Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• f0c68fbefd [diffusion] UX: aggregate expected dtype-cast logs during weight loading (#21552) Mick 2026-03-28 09:50:40 +08:00
• 7160b6cb76 [NVIDIA] Enable automatic NUMA configuration (#19452) Trevor Morris 2026-03-27 18:44:13 -07:00
• 83997080a6 docs: flesh out MAINTAINER.md oncall lists and link GitHub profiles (#21575) Lianmin Zheng 2026-03-27 17:39:16 -07:00
• c37200f5e4 Scope streaming backlog coalescing to incremental_streaming_output mode (#21037) Vladislav NosivskoyandLianmin Zheng 2026-03-28 03:29:54 +03:00
• a27651d5e0 Remove sync when enabling return_logprob (#20972) Qiaolin Yu 2026-03-27 16:36:28 -07:00
• e2b8463c80 [fix] qwen3.5 fuse_moe_triton_tune bug (#20232) zhangxiaolei 2026-03-28 07:23:24 +08:00
• 6d48719e31 [1/n] lora support - Auto detect lora target modules (#21439) Ethan (Yusheng) SuandBaizhou Zhang 2026-03-27 16:08:36 -07:00
• 9b29131961 fix tp capture in vit cuda graph (#17255) narutolhy 2026-03-27 15:38:18 -07:00
• ec29bbb286 Split workflow for releasing runtime docker (#21563) Baizhou Zhang 2026-03-27 15:05:52 -07:00
• 4a41aec844 Fix flaky test_pp_single_node (#21564) Qiaolin Yu 2026-03-27 14:33:46 -07:00
• 38ad251738 feat: add gc_threshold arg (#21481) Muqi Liandgemini-code-assist[bot] 2026-03-28 04:42:46 +08:00
• 4e905febd2 [CI] Relax several thresholds in flaky CIs (#21562) Baizhou Zhang 2026-03-27 13:16:49 -07:00
• 9a91323c9f test: point DSV3 int8 MLA CI models to lmsys Hugging Face org (#21561) Lianmin Zheng 2026-03-27 13:04:01 -07:00
• d864622a68 [Hicache & JIT_kernel] Support page first layout & mla jit kernel (#18311) huangtingwei 2026-03-27 23:54:36 +08:00
• 30397e0a1e [rl][sgl] fix tensor mismatch after pause (#21514) Bi Xue 2026-03-27 08:02:30 -07:00
• 279e7738c5 [diffusion] fix: return None instead of raising RuntimeError when no model info found (#21319) yang1002378395-cmykandMick 2026-03-27 22:42:39 +08:00
• 9238bd08a2 [CI] Register missing jit_kernel test files (#21547) Xiaoyu Zhang 2026-03-27 19:39:08 +08:00
• 6047d2c690 [AMD] Fix AMD CI monitor GitHub API rate limit exhaustion (#21527) Bingxu Chen 2026-03-27 17:55:56 +08:00
• f83b1b73a8 [diffusion] feat: add --strict-ports option for predictable port assignment (#21320) yang1002378395-cmykand阳虎 2026-03-27 16:40:50 +08:00
• 448b528720 [AMD] Adjust AMD 4gpu partitions (#21533) YC Yen-Ching Tseng 2026-03-27 15:59:26 +08:00
• 5fc5c18bed fix(security): replace unsafe pickle.loads with SafeUnpickler for CVE-2026-3989 (#20904) zwang86 2026-03-27 00:43:41 -07:00
• 8d4fca5908 [Security] 1/N: Bind ZMQ sockets to localhost to prevent unauthenticated remote access (#21435) Khoa Pham 2026-03-26 23:33:49 -07:00
• d633ab7349 [Diffusion] Add qknorm rope fuse kernel (#21440) Xiaoyu Zhang 2026-03-27 14:27:08 +08:00
• e8d46f145c Opt jit qknorm_across_heads cuda kernel (#21503) Xiaoyu Zhang 2026-03-27 13:30:46 +08:00
• 8a56a7b04d [jit_kernel] Migrate cast (downcast_fp8) from sgl-kernel AOT to JIT (#19103) Johnsonms 2026-03-26 22:21:44 -07:00
• c531be455e [jit_kernel] Add fused_qknorm_rope JIT kernel (#19059) JohnsonmsandXiaoyu Zhang 2026-03-26 22:21:28 -07:00
• 0138129d3c [CI] Fix nemotron nvfp4 test estimated time (#21516) Baizhou Zhang 2026-03-26 21:53:09 -07:00
• eaf392b9cc Remove redundant DeepSeek V3 FP4 PCG test (#21485) Mohammad Miadh Angkad 2026-03-27 12:52:47 +08:00
• 1487f80158 chore: bump mooncake version to 0.3.10 (#20942) Shangming Cai 2026-03-27 10:35:31 +08:00
• d7c4c57ace [diffusion] refactor: move format-specific weight loading hooks (quant-related) to a dedicated file (#21366) Mick 2026-03-27 09:58:49 +08:00
• e1ee68d0fc Release mm features on session close and support multiple /rerun-ut specs (#21501) Liangsheng Yin 2026-03-26 18:31:29 -07:00
• c2b3e42ad6 Fix sessions with mm inputs (#21269) Aurick Qiao 2026-03-26 17:38:23 -07:00
• 8a4cdcd538 Simplify flush_cache: reject concurrent requests, remove client-side retry (#21490) Liangsheng Yin 2026-03-26 16:31:04 -07:00
• 9dc266adb4 Fix concurrent /rerun-ut posting duplicate workflow URLs (#21495) Liangsheng Yin 2026-03-26 16:26:00 -07:00
• c580ddd19d Fix benchmark generating empty prompts when random_input_len is small (#21492) Liangsheng Yin 2026-03-26 16:24:35 -07:00
• a93065679b Revert "bugfix for weight loading for qwen3-next" (#21496) Baizhou Zhang 2026-03-26 16:17:18 -07:00
• 2e65c27b29 Api add flush cache timeout (#21413) SevenJ 2026-03-27 05:44:37 +08:00
• 8c3ccef2d9 Fix Kimi K2.5 dp attention+ spec decoding launch crash (#21391) Qiaolin Yu 2026-03-26 14:40:26 -07:00
• be0cca5596 Use torch.addmm instead of separate mm and add_ calls for LoRA torch.native (#20562) satyamk7054andSatyam Kumar 2026-03-26 14:35:20 -07:00
• e59ea4f6e9 fix: torch-native LoRA for multi-adapter case (#20564) satyamk7054andSatyam Kumar 2026-03-26 14:34:16 -07:00
• fb90c9d298 [Test] Consolidate eval accuracy test mixins into eval_accuracy_kit (#21047) Liangsheng Yin 2026-03-26 14:26:46 -07:00
• e5dd411f64 [CI] Add ci-workflow-guide skill and consolidate CI docs (#21429) Liangsheng Yin 2026-03-26 13:50:13 -07:00
• e5b7650353 Fix UnboundLocalError when DetokenizerManager constructor fails (#21471) Liangsheng Yin 2026-03-26 13:00:16 -07:00
• 17f43d1518 [CI] Skip stage health check for release branch cut (#21488) Liangsheng Yin 2026-03-26 12:54:51 -07:00
• 4b5f63e1b8 FIX: (NSA) Compute topk_indices_offset when NSA prefill flashmla_sparse is used with FP8 KV cache (#20606) Ho-Ren (Jack) Chuang 2026-03-26 12:50:50 -07:00
• 3867c6431a Fix bug in dbrx model (#21445) jianzhao-xuandJianzhao Xu 2026-03-27 02:23:30 +08:00
• 646573e4e8 fix: use get_rope_config() to support models without rope_parameters (#21135) shuwenn 2026-03-27 02:22:12 +08:00
• 0906e45cec bugfix for weight loading for qwen3-next (#21313) McZyWu 2026-03-26 21:21:00 +08:00
• 8febc3650d [NPU] multimodal-gen-test-8-npu-a3,Cache pytorch dependency (#21470) monkeyLovedingandKelon 2026-03-26 19:35:34 +08:00
• 013fa55630 [AMD] CI - fix amd 4-GPU PR test (#21444) YC Yen-Ching Tseng 2026-03-26 18:16:33 +08:00
• 238a4b8f8f [diffusion] CI: fix breaking import path in nightly (#21449) Mick 2026-03-26 16:33:22 +08:00
• 35720d9969 [diffusion] fix: fix qwen-image with nunchaku (#21415) Mick 2026-03-26 16:31:44 +08:00
• f289d173aa [Deps] Bump xgrammar to 0.1.32 (#21032) Anant Sharma 2026-03-26 04:22:37 -04:00
• fd535942ac [AMD]Integrate aiter's fused_topk for softmax scoring in topk function (#21421) Chen, ZhentaoandChen, Todd 2026-03-26 15:57:56 +08:00
• a305964159 [MLX] Add native MLX execution backend for Apple Silicon Mac (#20342) R0CKSTAR 2026-03-26 15:09:17 +08:00
• 7ca015fe65 [Diffusion] Refactor diffusion JIT kernel test layout and narrow CI triggers (#21385) Xiaoyu Zhang 2026-03-26 15:02:02 +08:00
• 79db3bec34 [CI] Add PID namespace and ps auxf diagnostics to killall.py (#21401) Liangsheng Yin 2026-03-25 23:57:15 -07:00
• 176c8e7b0e [CI] Simplify CI check actions and add maintenance gate to rerun-ut (#21455) Liangsheng Yin 2026-03-25 23:28:18 -07:00
• 01ccdb91b1 [Fix] Add EPLB rebalance support for Kimi K2.5 (#21004) MARATRIX 2026-03-26 12:01:40 +08:00
• f142608408 [MUSA] apply_vocab_mask support musa device (#21296) ori 2026-03-26 12:00:58 +08:00
• f420b9b4a5 [MUSA][Feature] Enable Piecewise CUDA Graph support for MUSA platform (#20758) MARATRIXandR0CKSTAR 2026-03-26 12:00:28 +08:00
• abf4f1a47a [MPS] Add StreamContext stub (#20782) R0CKSTAR 2026-03-26 11:59:51 +08:00
• 02521420b3 [MPS] Support sglang.check_env (#20753) R0CKSTAR 2026-03-26 11:59:25 +08:00
• d9e96153de [NPU] Support Hybrid KV Cache for Ascend backend (#18032) gjsheuandgengjinsong 2026-03-26 11:27:36 +08:00
• b835309f0c Reland: compute M-RoPE positions for preprocessed VL inputs (#21244) Simo Lin 2026-03-25 20:12:43 -07:00
• bb29893689 [Fix] Try to fix nvcc compilation error (#21246) DarkSharpness 2026-03-26 10:59:36 +08:00
• a34e9ed64a Add adjusted_filter_batch (#21260) Aurick Qiao 2026-03-25 19:59:05 -07:00
• 53c1d8e963 Fix customized_info offset truncation (#21262) Aurick Qiao 2026-03-25 19:57:51 -07:00
• 1100b9865c Fix MxInt4 MoE returning wrong output variable (#21348) Sam Shleifer 2026-03-25 22:57:09 -04:00
• 662635e7a7 fix(sgl-kernel): align wheel METADATA/WHEEL with +cu filename (#21437) Yanda Cheng 2026-03-25 22:44:50 -04:00
• 80389fec00 [AMD] Fix AMD CI: mark /sglang-checkout as git safe.directory in container (#21423) MichaelandBingxu Chen 2026-03-25 19:18:19 -07:00
• 6f2b51ade1 [Diffusion] Optimize diffusion Triton rotary embedding by processing multiple heads per token (#21387) Xiaoyu Zhang 2026-03-26 08:59:25 +08:00
• dbe871efdd Rollback flashmla to older version [1/2] (#21430) Baizhou Zhang 2026-03-25 17:49:54 -07:00
• 7c7b2a8c97 [Bugfix] Lazy-import CuteDSL KDA kernel to fix AMD/ROCm startup crash (#21428) Hubert Lu 2026-03-25 16:37:26 -07:00
• 75682f1d2f Remove noisy streaming backlog warning log (#21432) Liangsheng Yin 2026-03-25 16:25:16 -07:00
• 4dd4e06f1d [CI] Fix resource leak when setUpClass fails (#21338) Liangsheng Yin 2026-03-25 16:22:44 -07:00
• d5c5683d2b [CI] Use ETag conditional requests in wait-for-jobs and add CI infra to check-changes (#21345) Liangsheng Yin 2026-03-25 14:40:07 -07:00
• a12fea21ed perf(sgl-kernel): expose get_scheduler_metadata for FA3 decode optimization (#21103) Minglei Zhu 2026-03-25 13:17:27 -07:00
• e90cba715c Revert "[Bugfix] Disable ci for .md files" (#21420) Baizhou Zhang 2026-03-25 12:00:20 -07:00
• f1f13f8966 [Bugfix] Disable ci for .md files (#21410) Артем Савкин 2026-03-25 20:19:44 +03:00
• 77872a8d55 Update Nemotron Example docs to include Super v3 and Nano 4B (#21416) Nave Assaf 2026-03-25 18:03:19 +02:00
• 68f7f00174 [Diffusion] Speed up Qwen select01 Triton modulation kernels (#21318) Xiaoyu Zhang 2026-03-25 20:48:39 +08:00
• 04eb72801f [diffusion] CI: add performance tracking job to nightly (#21091) Mick 2026-03-25 19:01:33 +08:00
• 689e9ef05c [Diffusion] Add AKO4ALL kernel optimization skill (#21323) Xiaoyu Zhang 2026-03-25 18:46:21 +08:00
• e4ad10520b [diffusion] Skip automatic Wan/MOVA DiT layerwise offload on high-end GPUs (#21248) Xiaoyu Zhang 2026-03-25 18:45:30 +08:00
• 3d2a61cbf6 [Chore] Clean up JIT compilation flags (#21022) DarkSharpness 2026-03-25 18:08:40 +08:00
• 4480e6c237 [CI] Add retry loop to killall_sglang GPU cleanup verification (#21393) Liangsheng Yin 2026-03-25 02:16:20 -07:00
• c494e47843 [AMD] Fix stage-b-test-small-1-gpu-amd (test_tool_choice.py) (#19868) YC Yen-Ching Tseng 2026-03-25 16:10:21 +08:00
• 6425df5c8a [diffusion] doc: consolidate documentation (#21373) Mick 2026-03-25 16:01:32 +08:00
• f5c225eeba [CI] Fix TestQwen35WithHiCache (#21371) Baizhou Zhang 2026-03-25 00:04:59 -07:00
• 2d583799eb Update ascend docs (#20846) amote-i 2026-03-25 14:58:44 +08:00
• 5297a3cb46 [CI] Rewrite killall_sglang as Python with CI/local dual mode (#21331) 5 people 2026-03-24 23:54:01 -07:00
• 6cc5717e8a [diffusion] doc: update quantization.md (#21356) Mick 2026-03-25 14:48:38 +08:00
• 6b0f1e3b43 Update skip condition for TestQwen35PPAccuracy (#21370) Shangming Cai 2026-03-25 14:28:42 +08:00
• 17e41cfb21 Fix RDMA device mapping for non-zero GPU indices in disaggregation tests (#21303) 3 people 2026-03-24 22:56:57 -07:00
• 61a902ce88 [AMD][MoRI] Auto-select dispatch quantization type from MoE weight dtype. (#21040) Duyi-Wang 2026-03-25 13:53:57 +08:00
• 86e2622097 [AMD] Add mha fp8-kv support (#21253) kkandwunhuang 2026-03-25 13:38:02 +08:00
• 2b75fed0dd Workaround of DSA performance drop on B200 + DP (#21337) Baizhou Zhang 2026-03-24 22:21:07 -07:00
• d937d01fe6 [CI] Fix cancel workflow: use bypass-maintenance label (#21364) Liangsheng Yin 2026-03-24 21:43:00 -07:00
• c769398748 [CI] Add include_high_priority checkbox to cancel workflow (#21363) Liangsheng Yin 2026-03-24 21:30:44 -07:00