Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 76bc07a335 Move swa memory pool to a seperate file (#16347) Ke Bao 2026-01-04 22:39:30 +08:00
• b328cd20bb Skip local attn init metadata for mimo swa model (#16349) Ke Bao 2026-01-04 22:38:36 +08:00
• bf32cd8397 [npu] update model and feature supported for ascend npu (#16390) Hexq0210 2026-01-04 21:50:20 +08:00
• 27a0830593 Tiny add manual policy benchmark to CI (#16392) fzyzcjy 2026-01-04 19:10:35 +08:00
• 53479e226b Tiny refactor router benchmark github actions (#16374) fzyzcjy 2026-01-04 18:00:12 +08:00
• ff978e7d4c Update readme (#16391) Lianmin Zheng 2026-01-04 01:57:54 -08:00
• 16e00651dd fix: health check adding test delays (#16381) Douglas Yang 2026-01-04 00:28:32 -08:00
• 1b2b95d8d1 Super tiny cleanup unused function (#16376) fzyzcjy 2026-01-04 15:08:12 +08:00
• 9cac3c8628 Add liusy58 into CI_PERMISSION (#16373) Shangming Cai 2026-01-04 13:36:35 +08:00
• 8d58b3dcdd [model-gateway] Clear architectual debt in responses API (#16359) Chang Su 2026-01-03 21:35:28 -08:00
• 5f3eb377e0 [VLM] Support request level max_dynamic_patch for OpenAI request (#16268) Yuan Luoandluoyuan.luo 2026-01-04 13:04:43 +08:00
• 229938805f Make personal configs optional in SGLang's official docker image. (#16365) Liangsheng Yin 2026-01-04 12:27:13 +08:00
• d7aa0ce72f Support multiple engines for router simulation in schedule simulator (#16368) fzyzcjy 2026-01-04 11:46:05 +08:00
• e797f0c570 Support offline generation scenario for prefill delayer (#16363) fzyzcjy 2026-01-04 11:05:10 +08:00
• 5d4b7c78bf Fix memory leak in prefill delayer (#16358) fzyzcjy 2026-01-04 10:11:18 +08:00
• 9bd64d739b [VLM] Add doc for ViT CUDA Graph (#16343) Yuan Luoandluoyuan.luo 2026-01-04 10:05:23 +08:00
• 24e116ef6a Enhance schedule simulator and add several e2e scenarios (#16357) fzyzcjy 2026-01-04 10:02:38 +08:00
• 8ca9597057 Tiny support sticky routing algorithm in schedule simulator (#16355) fzyzcjy 2026-01-04 08:36:52 +08:00
• 216ea910ad Tiny support forwarding SMG routing key to engine for dumping and avoid string allocation (#16352) fzyzcjy 2026-01-04 08:33:35 +08:00
• f4ab2ec5be Add unified metrics collection framework (v1) (#16064) Hudson Xingandgemini-code-assist[bot] 2026-01-04 08:30:38 +08:00
• 25fa2ac290 Convert cu_seqlens to CPU for npu_flash_attention_unpad operator (#15434) fy 2026-01-04 08:16:42 +08:00
• ef5ac6f01e Support shared prefix (gsp) in schedule simulator (#16353) fzyzcjy 2026-01-04 08:03:16 +08:00
• c88aaf22c7 Support in-flight request age metrics for router (#16341) fzyzcjy 2026-01-04 07:38:53 +08:00
• e139d2aa76 Super tiny fix CI (#16351) fzyzcjy 2026-01-04 07:24:46 +08:00
• 2337b1bbb0 Refactor and fix prefill delayer (scheduler enhancer) (#16269) fzyzcjy 2026-01-04 07:15:06 +08:00
• 7bc13c906c [grpc] update api to scheduler in grpc request manager (#16350) Chang Su 2026-01-03 10:54:31 -08:00
• 877c8e3a96 Tiny refactor router test contexts (#16340) fzyzcjy 2026-01-04 01:44:10 +08:00
• 66dfb8c156 Tiny fix non-PD router http header missing whitelist (#16339) fzyzcjyandgemini-code-assist[bot] 2026-01-04 01:43:02 +08:00
• dcacc492d0 [diffusion] fix: fix RuntimeError in SageAttention3 on Blackwell with Qwen-Image (#16335) 3 people 2026-01-03 09:12:53 -05:00
• 87ef05e2e1 Support simple schedule simulator (#16344) fzyzcjy 2026-01-03 22:05:40 +08:00
• b65c9889a0 Fix incorrect running batch size in prefill stats (#15941) fzyzcjy 2026-01-03 20:49:54 +08:00
• c7c0d97fc6 Tiny support whitelisted headers in request logging (#16342) fzyzcjy 2026-01-03 20:10:58 +08:00
• 6bc5a52fd2 [NPU] Adapt qwen3-next W8A8 on NPU (#16164) shengzhaotian 2026-01-03 19:41:20 +08:00
• 2c09de343e [diffusion] improve: skip loading vision module for text encoders (#16304) Mick 2026-01-03 19:30:45 +08:00
• 38d48de93d [diffusion] CI: simplify warmup (#16303) Mickandgemini-code-assist[bot] 2026-01-03 18:09:38 +08:00
• 7f2fa2167b Tiny add --log-requests-target (#16338) fzyzcjy 2026-01-03 17:28:27 +08:00
• d0fb24ee7b [Diffusion] Flux2 tp support (#16219) Xiaoyu ZhangandMick 2026-01-03 17:02:27 +08:00
• 65b0b5b248 fix(logging): use Display format for model_id instead of Debug (#16337) Simo Lin 2026-01-03 00:21:39 -08:00
• 9a414b164c [Performance] Optimze the performance of Qwen25VL (#15640) Siyuan ChenandXinyuan Tong 2026-01-03 15:15:36 +08:00
• 5b4f790200 [diffusion] CI: add CI validation for diffusion model downloads (#16311) Alison Shao 2026-01-02 22:12:45 -08:00
• d8ac5eecf7 [model-gateway] bug fix on module name (#16332) Chang Su 2026-01-02 21:56:18 -08:00
• bdde949619 [HiCache] Add PP Support with suffix pp rank (#15175) 4 people 2026-01-03 13:49:21 +08:00
• b23e7ed13c refactor(gateway): shorten logging targets from sgl_model_gateway to smg (#16328) Chang Su 2026-01-02 21:12:33 -08:00
• 9821fae5c6 Tiny add explanations to realtime token metric labels (#16009) fzyzcjy 2026-01-03 12:36:55 +08:00
• 078d96213a [FEAT] optimize tensor zmq transfer for multimodal inputs (#13592) 3 people 2026-01-03 12:23:23 +08:00
• 8b111b20c3 feature: improvements to CI failure monitor (#16272) Douglas Yang 2026-01-02 20:09:41 -08:00
• 74a166cb86 [Fix] Only add SM90 and SM100 to check for auto-enabling TRT Allreduce Fusion (#16283) b8zhong 2026-01-02 19:43:17 -08:00
• 888e126ac9 [diffusion] comment: fix typo (#16257) triple-muandMick 2026-01-03 12:42:04 +09:00
• bb23a8fe77 [Tiny]Remove progress bar for fp8 ue8m0 quant when unneeded (#16177) Baizhou Zhang 2026-01-03 10:54:12 +08:00
• 8b869e326c [AMD] feat: add DLLM support for AMD GPUs with LLaDA2 testing (#15560) sunxxuns 2026-01-02 18:41:11 -08:00
• 6256936d09 Fix: Allow build-wheels to run when target_stage is set (#16324) Alison Shao 2026-01-02 17:58:11 -08:00
• 62f73a8c81 [Auto Sync] Update engine.py (20260102) (#16317) Lianmin Zhengandgithub-actions[bot] 2026-01-02 17:29:40 -08:00
• 7c1b4b1c4c Support swa allocator page size > 1 (#16296) Ke Bao 2026-01-03 08:49:47 +08:00
• 31ed68e7c1 [model-gateway] address embedding similarity threshold in ut (#16321) Simo Lin 2026-01-02 16:46:55 -08:00
• 24c91001cf [model-gateway] code clean up in tokenizer register step workflow (#16316) Chang Su 2026-01-02 16:05:10 -08:00
• c4edcac6d7 refactor(core): remove get_by_model_fast alias in worker_registry (#16313) Chang Su 2026-01-02 16:04:23 -08:00
• e93433892b refactor(steps): consolidate duplicate strip_protocol function (#16318) Chang Su 2026-01-02 16:02:42 -08:00
• a2d4f58a96 fix(http): use 504 Gateway Timeout for upstream timeouts (#16320) Chang Su 2026-01-02 16:01:44 -08:00
• f66b091699 refactor(http): improve pd_types.rs documentation and style (#16319) Chang Su 2026-01-02 16:01:11 -08:00
• 2f623368ab refactor(core): use idiomatic .min() in calculate_delay (#16314) Chang Su 2026-01-02 15:46:13 -08:00
• bd9a2ced47 refactor(core): use thiserror for WorkerError (#16315) Chang Su 2026-01-02 15:33:07 -08:00
• 0ca417d9ae chore(core): remove unused serde_json import in worker.rs (#16312) Chang Su 2026-01-02 15:31:41 -08:00
• 30cfb687fa fixed amd multimodal CI failures caused by refactor in #15812 #15813 (#16287) sunxxuns 2026-01-02 14:55:42 -08:00
• b7c7e03d93 Fix crash dump replay script for image data replay (#16277) Chenxi Li 2026-01-02 13:42:22 -08:00
• f0195627a9 Fix sgl-kernel jobs to skip when target_stage is specified (#16308) Alison Shao 2026-01-02 12:49:26 -08:00
• d401d23876 [model-gateway] Add embedding correctness test comparing against HuggingFace (#16092) Simo LinandChang Su 2026-01-02 10:58:20 -08:00
• 17041f4673 Enable /rerun-stage slash command to work on fork PRs (#16128) Alison Shao 2026-01-02 10:36:12 -08:00
• f07e76b229 Multiple refactors of DeepSeek V32 and context parallel (#16305) Baizhou Zhang 2026-01-03 02:21:22 +08:00
• 1cfd2b2ded [diffusion] chore: remove redundant ulysses nccl warmup (#16301) Xiaoyu Zhang 2026-01-03 00:19:35 +08:00
• 0d244116d2 [DeepSeek v3.2] opt Context Parallelism: support fused moe, multi batch and fp8 kvcache (#13959) Yongfei Xu 2026-01-02 23:49:14 +08:00
• 0eae831797 [diffusion] fix: Increase text length from 256 to 1024 in Qwen-Image (#16248) Changyi YangandMick 2026-01-02 10:25:31 -05:00
• c483a5f45f Tiny adjust hybrid swa handling (#16292) Ke Bao 2026-01-02 23:11:23 +08:00
• f26f6c2c99 [diffusion] fix: make lora compatible with layerwise-offload (#16298) Mickandgemini-code-assist[bot] 2026-01-02 22:33:26 +08:00
• b021332339 [NemotronH] Add latent MoE support (#16227) roikoren755 2026-01-02 16:08:58 +02:00
• 5062537b67 [diffusion] feat: support lightweight e2e warmup for benchmarking (#16213) Mick 2026-01-02 20:10:27 +08:00
• 698629d1f2 Add swa page table to FA3 backend forward metadata (#16267) Ke Bao 2026-01-02 20:00:43 +08:00
• dd93e4451c Support returning customized info (#16144) Qiaolin Yu 2026-01-02 02:32:18 -05:00
• 6c8587b5db [diffusion] fix: align negative prompt with official readme for new model (#16222) Yuhao Yang 2026-01-02 15:20:56 +08:00
• 027042604d [distributed] Clean up MoE groups in destroy_model_parallel (#15345) Yinghai Lu 2026-01-01 22:45:14 -08:00
• bd48ad5e6b [Diffusion] Fix broken ring_attention when use upstream fa3 (#16270) Xiaoyu Zhang 2026-01-02 14:33:12 +08:00
• 749736ba95 Tiny fix single-gpu dumper and add tests for dumper (#16285) fzyzcjy 2026-01-02 13:30:09 +08:00
• 7254986342 [VLM] feat: true on policy for vlm + fsdp (#14636) Nan JiangandXinyuan Tong 2026-01-01 19:54:39 -05:00
• 00562ee14a Support cache eviction for Manual Policy (#16263) fzyzcjy 2026-01-02 08:52:01 +08:00
• d7a8257ba5 docs: remove outdated TODO in GLM-4.7 MoE detector test (#16265) Hudson Xing 2026-01-02 02:25:23 +08:00
• f6f7af4068 [Refactor] Clean up custom op (#15995) DarkSharpness 2026-01-01 21:41:56 +08:00
• a3b1e8ef3d [Feature] add aligned_vector type for JIT kernel (#16162) DarkSharpness 2026-01-01 21:40:05 +08:00
• db499e1889 Tiny add filter, support duplications, add visualizations, fix error and robustness for dump comparator (#16262) fzyzcjy 2026-01-01 17:47:00 +08:00
• 6cf3a6dd69 Support HTTP control for dumper (#16261) fzyzcjy 2026-01-01 17:40:17 +08:00
• 90e24f5c31 Tiny add filter, dump dict, failable save to dumper (#16260) fzyzcjy 2026-01-01 17:35:38 +08:00
• 21de3e1406 [diffusion] webui: tiny fix loading output image (#16251) Mick 2026-01-01 14:34:49 +08:00
• e4c1e441af Fix deprecation warning of sre_parse for python 3.13 (#16247) Lianmin Zheng 2025-12-31 21:25:09 -08:00
• b5af283b1c ci: increase stage-b-test-small-1-gpu partitions from 4 to 5 (#16250) Alison Shao 2025-12-31 20:08:31 -08:00
• 130c6911f8 Add a GH action for cherrypick (#16243) Kangyan-Zhou 2025-12-31 20:05:38 -08:00
• e0e5084802 Fix parse args from file(#13911) (#14085) 3 people 2026-01-01 11:37:33 +08:00
• 70a769bc56 Fix NPU docker release workflow (#16253) Baizhou Zhang 2026-01-01 10:53:45 +08:00
• 3a42c5e341 [VLM] Adopt jit qk_norm kernel in VLM (#16171) Yuan Luoandluoyuan.luo 2026-01-01 10:10:36 +08:00
• 12b89e51d8 Add P90/99 e2e latency in bench_serving script (#16245) Kangyan-Zhou 2025-12-31 15:40:33 -08:00
• 851845579c perf(tree): optimize input_char_count calculation (#16238) Simo Lin 2025-12-31 15:29:37 -08:00
• 57d2ba9203 SGLang Tracing: Supports propagating trace headers through sgl.Engine… (#15814) Feng Su 2026-01-01 07:19:53 +08:00
• 417e75a60f fix: use 2-gpu-runner for pp 2 NemotronH ut (#16239) Yineng Zhang 2025-12-31 12:51:33 -10:00