EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity
4,274 Commits 8 Branches 0 Tags
4d16c88b6e96fea9f9fa684454f1d2c9a01b78bd
Commit Graph

2905 Commits

Author SHA1 Message Date
Elfie Guo
4d16c88b6e Update cutlass_moe.py (#8535) 2025-07-29 10:49:41 -07:00
Yineng Zhang
6478831be9 chore: bump v0.4.9.post6 (#8517) 2025-07-29 02:30:07 -07:00
Lifu Huang
fb16fbaf52 Fix incorrect KV cache allocation for MTP models. (#8482) 
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
 2025-07-28 22:54:50 -07:00
fzyzcjy
0ce84c822b Support colocating requests (#7973) 2025-07-28 22:51:49 -07:00
fzyzcjy
59d0bf012f Tiny add warnings for DeepEP when it is suboptimal (#8426) 2025-07-28 22:51:38 -07:00
fzyzcjy
7df2c0c2db Reduce memory usage for fp4 moe (#8413) 2025-07-28 22:51:23 -07:00
Yineng Zhang
8240a6b013 chore: add glm 4.5 fp8 tp4 config (#8480) 2025-07-28 16:14:01 -07:00
Yineng Zhang
3a04aa4be7 chore: add glm4 fp8 tp8 config (#8478) 2025-07-28 16:08:53 -07:00
Stefan He
74e7e45710 Fix DEEPEP BF16 compatibility for Deepseek Style model like GLM 4.5 (#8469) 
Co-authored-by: Minglei Zhu <mingleizhu1122@gmail.com>
 2025-07-28 14:36:08 -07:00
Cheng Wan
9c138a0445 [3/N] MoE Refactor: Simplify DeepEP Output (#8421) 2025-07-28 11:37:17 -07:00
Timofey
c8f549d96d Fix parsing ChatCompletionMessage (#7273) 
Co-authored-by: Timofey K <timosha1113@gmail.com>
 2025-07-28 11:35:14 -07:00
Kaixi Hou
134fa43e19 [NVIDIA] Change to use num_local_experts (#8453) 2025-07-28 10:38:19 -07:00
Yineng Zhang
ccfe52a057 fix: update dep (#8467) 2025-07-28 10:19:33 -07:00
harrisonlimh
747dd45077 feat: throttle requests at scheduler based on --max_queued_requests (#7565) 2025-07-28 22:32:33 +08:00
erictanjn
a9dd3ec3e9 fix:reorder topk experts to ensure shared expert replaces minimal score (#8125) 2025-07-28 20:36:46 +08:00
Yineng Zhang
45bc170b36 chore: bump v0.4.9.post5 (#8458) 2025-07-28 02:11:06 -07:00
Minglei Zhu
25f73c6cf3 fix GLM4_MOE launch with compressed_tensor quant model (#8456) 2025-07-28 01:31:20 -07:00
Binyao Jiang
581e7dcb92 GLM-4.5 Model Support Follow-up (#8445) 2025-07-27 23:35:20 -07:00
Yuxuan Zhang
6d6a8bc278 GLM-4.5 Model Support (#8224) 
Co-authored-by: Lifu Huang <lifu.hlf@gmail.com>
Co-authored-by: Binyao Jiang <byjiang1996@gmail.com>
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
 2025-07-27 22:54:07 -07:00
Shangming Cai
2fd5c7049f [PD] Fix abort_request for PD disaggregation (#8352) 
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
Co-authored-by: ybyang <10629930+whybeyoung@users.noreply.github.com>
 2025-07-27 21:48:27 -07:00
Stefan He
4ad9737045 chore: bump transformer to 4.54.0 (#8416) 
Co-authored-by: Binyao Jiang <byjiang1996@gmail.com>
Co-authored-by: Lifu Huang <lifu.hlf@gmail.com>
 2025-07-27 21:27:25 -07:00
Qiaolin Yu
2810338401 [feat] Support different attention backends for prefill and decode (#6338) 
Co-authored-by: tianqilin.99 <tianqilin.99@bytedance.com>
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
 2025-07-28 11:42:29 +08:00
Chang Su
dd487e5553 bugfix: Fix XGrammar backend to use model's EOS tokens for constrained generation (#8422) 2025-07-28 10:01:02 +08:00
Chang Su
b47eda3316 bugfix: Fix multiple finish_reason chunks and tool_calls finish reason check (#8417) 2025-07-27 13:31:06 -07:00
Binyao Jiang
e983d66680 Fix: Improve test_openai_function_calling unit test and fix reasoning_parser.py think_start_token logic (#8316) 
Co-authored-by: Chang Su <chang.s.su@oracle.com>
 2025-07-27 13:12:59 -07:00
fzyzcjy
b58c3c285e Support ue8m0 for triton quant kernel (#7603) 2025-07-27 13:04:35 -07:00
Lifu Huang
df90645525 Support overlapped lora updates (#8213) 2025-07-27 13:00:44 -07:00
Shangming Cai
22e00eeb4a [Bugfix] Prevent PD server crash from invalid grammar (#8062) 
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
 2025-07-28 00:17:51 +08:00
Yuan Luo
b3eac168e7 Support triton kernels v3.4.0 for fused_moe (#8258) 
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
Co-authored-by: Cheng Wan <cwan@x.ai>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
 2025-07-27 02:28:49 -07:00
Yineng Zhang
10ee89559e chore: upgrade flashinfer v0.2.9rc2 (#8406) 2025-07-27 01:41:22 -07:00
Cheng Wan
4d921f2b79 [hotfix] fix merge conflicts in FlashInferEPMoE (#8405) 2025-07-27 01:24:10 -07:00
Kevin Xiang Li
44d600cd67 Support precomputed_embeddings for Llama 4 (#8156) 
Signed-off-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
Co-authored-by: Xiang (Kevin) Li <lik@nvidia.com>
Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
 2025-07-27 01:14:49 -07:00
Elfie Guo
5c9c275bc8 Use FlashInfer FP4 gemm. (#8241) 2025-07-27 01:05:22 -07:00
Cheng Wan
bf0f448fe5 [2/N] MoE Refactor: Unify weight loader and quant methods (#8397) 2025-07-27 01:00:21 -07:00
Yingchun Lai
36d6f0ba5b fix: fix the missing metrics on non-rank0 nodes (#7720) 2025-07-27 00:55:25 -07:00
Li Hui
2a1936de96 Add A800 fused MoE kernel tuning configs for Qwen3-Coder-480B-A35B-Instruct (#8351) 2025-07-27 00:46:07 -07:00
Mick
0bcc195f4e fix: minor fix TransportProxyTensor under tp (#8382) 2025-07-27 00:38:49 -07:00
Kaixi Hou
85486b6f6f [NVIDIA] Add Flashinfer MoE blockscale fp8 backend (#8036) 2025-07-27 00:34:41 -07:00
fzyzcjy
62222bd27e Minor tool for comparison of benchmark results (#7974) 2025-07-27 00:27:50 -07:00
fzyzcjy
ed0fdbf35b Tool to dump and compare internal activation tensors (#7976) 2025-07-27 00:27:21 -07:00
Xinyuan Tong
b602f42354 Urgent Fix: intern-s1 chat-template matching (#8403) 
Signed-off-by: Xinyuan Tong <justinning0323@outlook.com>
 2025-07-27 00:22:31 -07:00
Zhiqiang Xie
528bd1ed85 HiCache, check before terminate prefetching (#8372) 2025-07-26 23:13:16 -07:00
Lifu Huang
761546315c Remove slot usage in code to be backward-compatible with python 3.9 (#8396) 2025-07-26 21:24:22 -07:00
Lifu Huang
5c705b1dce Add perf tests for LoRA (#8314) 2025-07-26 14:55:22 -07:00
RunningLeon
b7094a5ef1 model: support intern-s1 (#8350) 
Signed-off-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
Co-authored-by: zxy <zhou0493@e.ntu.edu.sg>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
Co-authored-by: Mick <mickjagger19@icloud.com>
Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
 2025-07-26 13:48:51 -07:00
fzyzcjy
da0c026084 Tiny assert EPLB is used together with expert parallel (#8381) 2025-07-26 03:20:39 -07:00
Mick
3212c2ad3f vlm: optimize tensor transport (#6003) 
Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
 2025-07-26 17:41:01 +08:00
Mick
534756749a chore: improvements on mm_utils (#7737) 2025-07-26 17:38:56 +08:00
Stefan He
ce32bc2ba9 Extract update_weights from RL Engine to SGLang to keep simplicity and fix torch reduce (#8267) 
Co-authored-by: CuiBo 82354186+SuperCB@users.noreply.github.com
Co-authored-by: GeLee 865038696@qq.com
Co-authored-by: 杨睿 yangruipis@163.com
 2025-07-26 02:00:59 -07:00
Cheng Wan
e236d8fee8 Save peak memory in logits processor (#8343) 2025-07-26 01:46:42 -07:00