Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• e9d6b9eb2d [HiCache & HybridModel] mooncake backend support DSA & mamba model (#21259) 5 people 2026-04-14 09:47:36 +08:00
• cc449ac4e5 feat(metrics): expose raw KV cache pool token counts as prometheus gauges (#22726) ishandhanani 2026-04-13 20:30:36 -05:00
• 945d73824f [HiSparse] Clarify decode token usage logs (#22331) huangtingwei 2026-04-14 09:03:25 +08:00
• c456cba7fd [gateway] Support SGLANG_LOG_MS for millisecond precision in router logs (#22506) Zhai Feiyue 2026-04-14 08:28:00 +08:00
• 1ec018f27a [Feature] Add SiMM as sglang HiCache Storage backend (#18016) yuki-brook 2026-04-14 08:12:37 +08:00
• ff61b2e470 [CI] Add workflow_dispatch and environment gate to GB200 nightly pipeline (#22733) Sahithi Chigurupati 2026-04-13 17:08:18 -07:00
• 33a3ba256f Delete dead rematch path in SessionAwareCache.release_session (#22735) Liangsheng Yin 2026-04-13 17:02:40 -07:00
• 9fb00ede15 Clean up TokenizerManager and req_time_stats: reduce overhead and simplify (#21646) Lianmin Zheng 2026-04-13 16:47:32 -07:00
• a2b5111962 perf: skip KV cache in FA backend for embedding mode (#21971) Jia GuoandClaude Opus 4.6 2026-04-13 16:27:52 -07:00
• 8f9553bccb [Misc] Migrate SGLANG_SET_CPU_AFFINITY to envs and refactor model config building (#22730) Lianmin Zheng 2026-04-13 16:10:31 -07:00
• f4f9e68189 [AMD] Add MoE weights and scales padding (#21097) mqhc2020andHAI 2026-04-14 06:50:15 +08:00
• b1efce342c env: add knob to control SWA eviction interval (#22645) Yilong Zhao 2026-04-13 15:37:59 -07:00
• f81b6e8f51 [Misc] Add @cache_once to is_arch_support_pdl in jit_kernel (#22724) Lianmin Zheng 2026-04-13 14:42:49 -07:00
• b441317aa4 Revert "Upgrade CI default CUDA version from 12.9 to 13.0" (#22727) Baizhou Zhang 2026-04-13 14:39:24 -07:00
• ba7bcca6b3 Use reshape instead of contiguous().view() in TRTLLMHAAttnBackend (#22517) Lianmin Zheng 2026-04-13 14:29:12 -07:00
• ff13dfee45 [lora][moe] Virtual experts for LoRA MoE (#22122) Kurt ShusterandYusheng Su 2026-04-13 17:19:30 -04:00
• 6b2bf66cd9 fix[glm4.7 flash]: properly detect gfx95_quant_format (#22720) ishandhanani 2026-04-13 15:10:07 -05:00
• f51ce2c92f Update CODEOWNERS for musa/mlx (#22593) R0CKSTAR 2026-04-14 03:41:07 +08:00
• 90ef8ce54d [Docker] Remove flashinfer cache copy (#22653) Mohammad Miadh Angkad 2026-04-14 00:48:22 +08:00
• 39810762d2 fix: use describe mode for SGLang version detection (#22600) Asish Kumar 2026-04-13 22:15:45 +05:30
• 13a4aafdbe [NPU] update glm5 running guide (#22712) 看海的人 2026-04-13 22:53:24 +08:00
• c6403a11cb Modify the optional values and constraints of parameter. (#22705) chx96642264 2026-04-13 22:50:48 +08:00
• b6a91b1afe [NPU] --attn-cp-size --init-expert-location --eplb-algorithm parameter docs update (#22704) jianzhao-xuandJianzhao Xu 2026-04-13 22:42:34 +08:00
• 8d904e50f2 [NPU]qwen3-8b and 32b md bugfix (#22687) Liwansi 2026-04-13 22:20:17 +08:00
• 2089ac86a7 Improve parameters usage constraints for npu deployment (#22700) loading66andh30064329 2026-04-13 22:02:56 +08:00
• 56c97c7738 [NPU] update npu doc (#22697) 看海的人andzhsurpass 2026-04-13 21:55:38 +08:00
• d01b2bf257 [Docs] Fix default values and options in Ascend server arguments documentation (#22698) xdtbyndandxdtbynd 2026-04-13 21:22:37 +08:00
• 314d6ecf08 [Feature][JIT Kernel] Fused TP QK norm For Minimax (#20673) DarkSharpnessandMingyang Jiang 2026-04-13 20:29:47 +08:00
• 4df60434d7 [diffusion] model: support stable-diffusion-3-medium-diffusers (#19225) 4 people 2026-04-13 01:07:06 -07:00
• 9e6d1c066e TestStreamingSessionAbortLeakRepro: inherit stdout/stderr instead of tempfile (#22668) Liangsheng Yin 2026-04-13 00:37:29 -07:00
• 4a746ea462 [AMD] Remove aiter hotfixes in Dockerfile covered by aiter v0.1.12.post1 (#22657) Thomas Wang 2026-04-13 15:01:37 +08:00
• 1e9eecfa36 [Intel GPU] Enable sgl-kernel-xpu fused_experts MoE kernel path for GPT-OSS bf16 models. (#22417) Chandrakant Khandelwal 2026-04-13 11:15:48 +05:30
• d524f110ac [diffusion] refactor: streamline denoising stages (#22633) Mick 2026-04-13 13:34:37 +08:00
• 7d2c11970c [Intel GPU] Upgrade pytorch xpu version to 2.11 (#21908) Polisetty V R K Jyothendra VarmaandMa Mingfei 2026-04-13 10:46:24 +05:30
• 5549d910c6 [hisparse]: Adding ci for hisparse kvcache-swap-in jit-kernel (#22155) Zhangheng 2026-04-13 12:50:29 +08:00
• 305b42935a [HiSparse]: Add benchmark for hisparse kernel (#22187) Zhangheng 2026-04-13 12:49:18 +08:00
• 3f4fbc165d Upgrade CI default CUDA version from 12.9 to 13.0 (#21441) Alison Shao 2026-04-12 23:48:40 -05:00
• 4dbd59850b Add bfloat16 KV cache validation for HiSparse (#22505) Mohammad Miadh Angkad 2026-04-13 12:41:42 +08:00
• fae0a2fc3c [codex] Add LTX-2.3 benchmark skill recipes (#22631) Xiaoyu Zhang 2026-04-13 12:23:32 +08:00
• c775c31cb3 Simplify test_chunked_prefill; remove redundant tests (#22652) Liangsheng Yin 2026-04-12 20:26:54 -07:00
• bf022e177c Revert "[Diffusion] Add FLUX.1-dev ModelOpt NVFP4 support (#22574)" (#22649) Mick 2026-04-13 11:17:32 +08:00
• bc59cc0f96 [RaidxTree Refactor]: Support Unified HybridRadixTree V2 (#21206) 5 people 2026-04-13 10:28:22 +08:00
• 5593539942 [RL] Refactor NVFP4 shuffling/swizzling to in-place replacement (#22204) Ziang Li 2026-04-12 19:08:45 -07:00
• 934e19a610 [CPU] Fix argument issues in qkv_proj_with_rope_fused_weight and bmm… (#21367) blzhengandMa Mingfei 2026-04-13 09:59:13 +08:00
• da6b8e1448 Extract pause_resume_in_place kit; rename test_abort to test_scheduler_control (#22647) Liangsheng Yin 2026-04-12 18:49:37 -07:00
• 28e40d873c fix(PD): respect pause_generation in disagg event loops (#20908) Lawrence Wu 2026-04-12 18:07:51 -07:00
• c1ab68b45e fix: streaming session race condition + some metrics (#21875) 4 people 2026-04-12 20:05:23 -05:00
• 37fc47c645 diffusion: fix layerwise offload for ModelOpt quantized DiTs (#22594) Xiaoyu Zhang 2026-04-13 08:01:54 +08:00
• 03a1a7b81c [Diffusion] Add FLUX.1-dev ModelOpt NVFP4 support (#22574) Xiaoyu Zhang 2026-04-13 07:57:41 +08:00
• f81b6df3a3 [lora] Fix partial MoE rank loading, VL lm_head, strict loading, deepseek on-demand (#21864) Kurt ShusterandYusheng Su 2026-04-12 19:25:02 -04:00
• 1f8df97054 Fix broken streaming response with --incremental-streaming-output (#22549) Khoa PhamandClaude Opus 4.6 2026-04-12 15:05:58 -07:00
• 701a0e0c25 [CI/Docker] Clean up redundant flashinfer cubin downloads (#22491) Mohammad Miadh Angkad 2026-04-13 03:30:41 +08:00
• d4ad30b94c [diffusion] quant: enable modelopt quantized FLUX deployment (#20082) ZhiyuandClaude Opus 4.6 2026-04-12 08:35:33 -07:00
• 495ef8ec64 [diffusion] model: support LTX2.3 two stage (#22182) Mick 2026-04-12 22:15:57 +08:00
• 31453bb76a [RL] Fix weight update for mxfp8 flashinfer_cutlass gemm backend (#22484) Ziang Li 2026-04-12 06:02:17 -07:00
• bcc0c65aa8 [DSA] Hopper FP8 FlashMLA KV padding (#22372) Mohammad Miadh Angkad 2026-04-12 17:19:17 +08:00
• 0e0091c6c8 [server] Add --quantization unquant to explicitly opt out of quantization (#21863) Kurt Shuster 2026-04-12 05:17:22 -04:00
• 4dfc8e1c3f VLM: support passing --mm-process-config for all models (#18467) Wenyao Gao 2026-04-12 02:08:05 -07:00
• f1eb4ca90c Fix streaming session busy check double-counting; add compat CI tests (#22213) Liangsheng Yin 2026-04-12 01:48:16 -07:00
• 3f60df8012 [Test] Fix flaky test_function_call_specific by adding strict=True (#22624) Xinyuan Tong 2026-04-12 09:22:57 +01:00
• bc1bfbf607 Fix swa input length limitation (#22597) Ke Bao 2026-04-12 16:03:35 +08:00
• f2377a00cb Add SWA support for runtime busy memory check (#21499) Liangsheng Yin 2026-04-12 00:39:51 -07:00
• 19cb918653 [Not-Merge][AMD] GLM-5 performance optimization (#21166) wufann 2026-04-12 14:58:11 +08:00
• edaa5973d4 [AMD][No-Merge] Simplify fused allreduce + RMSNorm and remove hidden_dim allowlist (#21986) Hubert LuandHAI 2026-04-11 23:47:08 -07:00
• 9a4e8089ff [Whisper] Batch encoder forward for concurrent prefill requests (#22361) Xinyuan Tong 2026-04-12 07:15:14 +01:00
• 52750129ef fix prefill tps log accuracy (#22497) Zhai Feiyue 2026-04-12 14:07:30 +08:00
• c6fd9a00c7 [tokenizer] eliminate O(n²) copy in non-incremental streaming (#22567) Alex NailsandClaude Opus 4.6 2026-04-11 23:05:36 -07:00
• 45472d70cc [diffusion] CI: dynamic load-balanced partitioning for diffusion CI (#15528) 3 people 2026-04-12 13:02:43 +08:00
• d6c9d9116b [CI] Update B200 est_times to prevent timeouts on slower machine (#22609) Alison ShaoandAlison Shao 2026-04-11 23:40:34 -05:00
• f21d23a211 ci: use local NVIDIA wheels to avoid re-downloading ~2GB every CI run (#22602) Alison ShaoandAlison Shao 2026-04-11 23:32:51 -05:00
• 870a21bf39 [CI] Remove Slack bot from CI failure monitor (#21581) Alison ShaoandAlison Shao 2026-04-11 22:34:48 -05:00
• cd2b2364ff [Bugfix] fix model_config deletion (#22281) Aurick Qiao 2026-04-11 20:24:26 -07:00
• 2e70e4f4f6 [CI] Little renaming of gb200 CI workflow (#22608) Baizhou Zhang 2026-04-11 17:52:42 -07:00
• 8da1cfb30d [lora][moe] Decoupled LoRA MoE backend with Marlin support (#21858) Kurt Shuster 2026-04-11 17:59:27 -04:00
• 78043d4448 [Misc] [MXFP8] Drop sm100 mxfp8 warning (#21881) Ziang Li 2026-04-11 04:11:28 -07:00
• 61a62c6503 [mem] Flatten memory checkers into composable per-pool invariant checks (#22562) Liangsheng Yin 2026-04-11 02:56:22 -07:00
• 8cca9747f5 [diffusion] model: support ERNIE-Image (#22439) dyhsup 2026-04-11 17:18:11 +08:00
• 7ab94e438c [Test] Fix flaky test_function_call_required by adding strict=True (#22586) Liangsheng Yin 2026-04-11 02:14:22 -07:00
• d14d368191 [Kernel] Set sgl_per_token_group_quant_8bit_v2 as default choice (#22467) Baizhou Zhang 2026-04-11 01:59:57 -07:00
• 4f45472f34 [NPU][Doc] add qwen3-30b-a3b low latency example (#22446) heziiop 2026-04-11 15:52:47 +08:00
• f855a0bde6 Introduce CUDA graph debug mode with breakable CUDA graph (#19102) 4 people 2026-04-11 00:36:56 -07:00
• d11da2403c Add hisparse staging + decode offload guards to is_fully_idle() (#22577) Liangsheng Yin 2026-04-11 00:12:10 -07:00
• e26c73c4e9 [diffusion] platform: support Intel XPU (#17920) sushil DubeyandMa Mingfei 2026-04-11 12:39:02 +05:30
• 3ce72252de [AMD] Fix Timeout: stage-b-test-2-gpu-large-amd,stage-b-test-1-gpu-large-amd (#22228) YC Yen-Ching TsengandHAI 2026-04-11 13:55:44 +08:00
• cf1436d6ae [AMD] Diffusion - Enabel rocm miopen tuning on vae (#22428) YC Yen-Ching Tseng 2026-04-11 13:47:25 +08:00
• 7e4e1dcd7a [AMD] Fuse RMSNorm + FP8 per-token quant for GLM-4.7-FP8 (#21403) Jacob0226andClaude Opus 4.6 2026-04-11 13:45:31 +08:00
• aeeff58cd4 [Spec][Ngram] Clean up unused stateless batchMatch (#22487) Khoa Pham 2026-04-10 21:52:56 -07:00
• 04bd8e1218 [Spec][Ngram] Return token counts in list_external_corpora API (#22471) Khoa PhamandClaude Opus 4.6 2026-04-10 21:50:02 -07:00
• 3c46ff2ac5 fix: restore CPU flash_attn test to use sgl_kernel directly (#22573) Baizhou ZhangandClaude Opus 4.6 2026-04-10 21:39:20 -07:00
• f2af00d05a [HiSparse-pd] Add device-buffer budget and fix logical pool admission in decode side (#22453) Zhangheng 2026-04-11 12:30:38 +08:00
• 8eac618a8d [tokenizer] lazy text accumulation + use deltas directly for streaming (#22548) Alex NailsandClaude Opus 4.6 2026-04-10 21:26:04 -07:00
• c7f93a2ce7 [metrics] Add PoolStats.update_scheduler_stats to deduplicate metrics assignment (#22559) Liangsheng Yin 2026-04-10 21:04:18 -07:00
• d30b3efa84 [sgl] _ATTN_TP and _ATTN_CP use message queue for broadcast on CPU (#22205) Bi Xue 2026-04-10 20:52:49 -07:00
• 7c6db40540 Fix tool call constrained decoding and parsing for models with native formats (#21593) Xinyuan Tong 2026-04-11 04:37:23 +01:00
• c2821dfbe9 [mem] Introduce PoolStats dataclass; unify pool metrics and token_usage (#22554) Liangsheng Yin 2026-04-10 20:35:50 -07:00
• 6cd183ff6b Remove redundant test_page_size.py (#22571) Liangsheng Yin 2026-04-10 20:35:04 -07:00
• 16f306fd85 [VLM] GPU Image Preprocessing for Kimi-K2.5 (#22368) Yuhao Yang 2026-04-11 11:13:30 +08:00
• 58f863956c cuda graph: adjust capture time num-non-padded-tokens to align capture with replay (#22404) Yilong Zhao 2026-04-10 19:27:50 -07:00
• 0b4f5c9fcb [diffusion] CI: improve readability and fix bug of early-return (#22507) Mick 2026-04-11 10:08:44 +08:00
• f41c810a2d [misc] update CI_PERMISSIONS.json (#22570) Qiaolin Yu 2026-04-10 18:58:55 -07:00