Commit Graph

Select branches

Hide Pull Requests

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 539253e1d5 Gate Rust extension builds (#30927) Lianmin ZhengandJialin Ouyang 2026-07-12 05:44:58 -07:00
• 80856aba85 Make the mxfp8 MoE runner backend list extensible (#30828) Yonghao Zhuang 2026-07-12 01:28:10 -07:00
• 81d273f73b Handle coredump dirs and cache hit updates (#30897) Lianmin ZhengandYongji Wu 2026-07-12 00:37:43 -07:00
• f1c247edf9 profile: add vlm prefill profiler ranges (#30871) Mick 2026-07-12 14:07:10 +08:00
• bce3fc987d perf: reuse MoonViT FA3 max-seqlen metadata (#30878) Mick 2026-07-12 14:05:21 +08:00
• 592c04381d Update test repository case scripts to the main community (#29939) pllimax 2026-07-12 11:36:30 +08:00
• a358abd651 chore: update vlm moe config and tune scripts (#30866) Mick 2026-07-12 08:35:59 +08:00
• 14bef7cd11 fix: lazy load TileLang MHC kernels (#30580) Mohammad Miadh Angkad 2026-07-12 08:32:27 +08:00
• af66370d81 bench: support random image resolutions (#30879) Mick 2026-07-12 08:28:56 +08:00
• 4884f6fbee [Fix] Unify ForwardBatch extend lens cpu fields to their declared list type (#30896) Liangsheng Yin 2026-07-11 17:30:53 -05:00
• d8ef76682e [Spec] Extract shared draft worker construction and generalize draft sampler capture (#30857) Liangsheng Yin 2026-07-11 12:34:09 -07:00
• 348e6fd29b [Fix] Guard kernel OOB accesses and harden runtime edge cases (#30847) Liangsheng Yin 2026-07-11 12:22:13 -07:00
• ed554aac17 Fix: add grammar sync in PP for structured output (#30747) 3 people 2026-07-12 02:54:50 +08:00
• 9b4bb415dd [cuda-graph] Size breakable-graph shared buffer from warmup output; slice by produced row count (#30834) Ming YangandClaude 2026-07-11 10:41:44 -07:00
• 32cb89d412 ci: prune uv cache in job teardown to bound its growth (#30813) Alison Shao 2026-07-11 03:01:23 -07:00
• 268b8e127f [NPU][bugfix] Fix NPU KernelLaunch Failure in rotate_input_ids_triton with Empty Batch (#30589) iridiumine 2026-07-11 17:32:42 +08:00
• 7bac9c8cdb [Spec] Enable draft extend cuda graph for DeepSeek-V4 attention backend (#30853) Liangsheng Yin 2026-07-11 02:29:49 -07:00
• 79f096d43b [DOCS][NPU]update npu support features and models (#30843) loading66 2026-07-11 15:21:29 +08:00
• 65abb23842 Add diffusion BCG prompt conditioning guard (#30782) Xiaoyu Zhang 2026-07-11 13:14:31 +08:00
• a91c2e6596 [Apple Silicon] [CI] Move the MLX lane to the check-changes + pr-gate composite (#30121) Jae B. 2026-07-11 01:00:15 -04:00
• e1d51be91f [Tiny] Fix a typo in cookbook (#30837) Baizhou Zhang 2026-07-10 21:47:35 -07:00
• 51c5ddbe65 [eplb] chunk expert-weight P2P on CUDA to prevent NCCL rebalance hang (#30829) Lucia Fang 2026-07-10 21:45:17 -07:00
• 07165d5daa Add DCP to runtime parallel context (#30478) Baizhou Zhang 2026-07-10 21:23:41 -07:00
• bbcfcaeefe fix(pd): do not abort when req.disagg_prefill_dp_rank is used (#27546) Lawrence WuandQiaolin Yu 2026-07-10 21:08:02 -07:00
• 90688366d9 test(disagg): set MC_GID_INDEX on RoCE hosts so mooncake KV transfer works (#30737) Alison Shao 2026-07-10 20:29:22 -07:00
• 0663ebc783 [minimax-m3] Split 4/4: model + VL + glue + function-call + fp8 quant + generic infra (#28715) 3 people 2026-07-11 11:11:06 +08:00
• e3ceccf781 [dLLM] Make FDFO a framework capability for all dLLM algorithms (#27551) Chenchen HongandXiaoyu Zhang 2026-07-11 11:05:05 +08:00
• fc2ef35308 [refactor] Move MLP collective flags onto ForwardFlags (#30802) Cheng Wan 2026-07-10 17:43:39 -07:00
• 7de33ce806 fix: fix mm processor double bos (#26505) Yihao Wang 2026-07-10 16:53:08 -07:00
• 649ce5dd3d model: support Pi0.5 (#30633) Mick 2026-07-11 07:50:58 +08:00
• e8646701c1 Update GLM-5.2 NVFP4 cookbook (#30826) Baizhou Zhang 2026-07-10 15:56:32 -07:00
• 4fcc994be1 Support priority request header override (#30811) Lianmin ZhengandAbhishek Parmar 2026-07-10 14:36:43 -07:00
• bc82b06400 Add ZYHowell to CI_PERMISSIONS.json (#30810) Lianmin Zheng 2026-07-10 14:09:32 -07:00
• 3f1694f5e0 Update sgl-deep-gemm to 0.1.4.post1 (#30697) Baizhou Zhang 2026-07-10 13:30:43 -07:00
• 7b99900980 Fix hybrid attention graph hook test fixture (#30790) Mohammad Miadh Angkad 2026-07-11 02:07:04 +08:00
• 7998fecfd1 [kv canary] Support UnifiedRadixCache in kv-canary and bracket nested model.forward (#30574) Lianmin Zheng 2026-07-10 10:58:55 -07:00
• 3dc93a12ca Improve EPLB dispatch handling and diagnostics (#30646) Lianmin ZhengandYonghao Zhuang 2026-07-10 10:40:19 -07:00
• 789bc3995c [AMD-miles] Bump miles rocm700-mi35x base image and migrate nightly test onto it (#30587) Zhiyao Jiang 2026-07-10 11:47:06 -04:00
• 0299393758 [UnifiedTree]: Sync mamba int8 checkpoint (#30626) Zhangheng 2026-07-10 23:46:38 +08:00
• 559854fe6a [diffusion] docs: sync cookbook and log hygiene (#30791) Mick 2026-07-10 22:55:14 +08:00
• 7090a49198 update codeowners (#30788) Mick 2026-07-10 22:35:34 +08:00
• 2286e25a21 [UnifiedTree]: Sync Replay SSM (#30636) Zhangheng 2026-07-10 21:54:59 +08:00
• 6ed9843b57 [Kernel] Introduce sglang.kernels namespace and migrate scattered triton_ops kernels (RFC #29630, Phase 2) (#30044) Xiaoyu ZhangandClaude Opus 4.8 2026-07-10 21:41:08 +08:00
• e9493a015c Fix diffusion BCG lifetime and add Z-Image-Turbo CI (#30584) Xiaoyu Zhang 2026-07-10 21:19:59 +08:00
• 94de28764c Fix TiktokenTokenizer missing num_special_tokens_to_add (#30643) 3 people 2026-07-10 18:07:30 +05:30
• 7045e0fdff Seed sgl-kernel topk sigmoid tests on all backends (#30754) Mohammad Miadh Angkad 2026-07-10 18:08:15 +08:00
• 6174f1cad8 [test] Set init-static attrs in mm_process_config mock fixtures (#30766) Xinyuan Tong 2026-07-10 18:00:00 +08:00
• fd19a76237 [BUG] fix strip streaming empty-string suffix from DSV4 tool arguments (#29883) BingjiaWang 2026-07-10 17:05:26 +08:00
• ecb7fb3989 Fix CuTe DSL DSA paged MQA export (#30627) Mohammad Miadh Angkad 2026-07-10 16:32:58 +08:00
• fef5eda4fb [Refactor] Split DeepSeek-V4 MQALayer into a reusable attention base (#30711) Liangsheng Yin 2026-07-10 01:14:03 -07:00
• 4a8e1b07a2 [diffusion] refactor: reorganize runtime utility and server_args modules (#30447) Mick 2026-07-10 15:53:00 +08:00
• 1d8e3c248b Fix DSV4 HiSparse SWA tail allocation forwarding (#30408) Junjie CaoandZhangheng 2026-07-10 15:27:02 +08:00
• b76dd0be69 Fix Mistral GSM8K chat eval (#27757) Xinyuan Tong 2026-07-10 12:08:48 +08:00
• a38cfc6768 [diffusion] doc: update cosmos3 cookbook (#27576) Richard WangandMick 2026-07-10 11:19:52 +08:00
• 7966f6be86 [NPU] fix npu import cutlass error (#30454) zhaozx-cn 2026-07-10 10:55:31 +08:00
• edd91cbdd5 [diffusion] feat: support action output for cosmos3 (#27168) 3 people 2026-07-10 04:48:04 +02:00
• 5e3dc5dd5f Fix immediate profiler step range boundary (#30215) weireweireandweireweire 2026-07-10 10:47:23 +08:00
• 23390589f7 [misc] Remove unit test cases that fail the admission criteria (round 3) (#30713) Liangsheng Yin 2026-07-09 19:42:51 -07:00
• 7b9b2e4798 [style] Extract init-static values in memory-cache path (#30710) Liangsheng Yin 2026-07-09 19:38:15 -07:00
• ccd2028def [style] Extract init-static values in tokenizer + multimodal path (#30709) Liangsheng Yin 2026-07-09 19:36:34 -07:00
• dda61b476e [style] Extract init-static values in forward path (#30708) Liangsheng Yin 2026-07-09 19:35:00 -07:00
• b5e75b9423 [style] Extract init-static values in scheduler hot path (#30707) Liangsheng Yin 2026-07-09 19:33:02 -07:00
• 073b36853f [CPU] update fla.cpp to support when num_head_v is not multiples of 16 (#30604) Ma Mingfei 2026-07-10 09:21:07 +08:00
• 5ce5e1ee3e [Diffusion] Revert CPU AMX optimizations (#30716) Mick 2026-07-10 09:09:38 +08:00
• 1e75ba236e Avoid implicit field-based side channel in Scheduler planning (#29408) fzyzcjy 2026-07-10 08:55:51 +08:00
• 32c8973ce8 Configurable decode retraction order (#30573) Lianmin Zhengandtanujtiwari1998 2026-07-09 17:55:26 -07:00
• 5be9c9f7c6 Localize cur_batch field in Scheduler to avoid field-based state access (#29407) fzyzcjy 2026-07-10 08:55:06 +08:00
• 69368d7593 Stop reading cur_batch in is_fully_idle and abort_request (#29406) fzyzcjy 2026-07-10 08:54:31 +08:00
• 2e66707399 Fix pipeline-parallel abort missing in-flight requests in non-current microbatch slots (#29405) 3 people 2026-07-10 08:53:18 +08:00
• 2c6cd1ef41 [Dep] Upgrade flashinfer to 0.6.14 (#29910) 3 people 2026-07-09 17:52:33 -07:00
• b2f9a95867 [NPU] use standalone group for moe ep (#29030) Even Zhou 2026-07-10 08:49:49 +08:00
• 295f85df08 Fix DFlash mamba verify init ordering (#30680) Mohammad Miadh Angkad 2026-07-10 08:40:12 +08:00
• cfc66e05c5 [tokenizer] Support pluggable tokenizer worker class in multi-tokenizer mode (#30630) Lianmin Zhengandtanujtiwari1998 2026-07-09 17:31:54 -07:00
• 77b7698cad Add tanujtiwari1998 to CI_PERMISSIONS.json (#30649) Lianmin Zheng 2026-07-09 17:03:46 -07:00
• 87992eeec4 [DeepSeek V2] Reorder dual-stream MoE to main-first to avoid CUDA graph stream explosion (#30460) 4 people 2026-07-09 16:59:39 -07:00
• 504570f425 [Refactor] Make DeepSeek-V4 attention backend tolerate an absent CPU seq_lens mirror (#30695) Liangsheng Yin 2026-07-09 16:39:47 -07:00
• a36c873147 [misc] Remove unit test cases that fail the admission criteria (round 2) (#30703) Liangsheng Yin 2026-07-09 16:35:59 -07:00
• b86466d54b Make KvVmmArena JIT stub unique per process (#30702) cctry 2026-07-09 15:56:26 -07:00
• c53559ba10 [misc] Remove unit test cases that fail the admission criteria (#30690) Liangsheng Yin 2026-07-09 15:31:28 -07:00
• 7e936f690e [Tiny] Fix Lint in #30645 (#30699) Baizhou Zhang 2026-07-09 15:30:48 -07:00
• c46b291c77 [misc] Add init-static value extraction to the general code style rule (#30701) Liangsheng Yin 2026-07-09 15:25:53 -07:00
• 40a522203c [Quantization][Bugfix]: Join multi-arg RuntimeError in Quark _check_scheme_supported (#25694) Spandan Tiwari 2026-07-09 15:07:14 -07:00
• 48d98b7c68 [Quantization][bugfix] Correct E8M0 NaN-sentinel detection in e8m0_to_f32 (#25519) Spandan Tiwari 2026-07-09 15:02:53 -07:00
• 966350408e [Quantization] Update error message strings with correct framework name in Quark/compressed-tensors (#25467) Spandan TiwariandSpandan Tiwari 2026-07-09 14:41:54 -07:00
• 10bb2eff3d [BCG] Restore Qwen3.5 MRoPE fusion under breakable CUDA graph (#27918) Kaixi 2026-07-09 23:30:05 +02:00
• bda1dc0d95 [DSA] Fix top-k v2 emitting invalid indices under tie overflow / inf scores (IMA in FA3 sparse decode) (#30645) DarkSharpnessandClaude Fable 5 2026-07-10 05:28:21 +08:00
• 8d0fd34150 [AMD] Enable unified-KV HiCache on DeepSeek-V4 (#29417) Thomas WangandHAI 2026-07-10 03:58:47 +08:00
• 7132af28de Fix garbage output for bare-tekken Mistral checkpoints (e.g. Leanstral) (#30396) Xinyuan Tong 2026-07-10 02:38:38 +08:00
• b717546fab fix(mtp): avoid mtp perf regression in deepseek when enable eplb (#28982) ICENaclandchunjiang.hc 2026-07-10 02:29:07 +08:00
• 462b6171bd [AMD] Fix stale SWA ring buffer on radix prefix reuse for DeepSeek-V4 with unified_kv backend (#30339) amd-danli103andamd-danli103 2026-07-10 02:28:22 +08:00
• b0ecbceed9 [Bugfix] Migrate retired parallel accessors (#30653) Mohammad Miadh Angkad 2026-07-10 02:22:34 +08:00
• 26ba3458d3 [AMD] Fix int32 offset overflow in Triton decode-attention kernels (#28788) 3 people 2026-07-10 02:20:04 +08:00
• 2e4d6368c3 [GDN] Auto-select FlashInfer GDN prefill on validated SM100 configs (#29734) YAMY 2026-07-09 09:26:00 -07:00
• 1959335997 refactor(load-snapshot): build LoadSnapshot directly, drop legacy get_loads IPC (#30525) cctryandcctry 2026-07-09 05:58:10 -07:00
• 7aab39a18b [Diffusion] SGLang backend for GLM Image AR. Step 1 - Separate server (#25381) 5 people 2026-07-09 15:54:50 +03:00
• 6ab7a65d94 Support grad injection and step override in the dumper's model dump (#30657) fzyzcjy 2026-07-09 20:26:05 +08:00
• 5b28465eb9 Cap diagnostic detail computation for failing tensors (#30656) fzyzcjy 2026-07-09 20:19:51 +08:00
• 287291c232 Fix rel_diff being nan for bitwise-identical tensors (#30655) fzyzcjy 2026-07-09 20:16:18 +08:00
• 0d7e8cfb85 Support per-regex diff-threshold predicates in the tensor comparator (#30654) fzyzcjy 2026-07-09 20:15:36 +08:00
• 4153410477 [NPU] [DOC] remove unsupported models from Ascend NPU models list (#30647) amote-i 2026-07-09 20:01:00 +08:00