Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 2c183350be [diffusion] fix: fix wrong dit config for qwen-image-edit-plus-2511 (#20123) Mick 2026-03-08 20:08:36 +08:00
• 1eed219f47 add mixed chunk unit test and make small refactors (#18776) Glen Liu 2026-03-08 07:56:09 -04:00
• ab9de886c5 [diffusion] reduce LayerwiseOffloadManager reserved GPU memory (#20042) Ratish P 2026-03-08 16:56:17 +05:30
• 29f3a5396e [Minor] Add SessionSlot.is_holding_kv property for readability (#20120) Liangsheng Yinandgemini-code-assist[bot] 2026-03-08 03:25:13 -07:00
• 0f0c8b2f18 chore: bump sgl-kernel version to 0.3.21.post1 (#20087) sglang-botandsglang-bot 2026-03-08 03:03:58 -07:00
• 36b557d2c9 Fix streaming session with paged KV cache (SWA/MLA) (#20070) 3 people 2026-03-08 03:00:32 -07:00
• 230fb55899 [Performance] Decode Offload improves the long texts performance 100% through dynamic block offload. (#17216) yuyu5333andzhangheng 2026-03-08 17:16:53 +08:00
• 97a2a9be0f [VLM] Replace conv3d proj with linear for GLM4V (#20033) Yuan Luoandluoyuan.luo 2026-03-08 14:50:47 +08:00
• 7fb282a96f [diffusion] fix: fix bug of copy_if (#20094) Fan LinandYihan Chen 2026-03-08 14:27:58 +08:00
• 7f9f85d4c8 [diffusion] feat: make QwenImageLayered resolution configurable (#20044) xingsy97 2026-03-08 14:26:05 +08:00
• a73369c39f [diffusion] chore: ensure CFG Zero Star numerical stability for Helios model (#20091) Lancer 2026-03-08 14:25:14 +08:00
• 72f6dfcc31 fix: add ModelScope cache lookup and speculative path support (#20098) shuwenn 2026-03-08 14:23:16 +08:00
• d02c515ee8 Decouple scheduler log printing from metrics collection (#20107) Liangsheng Yin 2026-03-07 22:09:10 -08:00
• 69ec15ad74 [CI] Re-enable streaming session tests with correctness and stress coverage (#20084) Liangsheng Yin 2026-03-07 18:18:11 -08:00
• d28f35240a [V32/GLM5] Change default setting of V32 nvfp4 on TP4 (#20086) Baizhou Zhang 2026-03-07 15:13:25 -08:00
• 0f62da6953 [CI] Show test partition assignments after checkout (#20085) Alison ShaoandAlison Shao 2026-03-07 13:50:49 -08:00
• 7bd3dd9270 fix: image URL in notebook to use raw.githubusercontent.com (#20100) shuwenn 2026-03-08 05:28:20 +08:00
• 45bd30e29d [NPU] make torch_native lora backend a little bit faster (#17228) 4 people 2026-03-07 20:14:46 +03:00
• 5867c3fa80 Support HiCache for MambaRadixCache (#19663) Ke Baoandhzh0425 2026-03-08 00:36:25 +08:00
• 17721b00fd [AMD] Fix Tensor Memory Aliasing (#19928) Bingxu Chen 2026-03-08 00:06:10 +08:00
• 7da590d4d0 [Qwen3.5] Support Qwen3.5 Pipeline Parallelism (#19670) Yuan Luoandluoyuan.luo 2026-03-07 23:34:08 +08:00
• 13bdc7bf4a [Feature][NPU]: add runtime support for AutoRound quantized models (#16699) 3 people 2026-03-07 23:03:55 +08:00
• fd79cd8d9c [Skills] Refine jit_kernel and sgl-kernel skills (#20095) Xiaoyu Zhang 2026-03-07 22:54:01 +08:00
• 5297b02c88 [Diffusion] [NPU] Wan2.2-T2V-A14B-Diffusers modelslim quantization support (#17996) Артем Савкинandronnie_zheng 2026-03-07 17:26:44 +03:00
• f8d4eb7022 [Docs] Add docstrings to JIT kernel include headers (#19770) xingsy97 2026-03-07 20:48:00 +08:00
• ef6540b439 [diffusion]: add width/height passthrough for OpenAI image API (#19970) Ratish P 2026-03-07 18:13:46 +05:30
• 19c51fe2fa fix(rope): restore K writeback in fused rope + kv store kernel (#19636) David Wang 2026-03-07 07:41:35 -05:00
• f016738f4c fix syntax error: "&&" unexpected (#20093) kk 2026-03-07 18:06:21 +08:00
• bd108a5971 Add workaround for aiter triton gemm config issue (#20090) kkandHAI 2026-03-07 17:21:31 +08:00
• 1aa6ab41de [Nightly] Replace MiniMax-M2 with MiniMax-M2.5 (#20083) Alison ShaoandAlison Shao 2026-03-07 01:15:34 -08:00
• 43d6a32045 [sgl-kernel] rebase FlashMLA 0217 (#18902) Fan YinandBaizhou Zhang 2026-03-07 16:30:52 +08:00
• c267bdb805 [AMD] Fix AMD CI - stage-b-small-1-gpu-amd (partition 7) (#20028) YC Tseng 2026-03-07 10:49:14 +03:00
• 011806c419 [Nightly] Add Kimi K2.5 nightly test (base + Eagle3 MTP), replace Kimi K2 (#19802) Alison ShaoandAlison Shao 2026-03-06 23:44:04 -08:00
• f8bbf56de7 Refactor NemotronHConfig to canonical layers_block_type and add MTP block-type support (#19950) danielafrimi 2026-03-07 09:22:03 +02:00
• b91fb8393e [diffusion] fix: fix multi-prompt generation and support multiple prompts in cli (#19960) Lancer 2026-03-07 13:01:59 +08:00
• 31e93e4486 [diffusion] fix: fix TeaCache silently fails with --enable-teacache (#19964) Eitan Turok 2026-03-07 07:00:11 +02:00
• c584158135 [CI] Temporarily disable flaky test_priority_metrics on CUDA (#20075) Alison ShaoandAlison Shao 2026-03-06 20:34:45 -08:00
• 925185f9ec Fix flashinfer backend with pcg (#20061) Qiaolin Yu 2026-03-06 20:01:43 -08:00
• 8a411a9a2a [Tracing] Remove the deprecated tracing code from mini_lb (#19409) Feng Su 2026-03-07 11:19:23 +08:00
• f88acf8780 [JIT Kernel] Reland NVFP4 kernels to JIT (#20012) Mohammad Miadh Angkad 2026-03-07 10:31:08 +08:00
• 6ffc74efd7 [Metrics] Add overlap bubble timing, full KV usage gauge, and prefill cuda graph tracking (#19982) Yilong Zhao 2026-03-06 17:41:27 -08:00
• 50bbdcf8e9 Relax flaky test thresholds for MLA DeepSeek V3 and AutoRound (#20068) Alison ShaoandAlison Shao 2026-03-06 17:26:28 -08:00
• a0d085c16d Adding correct path for module not found error while collecting test (#19778) shubham singhalandsys-lpot-val 2026-03-07 05:56:16 +05:30
• e818f8219a Fix none-comparison (E711) warnings (#19745) R0CKSTAR 2026-03-07 08:15:21 +08:00
• 0c4f98ed4e [diffusion] hardware: add set_musa_arch on MUSA (misc, 15/N) (#19381) R0CKSTAR 2026-03-07 08:14:41 +08:00
• 069d4c577b Fix Kimi K2.5 PP layer range exposure for PD disaggregation (#19959) MARATRIX 2026-03-07 08:14:02 +08:00
• ddcecdea49 [Core] Unify max_num_reqs dp_size division for pool sizing (#20063) Liangsheng YinandClaude Opus 4.6 2026-03-06 16:12:59 -08:00
• 7a12255b6e fix: set first_token_time before computing decode_throughput for single-batch completions (#19984) Kangyan-ZhouandClaude Opus 4.6 2026-03-06 16:11:41 -08:00
• 5c8e28698c Add cleanup for _ATTN_TP in parallel_state.py (#19978) Aurick Qiao 2026-03-06 15:43:31 -08:00
• 61de303f0a Fix fallback to default tactic (flashinfer autotuner) with trtllm_fp4_block_scale_moe (#19189) Shu Wang 2026-03-06 17:15:04 -06:00
• e89069ee64 Fallback to torch.cuda.mem_get_info() when nvidia-smi is unavailable (#18957) Kangyan-ZhouandClaude Opus 4.6 2026-03-06 15:00:08 -08:00
• 604db4471d [Core] Clarify memory variable naming in model runner (#20060) Liangsheng Yin 2026-03-06 14:00:46 -08:00
• ac453b253f Add Qwen3.5-397B-A17B nightly test (8-GPU) (#19906) 3 people 2026-03-06 13:49:28 -08:00
• 7a6cf0e9ba [Core] Extract _calculate_mamba_ratio and _init_pools from init_memory_pool (#20058) Liangsheng Yin 2026-03-06 13:37:22 -08:00
• 8cdb7e1fd4 [CI] Add GPT-OSS test for SM120 (#20056) Mohammad Miadh Angkad 2026-03-07 03:43:04 +08:00
• 759700c808 Fix SM120 triton_kernels MXFP4 block_k for GPT-OSS (#20040) Mohammad Miadh Angkad 2026-03-07 02:53:08 +08:00
• de1a0afcbc [MUSA][10/N] Add GGUF support (#18357) R0CKSTAR 2026-03-07 02:50:35 +08:00
• e8f2b80340 [diffusion] improve: improve code readability of DenoisingStage (#20003) JohnHerry 2026-03-06 23:23:44 +08:00
• 54634b9a40 [Kernel] Dispatch exp/sin/cos through dtype_trait (#19798) xingsy97 2026-03-06 22:57:52 +08:00
• 2d266c73ea Migrate renorm kernels from sgl-kernel to FlashInfer JIT (#18854) Johnsonms 2026-03-06 06:53:28 -08:00
• 6d22c9f369 [Diffusion] Move hf kernels diffusion cuda kernels skills to SGLD (#20001) Xiaoyu ZhangandMick 2026-03-06 22:16:06 +08:00
• f7de9375ac [GDN][Qwen3-Next][Qwen3.5] Fuse fused_gdn_gating and fused_recurrent_gated_delta_rule_update in verify_target (#19775) Yuan Luo 2026-03-06 21:42:44 +08:00
• e3b581ce6b [diffusion] fix: remove num_frames in wan2_1_t2v_1_3b_lora_1gpu test (#20009) Prozac614anddaiweitao 2026-03-06 21:36:43 +08:00
• 25e678d933 [diffusion] endpoint: add /server_info and /model_info endpoints for gateway discovery (#20020) Kangyan-ZhouandClaude Opus 4.6 2026-03-06 05:36:13 -08:00
• 550506894a [AMD] Upgrade aiter version (#19936) Thomas WangandHaiShaw 2026-03-06 18:31:04 +08:00
• 84aaa69795 [AMD] Use bfloat16 for correction_bias in AITER FP8 path to avoid runtime dtype conversion for dsv3 (#19843) inkcherry 2026-03-06 16:57:12 +08:00
• 27053aa5ed Fix MLA decode path returning unwritten (padded) rows (#19902) Clint 2026-03-06 02:54:29 -06:00
• 0252ca8255 [Bugfix] Fix the bug blocking the startup of Llama-3.2-11b-Vision-Instruct (#19638) xdtbyndandsglang-npu-bot 2026-03-06 16:21:50 +08:00
• 23eb24ddfe [AMD] Update/fix AMD CI workflow dispatch mechanism (#20014) YC Tsengandbingxche 2026-03-06 10:27:44 +03:00
• da27d9bff6 [Bug-Fix][EPD]: skip log waiting-image-req for zmq_to_tokenzer/mooncake (#19555) Zheng Wengang 2026-03-06 14:39:22 +08:00
• 04e364d538 [V32] Enhance deepseek v32 related tests (#19985) Baizhou Zhang 2026-03-05 20:12:49 -08:00
• be9a9e4819 refactor(multimodal/test): centralize model names and shared utilities in test_utils (#19354) MookandRatish P 2026-03-05 20:09:42 -08:00
• 51e5dc845a Revert "[Kernel Slimming] Migrate NVFP4 kernels to JIT" (#20005) Baizhou Zhang 2026-03-05 19:40:00 -08:00
• 6e5a2de354 [diffusion] fix: fix reading multiple prompts from prompt file (#19075) sushil Dubey 2026-03-06 08:53:31 +05:30
• 9502369488 fix(grpc): add server-side keepalive options to prevent GOAWAY (#19986) Simo Lin 2026-03-05 18:56:35 -08:00
• 5471e4a492 [NPU][Feature] eliminate dsv3 redundant rotary embed calculation (#19842) liupeng374 2026-03-06 09:02:14 +08:00
• b912d7ae19 [OPT]Skip the first delayer to maximize the BS of the decoding. (#19836) chenxu214 2026-03-06 08:53:19 +08:00
• 261be85ecc Support mrope_position_delta cache shadowxz109 2026-03-06 08:50:53 +08:00
• 9ebffef1ef [FIX] NSA backend page_table overflow in speculative decoding target_verify (#19016) Xinyuan Tong 2026-03-06 00:04:58 +00:00
• 13af7cbb02 fix: use consistent time denominator for throughput metrics in bench_one_batch_server (#19223) Ajay Anubolu 2026-03-05 15:58:17 -08:00
• dd2bbe6d62 fix(grpc): use context.abort() with proper status codes instead of in-band errors (#19972) Chang Su 2026-03-05 14:53:18 -08:00
• 46dced64ea Adjust padding size to improve triton_kernels moe performance (#19174) Qiaolin Yu 2026-03-05 14:50:40 -08:00
• 346a4131cf [Spec] Refactor NaN/OOB checks to async maybe_detect_* with env-var control (#19899) kpham-sglandhnyls2002 2026-03-05 13:51:05 -08:00
• 1c1712d8e5 [CI] Skip flashinfer-cubin reinstall when version matches (#19470) 4 people 2026-03-05 13:30:44 -08:00
• b3cfad0a80 Add Ray actor support for scheduler process management (DP=1) (#17684) Xinyu ZhangandCursor 2026-03-05 13:21:23 -08:00
• 07e7603c0c Update sgl-attn to include SWA decode optimizations (#19655) Minglei Zhu 2026-03-05 12:45:25 -08:00
• ebb66cc1de [misc] Priority scheduling metrics cleanup (#19927) sglang-botandClaude Opus 4.6 2026-03-05 12:42:42 -08:00
• ff6048fb9c rename nemotron reasoning parser (#19865) danielafrimi 2026-03-05 21:27:07 +02:00
• e58391dd7d Add --json-log flag to enable structured JSON logging (#19968) 3 people 2026-03-05 14:24:12 -05:00
• 41fd53fe37 Fix profile_activities parameter name in bench_one_batch_server_internal.py (#19954) Mohammad Miadh Angkad 2026-03-06 02:34:06 +08:00
• d605d811fb update CODEOWNERS (#19969) Mick 2026-03-06 02:19:05 +08:00
• 203cd8eb02 [AMD] [Z-Image-Turbo Day 0] Add Z-Image-Turbo nightly test for AMD GPUs (#19733) Michael 2026-03-05 08:36:17 -08:00
• 73d272bddb Revised fix for HybridAttnBackend forward for linear attn (#19369) akhilg-nv 2026-03-05 08:05:35 -08:00
• b5edab57f2 [AMD] CI - Add MI35x nightly/PR tests for kv-cache-fp8 and allreduce-fusion (DeepSeek) (#19834) YC Tsengandbingxche 2026-03-05 18:09:57 +03:00
• 0de0d74195 [EPD][Feat]support adaptive forward (#18118) Zheng Wengang 2026-03-05 21:12:30 +08:00
• 806d41ab65 [quant] fix fp32 downcasting (#19844) StonyPortandqiuxuan.lzw 2026-03-05 17:54:59 +08:00
• 472eef4071 fa4 cleanup (#19727) Rain Jiang 2026-03-05 01:54:25 -08:00
• c36de62bfc [diffusion] fix images/edit with 2 images (#17520) Chi McIsaacandMick 2026-03-05 03:56:39 -05:00
• dbc896f204 [Test] Enhance JIT kvcache store kernel test coverage (#19630) xingsy97 2026-03-05 16:17:15 +08:00
• 727face6c2 [DLLM] Add initial radix cache support (#18724) Tiwei Bie 2026-03-05 15:24:09 +08:00