Commit Graph

Select branches

Hide Pull Requests

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• d287880a7a Update deepep for SBO feature (#35450) Baizhou Zhang 2026-08-19 23:37:13 -07:00
• 0bda0b168a [Fix]: exclude SM120 from attn-res TMA dispatch (#35361) 5 people 2026-08-20 14:32:31 +08:00
• 628674a5c7 Remove unused MOONCAKE_COMPILE_ARG argument from Dockerfile (#35649) Shangming Cai 2026-08-20 14:19:52 +08:00
• 32d98aad13 [HiCache] Allow a retraction host pool smaller than the device pool (#35543) cctryandcctry 2026-08-19 22:59:57 -07:00
• 50dae2d99d Amd/dsv4 shared experts fusion top6 (#32340) 4 people 2026-08-20 00:57:58 -05:00
• 02b93e7e01 [AMD] Add GLM-5.2 MI35x nightly accuracy and perf benchmark (#32570) Michael 2026-08-19 22:49:43 -07:00
• 58e327480a update codeowner (#34802) siyuandliusy58 2026-08-20 13:43:15 +08:00
• ba433bb462 [Docs] Update contribution guide (#35419) Mohammad Miadh Angkad 2026-08-20 13:31:24 +08:00
• 09b7af1371 [CI] Surface AMD ROCm 7.2 state in the PR CI-states block (#34813) 3 people 2026-08-19 22:18:30 -07:00
• b6dcd393d6 [Fix] Support 128-aligned hidden sizes in the W4AFP8 DeepEP low-latency requant kernel (#35593) Alex NailsandClaude Opus 5 2026-08-19 22:18:17 -07:00
• f744607567 [diffusion] UX: report where a component's weights are (#35618) Mick 2026-08-20 13:05:14 +08:00
• a5a9d66baf [XPU] Fix/kimi linear xpu (#34546) 4 people 2026-08-20 10:13:41 +05:30
• db2eb47500 [diffusion] fix: keep cosmos3 T=1 fusion on blackwell only (#35612) Mick 2026-08-20 12:13:04 +08:00
• a3e4592f89 [diffusion] CI: use canonical residency selector in nightly (#35615) Mick 2026-08-20 10:50:29 +08:00
• 3b22f4f000 [diffusion] UX: reduce per-request log noise (#35614) Mick 2026-08-20 10:49:54 +08:00
• 360d10d6bc [Feature] Add process-local in-memory KV indexer and Router integration (#33370) 4 people 2026-08-20 10:45:35 +08:00
• 238ba40c27 [XPU][CI] key persistent JIT kernel cache by image content ID (#35337) ashwini rathi 2026-08-20 07:57:59 +05:30
• 9db4ba8da1 [DeepSeek-V4] Add Q8KV8 sparse MLA prefill runtime backend (#32327) 3 people 2026-08-20 10:23:49 +08:00
• a49560ce50 [misc] Add a comment style rule to .claude/rules (#35597) Liangsheng Yin 2026-08-19 18:52:48 -07:00
• f65961844a docker: fix CUDA-13 build — rename NCCL_VERSION ARG to avoid base image ENV collision (#35587) datdo-msft 2026-08-19 18:41:53 -07:00
• dc175b3ad2 [CI][AMD] Run the profiling suite without CUDA graphs on ROCm (#34452) 3 people 2026-08-19 18:41:28 -07:00
• c7478228dd [AMD] [Docker] Upgrade Python 3.12 + torch 2.11 + triton 3.7 in ROCm 7.2.4 (#30984) chuyehandChen 2026-08-20 09:18:31 +08:00
• ab203663c4 [diffusion] fix: reject unsupported modelopt checkpoint algorithms (#35182) Mick 2026-08-20 09:13:24 +08:00
• 1cf2b8c54d [Spec] Support quantized target lm_head in the DFlash2 selector (#35496) Jimmy ShongandLING ZHI 2026-08-19 18:06:41 -07:00
• 1f87d8f512 [diffusion] fix: stop reserving nccl device buffers for single-rank groups (#35538) Mick 2026-08-20 09:04:35 +08:00
• e805a8f98e [AMD] Keep the PTX-inline-asm diffusion norm fusions off on ROCm (fix FLUX warmup crash) (#34481) MichaelandCursor Agent 2026-08-19 17:55:28 -07:00
• f736895ce9 Make PR babysitter launcher fork-safe (#35575) Lianmin Zheng 2026-08-19 17:48:34 -07:00
• 99c12218c3 Support custom draft worker classes in DSpark (#35397) Lianmin ZhengandYichao Fu 2026-08-19 17:48:17 -07:00
• 9234e40aed [sampling] Fix int32 offset overflow in top-k renorm Triton kernels (#35571) Lianmin ZhengandXiaozhu Meng 2026-08-19 17:48:03 -07:00
• d216737e47 [Kernel] Support wider rows in mega_moe_pre_dispatch (#35372) Shiyan Deng 2026-08-19 17:47:04 -07:00
• 1df78c2cf1 chore: bump tilelang to 0.1.12 (#30874) elvischenvandClaude Opus 4.8 2026-08-20 08:01:19 +08:00
• 5375babbac [Quant] Load compressed-tensors quantized lm_head instead of value-casting it (#35228) Jimmy Shong 2026-08-19 15:37:45 -07:00
• c7e2c08d14 fix(constrained): reject NUL bytes in grammar specs to stop an xgrammar segfault (#34679) 4 people 2026-08-20 06:31:14 +08:00
• 082aac8fce [Bugfix] Fix min-new-token EOS handling (#31378) milesial 2026-08-19 15:29:22 -07:00
• 01814e110d [HiCache] Simple style change for buffer mode (#35574) Zhiqiang Xie 2026-08-19 15:23:21 -07:00
• 38b74d294b Add docs for TP LMHead optimizaiton (#35283) Siyuan Chen 2026-08-20 05:59:35 +08:00
• 1270204d2c Revert "[Feature] Add DeepEPv2 (ElasticBuffer) MoE A2A backend" (#35568) Liangsheng Yin 2026-08-19 14:14:38 -07:00
• a6bc0532c9 [Fix] Fix Nemotron-H Mamba illegal memory access under DP attention with CUDA graph (#34561) elvischenvandBrayden Zhong 2026-08-20 05:13:40 +08:00
• ed12d6827d fix(disagg): allow fake transfer with decode DCP (#35409) milesial 2026-08-19 13:54:37 -07:00
• defb2a3100 feat(openai): Accept the input_audio content part in chat completions (#33606) Jason Wiemels 2026-08-19 16:37:50 -04:00
• 746418a1ec [DSA] Trim top-k v2 output modes and tighten its PDL waits (#35041) DarkSharpnessandClaude Opus 5 2026-08-20 04:09:21 +08:00
• 1c82955861 [HiCache] Split the host-memory budget across co-located ranks (#35540) cctry 2026-08-19 13:01:58 -07:00
• b1707996e8 fix(gemma4): quantize MTP bridge projections (#32440) Ayushman Singh 2026-08-19 15:41:52 -04:00
• 6f69f927da [Scheduler] Add configurable decode interval after prefill (#35017) Po-Han Huang (NVIDIA) 2026-08-20 03:01:36 +08:00
• 4f8ecf6ae9 [Feature] Add DeepEPv2 (ElasticBuffer) MoE A2A backend (#29525) MengYuandmenyu 2026-08-20 02:52:45 +08:00
• 03cf2de2e3 [Qwen3.5][MTP] Preserve online NVFP4 draft quantization for mixed checkpoints (#35545) YAMY 2026-08-19 11:12:32 -07:00
• 157d8ad27a Support Intern-S2-Mobius FP8 (#34908) Xinyuan Tong 2026-08-20 01:58:01 +08:00
• 5f12839591 [Fix] Support Kimi-K3 ModelOpt mixed NVFP4/FP8 checkpoint (#35077) YAMY 2026-08-19 08:13:45 -07:00
• 41c018a9ec [UnifiedTree] feat: support runtime attach/detach (#35269) Shuwen Wangandhzh0425 2026-08-19 22:49:08 +08:00
• 1ef7882a5b [NIXL] Query EP top-k index dtype (#35294) Yoray Zack 2026-08-19 17:31:05 +03:00
• 23f2320c95 [Docs] PaddleOCR-VL: update which stage of the pipeline this serves and show real output (#35458) MickandClaude Opus 5 2026-08-19 21:27:58 +08:00
• e3445ed2bd [diffusion] fix: route quantized vae component repos safely (#35184) Mick 2026-08-19 21:02:46 +08:00
• 29f5d1c7c3 [diffusion] fix: fix multi-group layerwise offload startup memory (#35509) Mick 2026-08-19 20:32:31 +08:00
• c57ada81e1 [Diffusion] Use current_platform instead of hardcoded "cuda" in cosmos3 guardrails (#34612) Arseniy Mironovandronnie_zheng 2026-08-19 15:26:28 +03:00
• 574274660f [AMD] cookbook: serve Qwen3.5 MXFP4 on MI355X with an fp8_e4m3 KV cache (#35445) jacky.cheng 2026-08-19 18:49:32 +08:00
• 3e5ce26c2d fix: fix transcription & audio-understanding for ASR/audio/speech models (#32611) Rohit Kumar SinghandSingh 2026-08-19 16:14:23 +05:30
• f446e853e7 [AMD] DeepSeek-V4: route decode wo_a bf16 batched matmul to aiter batched_gemm_bf16 (#33313) 3 people 2026-08-19 05:04:48 -05:00
• ce1830c59b [AMD] DeepSeek-V4 MI355X: eliminate bpreshuffle fp8-scale relayout copy in dense w8a8 linear (#33165) karverma-amd 2026-08-19 05:02:40 -05:00
• f22442d3a4 Add three new test cases (#35502) HeYaoandHeYao 2026-08-19 17:59:45 +08:00
• adca19c497 [PD] Deferred decode-side KV release for the NIXL backend (#35360) Shangming Cai 2026-08-19 17:29:45 +08:00
• aa215e5523 [PD] Overlap prefill DP-rank bootstrap queries (#35071) YAMY 2026-08-19 02:25:52 -07:00
• 0e4a09480c [HiCache] Support DCP with DSpark (#35221) Khoa PhamandCursor 2026-08-19 01:42:19 -07:00
• 9113fc6d93 [docs] Add a fused-kernels page for SGLang Diffusion (#35436) Xiaoyu ZhangandClaude Opus 5 2026-08-19 16:32:42 +08:00
• 8a1e6e4e46 Qwen3.8-27B Model Support (#34859) 7 people 2026-08-19 16:31:43 +08:00
• ebec85f606 [AMD][DI][CI] Run MI355X disagg nightly at 7AM UTC (#35467) kangwangamdandbingxche 2026-08-19 16:27:20 +08:00
• 73e5fa4724 [diffusion] refactor: gate native encoder quantized checkpoints (#35183) MickandYiqi Yang 2026-08-19 16:18:36 +08:00
• ccbe380028 [CI] Trim the base-c 4-gpu-h100 stage from 5 shards to 4 (#35407) Liangsheng Yin 2026-08-19 00:48:07 -07:00
• e614121866 [Fix] Scale the req_to_token row headroom by attn_dcp_size (#35424) Liangsheng Yin 2026-08-19 00:37:26 -07:00
• f4158719d1 [AMD] Let the diffusion AITer backend take grouped-query K/V (fix Cosmos3-Nano startup) (#34485) 3 people 2026-08-19 00:17:40 -07:00
• c0c87e0547 VLM: feed the packed qkv projection output to vision backends uncopied (#35336) MickandClaude Opus 5 2026-08-19 14:39:52 +08:00
• a5c96362b6 [diffusion] chore: reuse shared checkpoint quant metadata resolver (#35174) Mick 2026-08-19 13:36:55 +08:00
• 593b1a9b8a [diffusion] optimization: reduce minimax h3 mps memory pressure (#33880) Mick 2026-08-19 13:35:01 +08:00
• 3391ab3712 [Constrained] Support MistralCommon tokenizers in the XGrammar backend (#35215) Alison ShaoandXinyuan Tong 2026-08-18 21:34:48 -07:00
• a72d8d29d3 Fix HiCache PP sync test fixture (#35446) Shuwen Wang 2026-08-19 12:25:17 +08:00
• c863760ae1 [Fix] DCP: advertise the logical KV-event block size (#35298) Jimmy ShongandClaude Fable 5 2026-08-18 20:55:11 -07:00
• 88f6074392 feat(api): add sglext_spec (#33518) Muqi LiandCodex 2026-08-19 11:54:18 +08:00
• 3b065a56b0 [HiCache] Batch PP write and load completion sync (#33473) luoroger37 2026-08-19 11:27:27 +08:00
• 1beb805356 [Perf] Restore the 16-token router GEMM threshold on SM10X (#34953) Mohammad Miadh AngkadandMohammad Angkad 2026-08-19 10:53:52 +08:00
• ee1f2e8dfd [diffusion][Minimax H3]support subblock sparse attention on SM90 (#34680) HuangJi 2026-08-19 10:31:45 +08:00
• eb085524c8 [diffusion] fix: make MiniMax-H3 AdaLN cache rebuild transactional (#34993) DayuxiaoshuiandMick 2026-08-19 10:27:38 +08:00
• 977412ae61 [HiCache] Buffer-only mode for HiCache host memory layer (#34798) Zhiqiang Xie 2026-08-18 19:21:24 -07:00
• 4cef72faee [diffusion] refactor: reuse srt qwen vision and text modules (#35006) Mick 2026-08-19 10:12:44 +08:00
• 58c5bee3ac Fix DP attention on CPU (#12961) Chunyuan WU 2026-08-19 09:56:21 +08:00
• 5d12280ae7 Add fmha_v2 attention backend for SM90/120 (#23112) akhilg-nvandYangmin Li 2026-08-18 18:41:03 -07:00
• baa2251847 [diffusion] chore: make --vae-tiling honest, fix the decode oom advice, gate nvfp4 on blackwell (#35353) Mick 2026-08-19 09:27:54 +08:00
• ef490853bb quant: extract shared checkpoint quant metadata resolver (#35172) Mick 2026-08-19 08:26:41 +08:00
• e73201e462 [diffusion] feat: support cache-dit, cfg gating, attention backend override as per-request param (#35339) MickandClaude Fable 5 2026-08-19 08:24:03 +08:00
• 77fc5c128e [perf] overlap page preprocessing, pack the vit, enable prefill CUDA graph for paddle-ocr (#35318) MickandClaude Opus 5 2026-08-19 08:20:55 +08:00
• 64e404263e [Doc] Fix TP and attention-TP group layout in initialize_model_parallel docstring (#34862) 3 people 2026-08-19 08:15:03 +08:00
• c14312a664 [Spec] DFlash2: local convolution + candidate selector (#35371) 4 people 2026-08-18 17:07:28 -07:00
• 3a8f522f65 install sglang in virtual env instead of system path (#30612) Augusto Yao 2026-08-19 07:36:50 +08:00
• cfc6dfb364 Apply latest DeepEP branch (#34923) Baizhou Zhang 2026-08-18 16:24:50 -07:00
• 37c09ff3d8 [Memory] Borrow CUDA graph pool storage for EAGLE sampling (#35375) cctryandcctry 2026-08-18 16:19:41 -07:00
• 7ebaa98f81 [Fix] Assert the page-aligned SWA evict floor on both PD decode prealloc paths (#35396) Liangsheng Yin 2026-08-18 15:48:55 -07:00
• b814a7e812 [CI] Skip fast-fail for scheduled stages (#35392) Baizhou Zhang 2026-08-18 15:07:33 -07:00
• 87a09494fa [Refactor] Share the page-aligned decode alloc lens between EAGLE and DFLASH (#35382) Liangsheng Yin 2026-08-18 14:39:41 -07:00
• 7b5410c999 Laguna: config-driven MoE router scoring (#35362) Joe Rowell 2026-08-18 22:42:36 +02:00
• 79dfef390b [Spec] Page-align the DFLASH decode KV reservation (#35265) Liangsheng Yin 2026-08-18 13:36:04 -07:00
• aa82229173 [Fix] Assert the page-aligned SWA evict floor at PD decode prealloc (#35286) Liangsheng Yin 2026-08-18 13:32:16 -07:00
• 955704544c [Fix] Skip padded state slots in the chunked GDN kernel (#33431) Yuwei AnandClaude Opus 5 2026-08-18 13:07:19 -07:00