Commit Graph

2810 Commits

Author SHA1 Message Date
Lianmin Zheng
55381a46ac Revert "[Feature] Simple Improve Health Check Mechanism for Production-Grade Stability" (#8181) 2025-07-19 22:41:30 -07:00
Atream
a589a07167 fix moe gate dtype, fix tbo, fix fake dispatch (#7825) 2025-07-19 22:13:46 -07:00
Yineng Zhang
f62d75b6a1 feat: add b200 tp 16 kimi k2 moe config (#8178) 2025-07-19 20:04:12 -07:00
Yineng Zhang
0f9b11e310 feat: add h200 tp 16 kimi k2 moe config (#8176) 2025-07-19 20:04:02 -07:00
Pavel Logachev
877e35d775 Add get_hidden_dim to qwen3.py for correct lora (#7312) 2025-07-19 19:31:16 -07:00
Clay
cbdfb77123 Enable FlashInfer support encoder models and add head_dim padding workaround (#6230) 2025-07-19 19:30:16 -07:00
ybyang
4540a4666a [Feature] Simple Improve Health Check Mechanism for Production-Grade Stability (#8115)
Signed-off-by: ybyang <ybyang7@iflytek.com>
2025-07-19 18:10:00 -07:00
Baizhou Zhang
8cddfa56a1 Clean warning logs for gate_proj loading in Lora (#8172) 2025-07-19 15:56:50 -07:00
Lifu Huang
4e3defe5a7 Support start up LoRA server without initial adapters (#8019) 2025-07-19 15:38:09 -07:00
Garry Fang
60468da4e2 bugfix: fix sglang crash in NVIDIA MIG container (#8167)
Signed-off-by: Garrybest <garrybest@foxmail.com>
2025-07-19 14:41:27 -07:00
Lifu Huang
3de617a75b Fix LoRA buffer contamination during adapter eviction (#8103) 2025-07-19 13:14:08 -07:00
Lianmin Zheng
bb0e8a32b5 Clean up server args (#8161) 2025-07-19 11:32:52 -07:00
Yineng Zhang
561dd7b2ce chore: upgrade sgl-kernel 0.2.6 (#8166) 2025-07-19 03:17:08 -07:00
Cheng Wan
15ad6c9086 [1/N] MoE Refactor: refactor select_experts (#7966) 2025-07-19 00:51:15 -07:00
Yingchun Lai
610381b75e [health_generate] fix: fix the /health_generate always success bug (#8028) 2025-07-18 22:08:46 -07:00
Shangming Cai
1403ea5694 [PD] Support non-MLA models PD different TP with DP attention (#7931)
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
2025-07-18 22:00:49 -07:00
Binyao Jiang
b7e951a6db Feat: Support audio in Phi4-mm model (#8048) 2025-07-18 21:03:53 -07:00
Haohui Mai
d918ab7985 Support NVFP4 quantized dense models on AMD CDNA2/CDNA3 GPUs (#7302)
Co-authored-by: HAI <hixiao@gmail.com>
Co-authored-by: Sai Enduri <saimanas.enduri@amd.com>
2025-07-18 19:59:39 -07:00
Mick
3964b352c3 chore: tune mem fraction static for vlm (#6881) 2025-07-18 17:19:27 -07:00
Lianmin Zheng
9c7a46180c [Doc] Steps to add a new attention backend (#8155) 2025-07-18 16:38:26 -07:00
Hubert Lu
7750b91ca8 [AMD] Add triton awq_dequantize kernel to support AWQ on ROCm (#7661) 2025-07-18 14:27:25 -07:00
Hongbo Xu
1f76fc8747 [3/n] chore: decouple AWQ implementation from vLLM dependency (#8113)
Co-authored-by: AniZpZ <zhuangsen.zp@antgroup.com>
2025-07-18 11:45:22 -07:00
Even Zhou
6737671c82 [Bugfix] Fix w8a8_int8 import error on NPU (#8147) 2025-07-18 11:34:55 -07:00
Enrique Shockwave
fd63b62eaa fix compressed tensors WNA16 imports (#8142) 2025-07-18 11:34:14 -07:00
Sai Enduri
d0510f08fe Revert "Fix different device type adjustment in PP" (#8141) 2025-07-18 01:12:11 -07:00
Zhiqiang Xie
9d33fcfb8e Hicache Storage Layer Prototype (#7704) 2025-07-18 15:20:19 +08:00
jianan-gu
7891bac16b [Quantization][w8a8_int8] Fix weight loading issue for w8a8_int8 path with "ignore" layer list in quantization config (#7820) 2025-07-17 22:03:56 -07:00
jianan-gu
48c1fa7bb6 [CPU][Llama4] Fix Llama4 MoE inputs with "apply_router_weight_on_input" (#7889) 2025-07-17 21:43:25 -07:00
yilian49
8aa5ae6b04 load draft model fix (#7506) 2025-07-17 21:41:32 -07:00
Minglei Zhu
8a32355704 Feat: Support Granite 3.0 MoE in SGLang (#7959) 2025-07-17 20:56:03 -07:00
Mick
e1020dc588 refactor: simply MultimodalTokens logic (#7924) 2025-07-17 17:59:15 -07:00
Zhao Chen
3586b4cef2 feat: add production metric for retracted requests due to insufficient kvcache (#7030)
Signed-off-by: Zhao Chen <zhaochen.zju@gmail.com>
2025-07-17 11:59:05 -07:00
Asher
4296021499 [Hunyuan]: Fix Dense Model Support (#8117)
Signed-off-by: Asher Zhang <asherszhang@tencent.com>
2025-07-17 10:00:11 -07:00
Ziqi Fan
01857fab61 fix: update HostKVCache init to report correct msg when available memory is not enough (#8102) 2025-07-17 21:24:34 +08:00
fzyzcjy
519ff5c8e6 Super tiny fix typo (#8046) 2025-07-17 21:15:51 +08:00
Cheng Wan
49b8777460 Refactor: move all quantization-related code to srt/layer/quantization (#7989) 2025-07-17 00:47:07 -07:00
hzh0425
5c08a36cbf [Fix] ensure DeepGEMM is only enabled for FP8_W8A8 models (#8110) 2025-07-16 21:33:29 -07:00
Cheng Wan
9069884b51 [ci] disable memory imbalance check for draft worker (#8108) 2025-07-16 20:41:47 -07:00
Yingchun Lai
795668dc73 feat: add tp_rank, pp_rank and dp_rank labels for scheduler metrics (#7597)
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
2025-07-16 17:55:59 -07:00
Mick
4395c87a9b refactor: unify names of the feature field of MultimodalDataItem (#8075) 2025-07-16 17:52:38 -07:00
Peng Zhang
c28ad1990d [1/n] chore: decouple quantization implementation from vLLM dependency (#7992) 2025-07-16 15:56:26 -07:00
Xiaoze Fan
570d33437b [Feature] Layer-wise Prefill (#7634)
Signed-off-by: jason-fxz <jason341132@qq.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2025-07-17 01:57:46 +08:00
YanbingJiang
b188a89a5d Fix CI xeon test with triton 3.3.1 (#8086) 2025-07-16 02:12:23 -07:00
Mick
497efe747d Revert "feat: replace Decord with video_reader-rs" (#8077) 2025-07-15 20:04:56 -07:00
Qiaolin Yu
69f453e5a4 Use device_group for all_gather when disabling overlap scheduling (#8001) 2025-07-15 19:38:58 -07:00
Qiaolin Yu
3bc43c683e Fix different device type adjustment in PP (#7760) 2025-07-15 19:37:14 -07:00
Xinyuan Tong
7498522f7d update transformers to 4.53.2 (#8029)
Signed-off-by: Xinyuan Tong <justinning0323@outlook.com>
2025-07-15 18:24:39 -07:00
strgrb
194841e329 remove kv_a.congigous in DeepseekV2AttentionMLA (#8058)
Co-authored-by: Zhang Kaihong <zhangkaihong.zkh@alibaba-inc.com>
2025-07-15 18:20:41 -07:00
kozo
ebff5fcb06 feat: replace Decord with video_reader-rs (#5163)
Signed-off-by: Xinyuan Tong <justinning0323@outlook.com>
Co-authored-by: Xinyuan Tong <justinning0323@outlook.com>
2025-07-15 18:17:34 -07:00
yhang
14f1f1514b H20 tune config for Kimi (#8047) 2025-07-15 13:48:31 -07:00