Commit Graph

Select branches

Hide Pull Requests

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• d1895cb60d [Spec] Extract move_accept_tokens_to_target_kvcache into spec_utils (#27764) Liangsheng Yin 2026-06-09 21:55:26 -07:00
• 2495c02c2c [Refactor] Cuda Graph Runner/Backend Refactor (#23906) 4 people 2026-06-09 21:36:57 -07:00
• 56f06278c6 [diffusion] refactor: refactor realtime control state and adapters (#27698) Mick 2026-06-10 12:27:16 +08:00
• 047e5df3b1 Revert "Share BCG output buffers across capture sizes" (#27758) Baizhou Zhang 2026-06-09 20:51:40 -07:00
• 165331a200 Share BCG output buffers across capture sizes (#27659) Lianmin Zheng 2026-06-09 20:33:46 -07:00
• d21c31f681 fix: forward update_mamba_state_after_mtp_verify in HybridAttnBackend (#25883) fatSheep 2026-06-10 11:06:50 +08:00
• f101b287ef [Unified Tree]fix compatibility with eagle key and l3 hicache (#27655) huangtingwei 2026-06-10 10:54:45 +08:00
• f42a093261 [AMD] Migrate 2-GPU kernel allreduce tests into the registered system (#27722) Michaelandmichaelzhang-ai 2026-06-09 19:39:03 -07:00
• 7e3e616159 Add Arm64 INT8 MoE test coverage (#25007) Mandepudi Rani Chowdary 2026-06-09 21:36:57 -05:00
• 854d232a40 Fix flaky hicache l3 mmlu nightly test (#27688) Ke Bao 2026-06-10 10:01:14 +08:00
• 6565b7c464 [Docs] Update MegaMoE handling and rerun benchmarks (#27726) zijiexia 2026-06-09 19:00:43 -07:00
• 6110ed671f ci(xpu): clean build artifacts in cleanup (#27648) 3 people 2026-06-10 07:16:54 +05:30
• 5809bbe35d Mistral3 add tensor parallel support for diffusion text encoder (#25950) sushil Dubey 2026-06-10 07:13:21 +05:30
• af55025644 [diffusion] refactor: refactor realtime and model-specific stage modules (#27697) Mick 2026-06-10 09:39:06 +08:00
• bcd9c5a903 update pytorch-xpu to 2.12 (#27133) 3 people 2026-06-10 09:25:30 +08:00
• 77c4d53f19 [PD] Fix prefill bootstrap registration failure with --host 0.0.0.0 (#27608) Jianhong Zhang 2026-06-09 18:15:26 -07:00
• 2947781ce6 [NPU] MiMo-V2-Flash Adaptation (#25455) iridiumine 2026-06-10 09:13:55 +08:00
• f3ecc3688f Fix Gemma3 ModelOpt kv-scale loading (#25794) Jan Bernlöhr 2026-06-10 02:53:21 +02:00
• f332e52611 Add UT guarding per-request bookkeeping clock ownership (#27710) Liangsheng Yin 2026-06-09 17:11:49 -07:00
• 365b7dab9a fix(schema): update tokens_after_end (#27017) Muqi Liandzqlcode 2026-06-10 07:52:35 +08:00
• bc82086ef8 Remove FlashInfer GB transport workaround (#27453) Mohammad Miadh Angkad 2026-06-10 07:48:49 +08:00
• 98fe7e326e fix(gemma4): register image/video/audio token_regex for HF-expanded prompts (#26320) Wenqiandwenqi 2026-06-10 00:30:13 +01:00
• 9ab7a64ee1 [AMD] Update amd qwen3.5 cookbook (#27660) Thomas Wang 2026-06-10 07:26:33 +08:00
• ca716f4734 Add TP server GPU process regression test (#27721) Lianmin Zheng 2026-06-09 16:25:27 -07:00
• 53a4b51f8c Fix GLM NextN draft value head dim (#26049) Yueming Yuan 2026-06-09 16:13:37 -07:00
• 4455abd164 dflash piecewise cuda graphs support (#27468) David Wang 2026-06-09 18:44:19 -04:00
• decb88e0e3 Support spec v2 for Frozen-KV MTP; remove v1 worker (#27607) 3 people 2026-06-09 15:30:20 -07:00
• 7f730edfdc fix: correct off-by-one in vocab boundary check for token validation (#22367) 3 people 2026-06-10 06:24:39 +08:00
• fde4004429 [Fix] Reset positions tensor in CUDA graph runner when batch size differs from captured size (#24401) weizhoublue 2026-06-10 06:03:58 +08:00
• 2fef951fe8 [AMD] Replace fp8 mla with fp8 mha kernel for diffusion model aiter backend (#23927) jacky.cheng 2026-06-10 05:46:41 +08:00
• 42322947aa [BUG FIX]Fix DSA CPU offload mamba indices signature (#27645) ziang663 2026-06-10 05:03:38 +08:00
• eb8dceda44 Defer DeepGEMM PDL setup to worker init (#27671) Lianmin Zhengandlmzheng 2026-06-09 13:52:30 -07:00
• 186f1e300a [CI] Move JIT kernel tests + benchmarks to test/registered/jit; add in-package guard (#27644) Liangsheng Yin 2026-06-09 12:37:39 -07:00
• 8ae328e5f0 [sgl] Fix kimi-k2.5 EAGLE3 MLA draft embeds for batched MM prefill (#27647) Bi Xue 2026-06-09 11:26:48 -07:00
• 5babb902a9 [AMD] fix: handle per-frame 4D shift in native scale-shift kernel (#27581) Michael 2026-06-09 10:31:00 -07:00
• aa18a68ac5 [diffusion] Run LTX-2 VAE decode in channels_last_3d (faster decode, lower peak memory) (#27431) Xiaoyu ZhangandClaude Opus 4.8 2026-06-09 23:26:40 +08:00
• 17d8c5801d [AMD] Fix test_deepseek_r1_mxfp4_8gpu.py : disable async-assert probes on AMD CI (#27505) YC Yen-Ching Tseng 2026-06-09 23:12:35 +08:00
• badab6b136 [router] Add request/TTFT/worker metrics + Grafana dashboard to experimental sgl-router (#27591) Kangyan-ZhouandClaude Opus 4.8 2026-06-09 07:25:56 -07:00
• 1368717248 Add more testing for chunked prefill (#27506) fzyzcjy 2026-06-09 20:19:30 +08:00
• 609f5f549c Add mixed-prefix gsm8k eval and its CPU unit test (#27502) fzyzcjy 2026-06-09 20:17:41 +08:00
• fdcd28a08d [NPU] Enable consistency checking for diffusion tests (#27283) 3 people 2026-06-09 15:01:45 +03:00
• c6be251c5b [NPU] RL update_weights_from_disk/ tensor /distributed (#26717) McZyWu 2026-06-09 16:52:36 +08:00
• cd6efcb947 [NPU][Bugfix] fix MTP accuracy regression on Qwen3 hybrid models (#27202) AndyLi429 2026-06-09 15:47:23 +08:00
• 2218622f50 Fix spec v2 stop output boundary (#25980) 4 people 2026-06-09 15:32:21 +08:00
• d981b7b9c4 [Fix] Avoid applying cuda graph input-buffer registry on non-cuda devices (#27549) Zaili Wang 2026-06-09 15:24:55 +08:00
• a287ab83c0 Fix Gemma4 NVFP4 MoE default attention backend (#26791) Mohammad Miadh Angkad 2026-06-09 14:33:08 +08:00
• d145a6127a fix: stop-string check misses early matches during speculative decoding (#23802) 4 people 2026-06-09 14:25:58 +08:00
• 991689fd0d fix(hiradix): wait for extra pool IO (#27550) LJL36andZhangheng 2026-06-09 14:17:36 +08:00
• dff695c76b [lora] Exclude finished requests from running_loras (#27597) Erik Wijmans 2026-06-08 23:05:43 -07:00
• 9a3e845fc1 [Spec] Add nvtx to spec regions (#27615) Khoa PhamandCursor 2026-06-08 22:58:03 -07:00
• 5c0b2859e8 [diffusion] rl: extract post-training weight apis into mixins and add tensor update/checker paths (#22817) 7 people 2026-06-08 22:57:47 -07:00
• cae022aa88 [JIT] Reuse JIT kernel build cache across CI runs (#27605) Liangsheng Yin 2026-06-08 22:14:48 -07:00
• a32aeb688a [AMD] Fix AttributeError in GeneratedSharedPrefixDataset.from_args for in-process callers (#27580) Michael 2026-06-08 21:53:14 -07:00
• f6e6394d86 ci: run pr-test-extra on release branch cuts (not just base suites) (#27627) Alison Shao 2026-06-08 21:18:50 -07:00
• c2eae96c56 MSCCL++ Integration (#22734) 3 people 2026-06-08 21:13:13 -07:00
• 9c53031d2b ci: partition the H200 nightly 'Run test' step across the matrix (#27621) Alison Shao 2026-06-08 20:49:12 -07:00
• f6d53d6d16 docs: update SANA-WM cookbook serve examples (#27626) Mick 2026-06-09 11:44:31 +08:00
• 0f8673851c test: fix gemma GSM8K thresholds in nightly text eval (#27342) Khoa Pham 2026-06-08 19:58:42 -07:00
• bdf47315ce docs: add cookbook for SANA-WM (#27198) 3 people 2026-06-08 19:47:32 -07:00
• 15c801f726 fix(server): clamp piecewise_cuda_graph_max_tokens to context_length (#22516) Jan BernlöhrandXueLiang Yang 2026-06-09 04:43:54 +02:00
• 95090b837e [router] Add /flush_cache endpoint to experimental sgl-router (#27612) Kangyan-ZhouandClaude Opus 4.8 2026-06-08 18:53:29 -07:00
• db143e5212 [Intel GPU][Encoder] Add xpu_attn backend for encoder vision attention (#26460) jianan-guandMa Mingfei 2026-06-09 09:47:44 +08:00
• ea66b2cca7 [XPU] Enable NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 on Intel XPU backend (#24390) 4 people 2026-06-08 18:46:12 -07:00
• 71e8258783 Improve registration in cpu_graph_runner (#26635) Cao E 2026-06-09 09:35:33 +08:00
• 009a0ceefa [XPU CI] Re-enable stage B with docker-pull flow and split tests (#27526) ashwini rathiandClaude Opus 4.7 2026-06-09 06:56:11 +05:30
• d7c8b9ab9f [Intel GPU] Enable fused_experts in fp8.py for quantized models on XPU (#27533) Polisetty V R K Jyothendra Varma 2026-06-09 06:52:30 +05:30
• 6abef38627 [NPU]adaptation to support deterministic inference (#21197) Estrella-xx 2026-06-09 09:22:09 +08:00
• ab70153b62 [XPU][NIXL] Use uint64 for XPU address arithmetic in prep handle builders (#27415) Jianhong Zhang 2026-06-08 18:18:36 -07:00
• 317fc6a9dd refactor: replace oversized 1.3MB tiny_tokenizer.json fixture with a genuinely tiny byte-level BPE fixture (#27235) Matt Van HornandMatt Van Horn 2026-06-08 17:40:32 -07:00
• a09e85d677 Fix(spec): Fix the crash issue in the FA3 backend when running with top-k > 1 and page_size > 1 (#25077) Zhiy-Zhangandqiufan.zzy 2026-06-09 08:14:29 +08:00
• 3fe6bc390b [Spec] Naming cleanup: contiguous draft-loc kernel + accepted->accept (#27599) Liangsheng Yin 2026-06-08 15:04:58 -07:00
• c95179bc85 [Spec] Fuse small kenrels under gather_spec_extras (#27233) Khoa Pham 2026-06-08 15:02:04 -07:00
• b5c64b94d5 [Spec] Rename token resolver to _resolve_spec_v2_tokens; remove dead V1 helpers (#27552) Liangsheng Yin 2026-06-08 14:42:21 -07:00
• ca66e6fb5e [BCG] Support breakable CUDA graph for DeepSeek V4 DP attention (#25195) YAMYandYuwei An 2026-06-08 13:54:58 -07:00
• 801fe5e0f2 docs: sync LMSYS SGLang blog cards (#27517) 3 people 2026-06-08 13:31:30 -07:00
• dc24a26821 Fix GPT-OSS MXFP4 hidden size reshape on SM10X (#27528) Mohammad Miadh Angkad 2026-06-09 04:26:54 +08:00
• 28c1a3cb45 [Spec] Deprecate Spec V1 (#25464) Liangsheng Yin 2026-06-08 13:10:27 -07:00
• ea1d190ed0 [ROCm] dsv4: remove the redundant fp8 scale transpose-copy on decode (#27289) 3 people 2026-06-08 14:49:34 -04:00
• fca4ef9d69 Fix SWA pool resolution for EAGLE draft workers (#27491) Lianmin Zheng 2026-06-08 11:00:29 -07:00
• bcb5645629 Fix TRTLLM target verify query metadata (#27495) Lianmin Zheng 2026-06-08 10:58:48 -07:00
• 593eb2e0fa [NPU] Fix CI (#27577) Elizaveta MartirosianandElizaveta Martirosian 2026-06-08 20:44:57 +03:00
• 32bedbf88e [diffusion] model: support SANA-WM with streaming support (#27531) 4 people 2026-06-08 10:10:14 -07:00
• b047bb3e92 build(sgl-kernel): support configurable mirrors for restricted networks (#27387) Yingchun Lai 2026-06-08 23:52:26 +08:00
• 3607cbd65a [AMD] update ROCm AITER commit (#27555) Bingxu Chen 2026-06-08 23:49:10 +08:00
• 61e4132bc2 [MUSA] bump torchada version to 0.1.59 and workaround PCG limitation. (#27537) MARATRIX 2026-06-08 23:45:39 +08:00
• 12de907bc2 [MUSA][23/N] CI: Fix torchada preflight lock cleanup and add LLM server smoke test (#27242) Stella-17andxinyue.fan 2026-06-08 23:45:16 +08:00
• eb646c7b78 [srt] Add sglang:weight_load_duration_seconds gauge with source label (#27363) Leon Gao 2026-06-08 08:41:54 -07:00
• 40030d8af8 [NPU] Add GitHub test summary and deduplicate test code. Part 2 (#24689) 4 people 2026-06-08 18:08:45 +03:00
• 62c505a196 [HiCache][Dsv4] Don't cache C128 State pool in L3 (#27293) Vladislav Nosivskoy 2026-06-08 16:23:26 +03:00
• b0cd533a96 [diffusion] feat: progressive resolution growing for image and video models (#27524) Brian Chao 2026-06-08 20:44:07 +08:00
• 1ff7c627cd [UnifiedTree]: Support hicache metrics (#27554) Zhangheng 2026-06-08 20:18:02 +08:00
• 6394a8b381 [EPD] Dynamic encoder registration cleanup (#27542) Shangming Cai 2026-06-08 19:08:29 +08:00
• 1f5dc2cdca [GPTQ] Refactor CPU quantization schemes (#26786) Yaochen Hanandronnie_zheng 2026-06-08 18:14:40 +08:00
• a26587dd4e [AMD][diffusion] Add FlyDSL fused normalization kernels for ROCm diffusion models optimization (#22786) YC Yen-Ching Tseng 2026-06-08 17:42:39 +08:00
• 57ea09badb Fix NaN in triton EAGLE spec-v2 draft-extend CUDA graph at topk>1 (wrong qo_indptr stride) (#27545) Liangsheng Yin 2026-06-08 02:28:48 -07:00
• 8ff0c9fef9 [PD] Downgrade propagated rank failure logs from error to debug (#27534) Shangming Cai 2026-06-08 16:56:46 +08:00
• df6b9c2d9d [AMD] ci: reinstall MoRI if Different from Dockerfile-pinned commit during install_dependency (#27538) Bingxu ChenandCursor 2026-06-08 16:29:01 +08:00
• d03182cd2d Fix TP deadlock in unified radix cache writing_check / loading_check (#27489) Ke Bao 2026-06-08 16:08:47 +08:00
• 13dda3b8de [EPD] Support dynamic encoder register (#22253) siyu 2026-06-08 16:05:16 +08:00
• 18d728967a [PD][MoRI] Drive KV transfers with a sharded synchronous worker pool (#26922) Niko Ma 2026-06-08 15:49:25 +08:00