Commit Graph

Select branches

Hide Pull Requests

bot/update-est-time-20260928

dsv41-pd

dsv41-pd-visioncp

main

Mono Color

• 4b74c3fcca [chore] Clean up redundant lora_weight_names concept to simplify code (#9131) Lifu Huang 2025-08-17 12:36:58 -07:00
• ce3ca9b02f [router] add cargo clippy in CI and fix-up linting errors (#9242) Jeff Nettleton 2025-08-17 11:03:56 -07:00
• 4d98e48649 Revert "[Misc] feat: Deepgemm update for sgl-kernel (#8790)" to fix kernel CI (#9260) Liangsheng Yin 2025-08-17 22:59:50 +08:00
• 3d77a31885 from python.sglang.srt -> from sglang.srt (#9268) Netanel Haber 2025-08-17 12:45:45 +03:00
• 845d12a979 model: support nvidia/Llama-3_3-Nemotron-Super-49B-v1 (#9067) Netanel HaberandKyle Huang 2025-08-17 11:48:15 +03:00
• e47800e176 Quick Fix GLM (#9264) Stefan He 2025-08-16 23:43:41 -07:00
• bb10e3a1c3 [router] fix pd prefill http request complinace issue (#9237) Simo Lin 2025-08-16 22:36:45 -07:00
• fda762a27d [Bugfix] Change vLLM install order & Add A2 support (#9232) Even Zhou 2025-08-17 13:36:14 +08:00
• 1df84ff414 ci: simplify multi-modality tests by using mixins (#9006) Mick 2025-08-17 13:25:02 +08:00
• 66d6be0874 Bug fix: use correct mm_items in embed_mm_inputs (#8893) Binyao Jiang 2025-08-16 19:55:56 -07:00
• 1c1f8a118e Combine fp4.py and mxfp4.py into one file and support dynamic mxfp4 quantization in mxfp4.py (#9049) kkandwunhuang 2025-08-17 10:01:54 +08:00
• 384f8ab5ce [PD] Support PD disaggregation with Prefill PP (#8846) 5 people 2025-08-17 09:31:31 +08:00
• 6a9d6ca33c fix unexcepted answer in EAGLE mode (#9252) zyksir 2025-08-17 08:45:36 +08:00
• 94371dbbd6 [feature] Ascend NPU graph support (#8027) 6 people 2025-08-17 03:25:17 +03:00
• 740f063035 Fix Custom All Reduce CI job. (#9258) Sai Enduri 2025-08-16 16:29:43 -07:00
• 81da16f6d3 [CI] add deepseek w4a8 test on h20 ci (#7758) Hank Han 2025-08-16 16:54:13 +08:00
• bc938ea13f Fix DP load for embedding (#9165) Brayden Zhong 2025-08-16 02:58:44 -04:00
• eff4eb3fdd Add fp4 quantize before all-gather for Flashinfer cutlass MoE DP (max throughput) (#7667) Trevor Morris 2025-08-15 22:08:11 -07:00
• 87dab54824 Revert "chore: bump sgl-kernel v0.3.6 (#9220)" (#9247) Yineng Zhang 2025-08-15 17:24:36 -07:00
• 5121af4627 Revert "chore(docker): update sgl_kernel version to 0.3.6 in Dockerfi… (#9246) Yineng Zhang 2025-08-15 17:19:38 -07:00
• 983aa4967b Fix nan value generated after custom all reduce (#8663) kkandwunhuang 2025-08-16 03:33:54 +08:00
• 9c3e95d98b [AMD] Expand test coverage for AMD CI and enable apply_token_bitmask_inplace_cuda in sgl-kernel (#8268) Hubert Lu 2025-08-15 12:32:51 -07:00
• e52c3866eb chore(docker): update sgl_kernel version to 0.3.6 in Dockerfile.gb200 (#9243) ishandhanani 2025-08-15 12:06:52 -07:00
• da53e13cbb [router] preserve original worker response header in router (#9236) Simo Lin 2025-08-15 11:01:47 -07:00
• d7e38b2f6d [router] clean up lint warnings with clippy execution (#9201) Jeff Nettleton 2025-08-15 11:01:21 -07:00
• 21b8846066 [router] allow more health check configuration (#9198) Simo Lin 2025-08-15 08:07:45 -07:00
• 0c8594e67d Optional extension for green context (#9231) Liangsheng Yin 2025-08-15 21:33:52 +08:00
• c186feed7f chore: bump sgl-kernel v0.3.6 (#9220) Yineng Zhang 2025-08-15 02:50:50 -07:00
• 84b006b278 Cleanup MoE Refactor (#9223) Cheng Wan 2025-08-15 02:28:33 -07:00
• 8ca07bd948 [CI] Fix sgl-router disaggregation test (#9222) Shangming Cai 2025-08-15 17:24:44 +08:00
• 4fc09e0df0 Fp4 MOE quant kernel optimization (#8777) jy-song-hubandRain Jiang 2025-08-15 01:46:16 -07:00
• a3d99d6dcd [Misc] feat: Deepgemm update for sgl-kernel (#8790) PGFLMG 2025-08-15 16:05:27 +08:00
• 189af90896 [Eagle Warning fix] replace the deprecated 'and' with & (#9215) Xuchun Shang 2025-08-15 15:43:36 +08:00
• f8644a5632 Tiny update tmux history limit on dev container (#9218) fzyzcjy 2025-08-15 15:22:08 +08:00
• e3e75a786a Fix the deprecation warning for enable_flashinfer_mxfp4_moe (#9214) Cheng Wan 2025-08-14 23:59:35 -07:00
• d4db9b028b fix: the store_dtype typo for ascend mla (#9208) shilinlee 2025-08-15 14:58:42 +08:00
• f7dd651dbd feat(hicache-3fs): 3FS-SGLang Hierarchical Cache Deployment Guide​ (#9213) hzh0425 2025-08-15 14:54:31 +08:00
• 9d54c6e6dd feat: remove sm75 (#9207) Yineng Zhang 2025-08-14 22:27:14 -07:00
• 1f9d65f57d use fast math for per_token_group_quant_8bit. (#9177) strgrbandZhang Kaihong 2025-08-15 13:19:56 +08:00
• 295895120d [6/N] MoE Refactor: Cleanup MoE-related configs (#8849) Cheng Wan 2025-08-14 21:14:53 -07:00
• 584e1ab2d0 fix: fix unsupported palette mode of images in bench_serving for mmmu (#9206) Mick 2025-08-15 09:44:46 +08:00
• 392de007cb Minor fix docker container DeepEP on multi platforms (#9205) fzyzcjy 2025-08-15 08:41:49 +08:00
• 004f7f1972 [typo fix] Fix a typo in communicator.py (#9183) Philo 2025-08-15 08:29:38 +08:00
• d2fbf2de0c feat: add fused moe config for Qwen3-235B-A22B-FP8 on B200 (#9204) zixuanzhang226 2025-08-14 17:21:30 -07:00
• fab0f6e77d chore: bump v0.5.0rc2 (#9203) Yineng Zhang 2025-08-14 16:11:16 -07:00
• 27985c27aa feat: update model config (#9202) Yineng Zhang 2025-08-14 15:15:27 -07:00
• ac474869d4 chore: upgrade transformers 4.55.2 (#9197) Yineng Zhang 2025-08-14 13:51:02 -07:00
• 0b1e04f083 [VLM] Improving multimodal tensor hash kernel (#9008) Adarsh Shirawalmath 2025-08-15 02:15:55 +05:30
• c1c7dc4534 feat: Add model version tracking with API endpoints and response metadata (#8795) Chengxing Xie 2025-08-15 03:13:46 +08:00
• 2cc9eeab01 [4/n]decouple quantization implementation from vLLM dependency (#9191) 3 people 2025-08-15 03:05:46 +08:00
• 63d82a776a refine mxfp4 shuffling log (#9194) Xiaoyu Zhang 2025-08-15 01:57:29 +08:00
• 53dcc750b6 [sgl-kernel] Support FlashInfer top_k_top_p_sampling_from_logits (#9060) Yuan Luoandluoyuan.luo 2025-08-15 01:56:36 +08:00
• 432f2053dd [sgl-kernel] 1/N Refactor sglang cutlass 3x - gemm fp8 blockwise sm90 (#8913) Yuan Luoandluoyuan.luo 2025-08-15 01:55:54 +08:00
• 1fea998a45 chore: bump sgl-kernel v0.3.5 (#9185) Yineng Zhang 2025-08-14 03:20:48 -07:00
• 5aa1ebd242 [2/n]decouple quantization implementation from vLLM dependency (#8112) 3 people 2025-08-14 18:19:03 +08:00
• 4dbf43601d fix: zero_init buffer (#9065) eigenandYineng Zhang 2025-08-14 05:39:09 -04:00
• 3d6be1fbce add w8a8-fp8-block-wise H20-3e triton config (#8018) lukec 2025-08-14 14:15:09 +08:00
• 4063234c1a Add H200 fused MoE kernel configs for DeepSeek-V3 in triton 3.3.1 (#7687) Jun Liu 2025-08-14 15:14:09 +09:00
• 83feef5b2c Add H20 fused MoE kernel configs for Dpsk & Qwen3 (#7631) Tommy Yang 2025-08-14 14:13:22 +08:00
• 2871eacc05 Add Triton Fused MoE kernel config for E=16 on B200 (#7004) Brayden Zhong 2025-08-14 02:12:27 -04:00
• ac15bdc194 Add H200 fused MoE kernel tuning configs for Qwen3-Coder-480B-A35B-Instruct (#8852) forestlee95 2025-08-14 14:11:11 +08:00
• d6451c3f65 Add A800 fused MoE kernel tuning configs for GLM4.5 and GLM4.5-Air (#8808) Li Hui 2025-08-14 14:03:17 +08:00
• 841810f227 [Perf] Tunings for SM100 FP8 CUTLASS kernel (#8818) henryg 2025-08-13 21:59:22 -07:00
• 733446dd36 fix io group (#9154) pansichengandZhiqiang Xie 2025-08-14 12:46:42 +08:00
• 4c22897a66 Feature: support qwen and llama4 reducescatter for dp attention padding (#9101) wxzhoucs 2025-08-14 12:10:29 +08:00
• 1bc183c6de Faster weight processing (trtllm-gen moe nvfp4) (#9162) Alex Yang 2025-08-13 21:09:34 -07:00
• b87aacb5c5 [DP Attention] Refactor: adding some utility functions (#9136) Cheng Wan 2025-08-13 21:08:06 -07:00
• b3363cc1aa Fix docker container DeepEP error on Blackwell (#9171) fzyzcjy 2025-08-14 12:06:48 +08:00
• 98457c0453 [Bugfix] Avoid unnecessary reduce-scatter call in prepare_mlp (#9169) Huaixin Chang 2025-08-14 12:04:41 +08:00
• 0fc8bf2cd4 [AMD] Update fallback images for AMD CI (#9159) michael-amdandgemini-code-assist[bot] 2025-08-13 20:15:10 -07:00
• a669bc2f74 Replace sglang.srt.layers.quantization.scalar_types with sgl_kernel.scalar_type (#8951) Hongbo Xu 2025-08-14 10:41:41 +08:00
• 6b7c24712c Fix broken trtllm_mha attn backend with gpt-oss (#9161) Nicolas Castet 2025-08-13 18:40:55 -05:00
• a027a9b4b3 [Generative Score API] Optimization to Remove Decode. (#8840) Sundara Raman Ramachandran 2025-08-13 14:12:24 -07:00
• 9e426466af Clean up allocators (#9134) Lianmin Zhengandgemini-code-assist[bot] 2025-08-13 13:56:04 -07:00
• 2f20f43026 Swap xeon ci to gnr server (#9042) DiweiSun 2025-08-14 03:39:19 +08:00
• 65736dc524 [Model] Support Qwen3ForSequenceClassification for Qwen3-Embed Model (#7957) Zhihao Liu 2025-08-14 02:14:54 +08:00
• 7b56e494be chore: bump v0.5.0rc1 (#9069) Yineng Zhang 2025-08-13 10:44:14 -07:00
• 0ff6d1fce1 Support FA3 backend for gpt-oss (#9028) Ke Bao 2025-08-14 01:41:50 +08:00
• 4a16a71c36 [PD] feat: mooncake use batch reg/dereg (#8910) Teng MaandShangming Cai 2025-08-14 00:54:34 +08:00
• a16923efab [PD] optimize kv cache transfer directly using batch transfer (#9149) FrancisandShangming Cai 2025-08-14 00:54:14 +08:00
• 6337d9057c [router] optimize Rust compilation and development workflow (#9133) Simo Lin 2025-08-13 05:14:25 -07:00
• 71fb8c9527 feat: update fa3 (#9126) Yineng Zhang 2025-08-13 05:07:08 -07:00
• 94f44b88d1 Update fa3 interface and add unit test (#9150) Ke Bao 2025-08-13 20:05:02 +08:00
• 3b3b3baf9f Double vision prefill throughput by defaulting to optimal vision attention backend (#8484) Kevin Xiang LiandXiang Li 2025-08-13 02:08:30 -07:00
• 35e6bc92e3 Update docker file for MI35x base image update to support gpt-oss mxfp4 model (#9111) kkandwunhuang 2025-08-13 15:55:31 +08:00
• 9394ed6386 Fix gpt-oss ~2x memory consumption issue (#9146) fzyzcjy 2025-08-13 15:11:43 +08:00
• 930fe467bd Support Triton FP8 Gemm can handle hidden_dim not divisible by 16 (#9093) Stefan HeandXiaoyu Zhang 2025-08-12 21:21:55 -07:00
• 13c48dcf88 [1/2][resubmit again] sgl-kernel: Fuse routed scaling factor into moe_fused_gate (#9088) Trevor Morris 2025-08-12 20:12:38 -07:00
• 8723b4f146 Use FlashInfer's TRTLLM FP8 Blockscale GEMM (#8588) Elfie Guo 2025-08-12 20:08:40 -07:00
• 62f99e08b3 fix: wrong docker hub org name (#9137) li chaoran 2025-08-13 10:26:19 +08:00
• 86a0be65d8 [Feature] Support custom set kv buffer kernel (#8884) DarkSharpness 2025-08-12 16:56:51 -07:00
• 0edda32001 Support page first layout zero copy for mooncake store (#8651) huangtingweiandZhiqiang Xie 2025-08-13 06:59:26 +08:00
• 924827c3de chore: use cp310 (#9130) Yineng Zhang 2025-08-12 15:33:22 -07:00
• c81daf838d fix: update Dockerfile (#9129) Yineng Zhang 2025-08-12 15:01:29 -07:00
• 25caa7a8a9 [AMD] Support Wave attention backend with AMD GPU optimizations (#8660) 9 people 2025-08-13 04:49:11 +08:00
• 03d114496f Fix typos in supported models documentation (#9119) Hangzhi 2025-08-12 13:35:24 -07:00
• 83123f481e [Quantization] Supported w8a8 int8 quantized Gemma3 and Qwen-VL models (#8619) ichernobandronnie_zheng 2025-08-12 23:31:18 +03:00
• 48afa8f14f [feat] Enable Ascend profiling on SGLang (#8610) ronnie_zhengandliyou_b 2025-08-13 04:28:31 +08:00
• 2ecbd8b8bf [feat] add ascend readme and docker release (#8700) 3 people 2025-08-13 04:25:42 +08:00
• 305b27c124 fix: update Dockerfile (#9125) Yineng Zhang 2025-08-12 13:23:10 -07:00