• 7bc7775260 [CI] Fix stage-b-test-4-gpu-b200 silently skipped, hanging wait-for-stage-b (#24208) Baizhou ZhangandClaude Opus 4.7 2026-04-30 22:29:08 -07:00
• 357e3789ea ci: limit nightly test parallelism to 1 job per hardware type (#23314) Alison Shao 2026-04-30 21:49:18 -07:00
• 9d95783603 Add Docker image provenance metadata (#24090) AlecandKangyan-Zhou 2026-04-30 21:40:42 -07:00
• 9d84268705 [diffusion] refactor: introduce component residency manager (#23771) Mick 2026-05-01 11:10:41 +08:00
• 108bfd8b6a [MoE] Add Aiter MoE runner backend and purge aiter.fused_moe from quant methods (#23597) Cheng WanandClaude Opus 4.7 2026-04-30 19:50:52 -07:00
• 0acc569edd [Bench] extend MMMU answer extractor with explicit-commit patterns (#24084) Yihao Wang 2026-04-30 22:08:08 -04:00
• 1742bfb610 [CI] release-whl-kernel: skip musa wheels in update_kernel_whl_index (#24191) Baizhou ZhangandClaude Opus 4.7 2026-04-30 17:17:50 -07:00
• cf346bb15d [CI] Skip worker-dependent SMG e2e tests pending runner-image debug (#24166) Kangyan-ZhouandClaude Opus 4.7 2026-04-30 16:45:10 -07:00
• 5f88c8593a [Misc] Redirect default sglang nightly wheel to cuda 130 (#24183) Baizhou Zhang 2026-04-30 16:32:50 -07:00
• f67292539f spec: gate dp mlp sync with server args (#24177) Yilong Zhao 2026-04-30 16:29:41 -07:00
• da7f890788 [Intel GPU] Integrate flash_mla_decode in Intel XPU attention backend (#23557) 4 people 2026-05-01 04:51:28 +05:30
• e35ac95cdc [Test] Add XPU device support to unit tests (#22236) 3 people 2026-05-01 04:48:51 +05:30
• 8b23d32ec1 feat: implement workflow to sync LMSYS SGLang blog (#23438) zijiexia 2026-04-30 16:17:29 -07:00
• 918f910cf0 ci: temporarily disable multimodal-gen-test-1-b200 (#24174) Alison Shao 2026-04-30 16:16:43 -07:00
• 9c5cad3914 Use device-agnostic helpers for Mamba tests and core ops (#20234) 3 people 2026-05-01 04:44:53 +05:30
• 8a9e424faa Replace hardcoded CUDA device with get_device() for XPU support (#13599) Kalyan KumarandMa Mingfei 2026-05-01 04:43:46 +05:30
• c5f1339773 Revert "ci: add rebase-required mode to check-maintenance action" (#24179) Kangyan-Zhou 2026-04-30 16:11:36 -07:00
• e45b8ec1ec [CI] Publish nightly sglang wheel under both cu129 and cu130 indexes (#24176) Kangyan-ZhouandClaude Opus 4.7 2026-04-30 16:08:59 -07:00
• 694ef516cb Revert "[ci] split stage-c-test-4-gpu-b200 to enable a low-disk runner pool" (#24163) Alison ShaoandAlison Shao 2026-04-30 15:57:19 -07:00
• cdc4078815 ci: add rebase-required mode to check-maintenance action (#23109) Alison ShaoandAlison Shao 2026-04-30 15:47:24 -07:00
• f75a8b6220 fix: support HybridLinearAttnBackend in TboAttnBackend (#20114) Lawrence Wu 2026-04-30 15:40:13 -07:00
• d57671527a Fix LFM2 ShortConv Mamba State Indexing (#23975) Hubert Lu 2026-04-30 15:23:39 -07:00
• 2e027b1afe chore: bump sgl-kernel version to 0.4.2 (#24170) sglang-botandsglang-bot 2026-04-30 15:02:07 -07:00
• 989a16187d [Bench] Fix bench_serving missing reasoning_content stream chunks (#23954) Xinyuan Tong 2026-04-30 23:00:27 +01:00
• 340efca244 [sgl-kernel] Prep for torch 2.11 upgrade and switch PyPI default to cu130 (#24162) 3 people 2026-04-30 14:54:39 -07:00
• c04b20dc88 Fix KeyError in prepare_lora_batch when lora_ids contains None (#21974) Erik Wijmans 2026-04-30 11:50:16 -07:00
• 71e89e9003 [MUSA][19/N] Support qwen series models (#23654) oriandzhiguo.qin 2026-05-01 02:26:47 +08:00
• dc395bc059 ci: run setup_ld_library_path before install_sglang_kernel (#24141) Alison Shao 2026-04-30 10:55:45 -07:00
• 7bb7f6049a ci: add per-host utilization view to runner-utilization report (#24102) Alison Shao 2026-04-30 10:05:16 -07:00
• 651af06a0b [Feature] Xiaomi MiMo-V2.5 day0 support (#23811) 6 people 2026-05-01 00:02:26 +08:00
• cf4f462094 [AMD] Nightly image release for deepseek v4 (#24155) YC Yen-Ching Tsengandbingxche 2026-04-30 23:49:13 +08:00
• aa74911448 [NPU] fix some npu error with OffloaderV2 (#19541) 3 people 2026-04-30 20:05:35 +08:00
• 577dbc4ab9 [4/N] Quantization Refactor: AWQ schemes and Kernel call and weight init split (#21126) Yaochen Han 2026-04-30 19:51:01 +08:00
• b1ef99f65f [CI] Remove orphaned test/srt/ascend and test/srt/configs (#24145) Lianmin Zheng 2026-04-30 04:43:11 -07:00
• 583929c0a1 fix the compatibility between --moe-dense-tp-size 1 and piecewise cuda graph (#23972) Qiaolin Yu 2026-04-30 17:12:13 +08:00
• 99c0b62f1e allow requests with exactly context_len total tokens (#22546) Opher LieberandEthan Su 2026-04-30 01:12:06 -07:00
• 125f75db72 fix(lora): avoid CUDA graph-breaking scalar assignment in seg_indptr (#23738) Ethan (Yusheng) Su 2026-04-30 01:11:45 -07:00
• da07b22949 [Docs] quick fix delete --enable-dp-attention in sgl-jax (#24052) Brian 2026-04-30 15:39:03 +08:00
• 692979a8d9 [AMD] Support sdma path for moriep (#23929) billishyahao 2026-04-30 14:57:00 +08:00
• 4f0b44c5c6 [fix] moss-vl: use Conv3dLayer and remove no-op flat_encoder_result (#23932) Shaojun Zhou 2026-04-30 14:19:45 +08:00
• 936c9c2355 fix(qwen3_5): broadcast per-tensor scale in _make_packed_weight_loader for FP8 models (#23062) kkyyxhll 2026-04-30 14:16:57 +08:00
• bcb34da9f9 Add deterministic mode for XPU operations (#16793) Jay ThakurandMa Mingfei 2026-04-30 11:09:06 +05:30
• c8c1c9261d LoRA support for qwen3.5 and nemotron3 (#23594) Opher LieberandYanbin Jiang 2026-04-29 21:51:53 -07:00
• 0b1fbdba15 [diffusion] CI: change ground truth upload path and improve publish script (#24120) Mick 2026-04-30 12:26:10 +08:00
• c54ada994b fix: rename mimo spec threshold attr to num_accepted_drafts_thres (#24118) Liangsheng Yin 2026-04-29 21:00:55 -07:00
• d040333c95 [Bug Fix] missing index/KV transfer for MTP layer in NSA disaggregation (#23539) Yuxuan Zhang 2026-04-30 11:55:45 +08:00
• 9e68c8527e [AMD] Update AMD Nightly Test checkout mechanism (#24112) YC Yen-Ching Tseng 2026-04-30 11:36:18 +08:00
• 2d2be5d7b2 [PD][Bugfix] fix mamba cache capping (#22462) 3 people 2026-04-30 10:57:55 +08:00
• fdfe53e872 [misc] update CI_PERMISSIONS.json (#24062) Chi McIsaac 2026-04-29 22:24:22 -04:00
• 62136073f9 pin the version of xgrammar to v0.1.32 (#24010) MingxuZhandMa Mingfei 2026-04-29 19:13:08 -07:00
• 3553fd0322 [NPU] add split_qkv_tp_rmsnorm_rope ops for minimax2 & fix eagle3 hidden states capture in dp attn mode (#23190) heziiop 2026-04-30 08:51:22 +08:00
• e60c60eff0 [SWA] Fix missing mamba_indices parameter in cpu copy interface (#24026) Lianmin Zheng 2026-04-29 17:33:38 -07:00
• 2bbd30a27a relax the threshold in test_step3p5_flash_chain_mtp (#24105) Qiaolin Yu 2026-04-30 07:53:35 +08:00
• ea794deefc [misc] fix lint in main branch (#24095) Qiaolin Yu 2026-04-30 06:01:35 +08:00
• 6575aea128 [CI] Fix black formatting on main (unblocks PR #21247 lint) (#24093) Kangyan-ZhouandClaude Opus 4.7 2026-04-29 14:59:17 -07:00
• 3d31ac2672 [Fix] FP8 Qwen3-Next quant error by removing fallback fused shards (#23973) Jimmy Shong 2026-04-29 16:33:47 -05:00
• 850021378a [Score API] Hoist query placeholder scan and specialize PositionalEmbeds stacking (#23513) jsheng_LinkedinandClaude Opus 4.7 2026-04-29 13:51:53 -07:00
• 79dbfe4505 Use spec v2 by default (#21062) Qiaolin Yu 2026-04-30 04:40:42 +08:00
• c3ab5bec7d ci: consolidate rust + protoc install across workflows (#23700) Alex NailsandClaude Opus 4.7 2026-04-29 13:39:02 -07:00
• b3ead32d3c [minor] Remove incorrect note after supporting w4a16 moe for DeepSeek V4 (#24035) Baizhou Zhang 2026-04-29 13:26:13 -07:00
• 7389743d85 feat: Support modelexpress p2p RDMA transfer (#23105) Zhongdongming Dai 2026-04-29 12:57:40 -07:00
• db84a8ebbb [Model] Qwen3ForPooledOutput: forward get_input_embeddings to inner model (#23434) jsheng_Linkedin 2026-04-29 12:25:06 -07:00