EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity
2,651 Commits 8 Branches 0 Tags
5cb552b1d461844e4be280060301509fc6ff6cc6
Commit Graph

276 Commits

Author SHA1 Message Date
Mick
5cb552b1d4 refactor: multimodal data (#4754) 2025-03-31 09:57:51 -07:00
Baizhou Zhang
4a63bc32b7 [Fix] Add torch compile for torch.clamp back (#4936) 2025-03-30 20:46:07 -07:00
Baizhou Zhang
e62d60fe6d [Fix] avoid stream sync and torch compile in prefill for fa3 backend (#4932) 2025-03-30 13:53:44 -07:00
Lianmin Zheng
b26bc86b36 Support page size > 1 + eagle (#4908) 2025-03-30 00:46:23 -07:00
Baizhou Zhang
20c90be23d [Feature] Support FA3 backend for MLA (#4831) 2025-03-28 18:30:14 -07:00
tarinkk
7f19e083c1 Support (1 <= dp < tp) in the dp attention in DeepEP (#4770) 
Co-authored-by: Cheng Wan <cwan39@gatech.edu>
 2025-03-27 17:09:35 -07:00
AinL
17000d2b3a Remove Unintended Capture Batch Sizes in AMD HIP Graph Runner (#4638) 2025-03-27 08:41:33 -07:00
fzyzcjy
92bb49a7f9 Patch PyTorch's bug that cross-process tensor transfer will lead to wrong device (#4565) 2025-03-27 00:22:33 -07:00
fzyzcjy
26f07294f1 Warn users when release_memory_occupation is called without memory saver enabled (#4566) 2025-03-26 00:18:14 -07:00
Mick
1e86457c90 model: Minicpmo (#3023) 2025-03-24 20:08:40 -07:00
Stefan He
5d7edc8e55 Support FA3 as Attention backend by using --attention-backend fa3 (#4680) 
Co-authored-by: qsong <qsong@linkedin.com>
Co-authored-by: qingquansong <ustcsqq@gmail.com>
 2025-03-23 23:28:11 -07:00
Mick
11577cedb7 refactor: bug fixes and refactor for vlm (#4661) 2025-03-22 22:48:49 -07:00
JieXin Liang
9e93ef3f8e [fix] fix illegal mem access and clean up triton attention backend (#4571) 2025-03-20 02:01:52 -07:00
Jinyan Chen
f44db16c8e [Feature] Integrate DeepEP into SGLang (#4232) 
Co-authored-by: Cheng Wan <cwan39@gatech.edu>
Co-authored-by: Xuting Zhou <xutingz@nvidia.com>
 2025-03-19 08:16:31 -07:00
JieXin Liang
c0e9a36c5f Optimize Triton decoding kernel for dynamic workload (#4553) 2025-03-18 21:25:38 -07:00
aoshen524
588865f0e0 [Feature] Support Tensor Parallelism and Weight Slicing for Lora (#4274) 
Co-authored-by: ShenAo1111 <1377693092@qq.com>
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
 2025-03-18 20:33:07 -07:00
James Liu
9e0186f352 [Feature] Support EAGLE 3 (#4247) 2025-03-18 07:35:23 -07:00
Mick
d373a48c98 fix: second_per_grid_ts should be used to get mrope position (#3682) 2025-03-17 18:12:38 -07:00
Lianmin Zheng
5493c3343e Fix data parallel + tensor parallel (#4499) 2025-03-17 05:13:16 -07:00
萝卜菜
d6d21640d3 [Feature] Support Deepseek-VL2 (#2798) 
Co-authored-by: Edenzzzz <wtan45@wisc.edu>
Co-authored-by: Chayenne <zhaochen20@outlook.com>
Co-authored-by: Yi Zhang <1109276519@qq.com>
 2025-03-16 23:07:59 -07:00
Mick
9d02bb3e2a Urgent model support: support gemma-3-it (#4424) 2025-03-16 17:37:32 -07:00
lukec
a53fe428f9 Support FlashMLA backend (#4472) 
Co-authored-by: yinfan98 <1106310035@qq.com>
 2025-03-16 09:07:06 -07:00
Ying Sheng
1b859295f4 [Eagle] Remove the greedy branch and some redundant code (#4363) 
Co-authored-by: Sehoon Kim <sehoon@x.ai>
 2025-03-16 02:48:55 -07:00
JieXin Liang
1a3fa75f2f [Fix] use torch.cat instead of torch.concat to prevent entering the Autograd backends. (#4466) 2025-03-16 00:02:47 -07:00
wangyu
1ce4878d31 feat(remote_model): support variable remote backend for model loader (#3964) 
Signed-off-by: wangyu <wangyu.steph@bytedance.com>
 2025-03-14 00:40:44 -07:00
Lianmin Zheng
8e66fbecee Improve DP attention (#4390) 
Co-authored-by: dhou-xai <dhou@x.ai>
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
 2025-03-13 08:23:56 -07:00
Lianmin Zheng
45de89719c Revert "[XPU][CPU] Enable the native path of DeepSeek" (#4367) 2025-03-12 23:45:52 -07:00
Meng, Hengyu
71046fcd71 [XPU][CPU] Enable the native path of DeepSeek (#4086) 
Co-authored-by: Zhang, Liangang <liangang.zhang@intel.com>
 2025-03-12 22:26:29 -07:00
Lianmin Zheng
c76040e31b Support page size > 1 (#4356) 2025-03-12 22:22:39 -07:00
Lianmin Zheng
e35a93fa8a Move output processing logic from scheduler.py into a separate file (#4354) 2025-03-12 16:21:49 -07:00
Lianmin Zheng
d40ee62b5d Update nightly tests (#4352) 2025-03-12 15:36:13 -07:00
Yineng Zhang
d1da58e275 unify is_cuda and is_hip (#4321) 2025-03-11 18:12:56 -07:00
Lianmin Zheng
00d25a7f5e Fix quantization and nightly tests (#4258) 2025-03-10 03:06:21 -07:00
Lianmin Zheng
08c4d764a5 lazy import attn backends (#4200) 2025-03-08 00:41:35 -08:00
Lianmin Zheng
d4017a6b63 [EAGLE] many fixes for eagle (#4195) 
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
Co-authored-by: Sehoon Kim <sehoon@x.ai>
 2025-03-07 22:12:13 -08:00
Yineng Zhang
eb61f5c9af Revert "ROCm: Flex Attention Enablement with custom backends (#4178)" (#4186) 2025-03-07 10:27:52 -08:00
HAI
0beea4503f ROCm: Flex Attention Enablement with custom backends (#4178) 
Co-authored-by: linsun12 <linsun12@amd.com>
 2025-03-07 04:38:53 -08:00
Lianmin Zheng
bc1534ff32 Fix a draft model accuracy bug in eagle; support step=1; return logprob in eagle (#4134) 
Co-authored-by: Sehoon Kim <kssteven418@gmail.com>
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
Co-authored-by: Sehoon Kim <sehoon@x.ai>
 2025-03-06 06:13:59 -08:00
Lianmin Zheng
98c73d71cb [Minor] make the __init__ function of model_runner.py shorter (#4132) 2025-03-06 01:51:12 -08:00
Zhiqiang Xie
aee30630d8 Add a pointer to the real KV cache pool (#4113) 2025-03-05 21:39:07 -08:00
Ke Bao
ef9d3b3c2c Fix triton kernel illegal memory issue for eagle (#4100) 2025-03-05 11:23:53 -08:00
Baizhou Zhang
fc91d08a8f [Revision] Add fast decode plan for flashinfer mla (#4012) 2025-03-05 11:20:41 -08:00
Ying Sheng
d3d4d76758 [Eagle] Refactor eagle speculative decoding (#3986) 
Co-authored-by: Ke Bao <ISPObaoke@163.com>
 2025-03-05 08:06:07 -08:00
Lianmin Zheng
e074d84e5b [Minor] more code cleanup (#4077) 2025-03-04 21:23:47 -08:00
Chen Shengzhi
61261b3996 [XCCL] Use xccl for xpu backend since xccl is ready in latest PyTorch. (#3954) 2025-03-04 04:05:56 -08:00
Ke Bao
9fafa62db7 Share target model embed and head weights for nextn (#4033) 2025-03-03 13:30:04 -08:00
Lianmin Zheng
935cda944b Misc clean up; Remove the support of jump forward (#4032) 2025-03-03 07:02:14 -08:00
Lianmin Zheng
66301e124f Improve code styles (#4021) 2025-03-03 03:20:23 -08:00
Lianmin Zheng
ac2387279e Support penalty in overlap mode; return logprob with chunked prefill; improve benchmark scripts (#3988) 
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
Co-authored-by: dhou-xai <dhou@x.ai>
Co-authored-by: Hanming Lu <hanming_lu@berkeley.edu>
 2025-03-03 00:12:04 -08:00
Lianmin Zheng
9e1014cf99 Revert "Add fast decode plan for flashinfer mla" (#4008) 2025-03-02 19:29:10 -08:00