Commit Graph

3839 Commits

Author SHA1 Message Date
Stefan He
00fbd8a484 Fix typo of flash_cache (#7513) 2025-06-25 02:04:41 -07:00
valarLip
802815e40b take aiter get_rope back (#7521) 2025-06-25 02:03:33 -07:00
valarLip
4c6675c4fc enable aiter fp8 blockscale quant (#7520) 2025-06-25 02:02:31 -07:00
Hongbo Xu
e21aa1df67 [PD] Add different TP sizes support for no-MLA models (#6793)
Co-authored-by: shangmingc <csmthu@gmail.com>
Co-authored-by: Shangming Cai <caishangming@linux.alibaba.com>
2025-06-25 02:00:22 -07:00
zixuanzhang226
f3cbd24541 feat: send kvmetrics from sglang scheduler (#6721) 2025-06-25 01:57:49 -07:00
ll819214
506a2d5934 npu fused op (#7386)
Co-authored-by: Li Junwen <lijunwen13@hisilicon.com>
2025-06-25 01:54:20 -07:00
Shangming Cai
a07f8ae4b7 [CI] Upgrade mooncake to v0.3.4.post2 to fix potential slice failed bug (#7522)
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
2025-06-25 01:49:22 -07:00
Chunyuan WU
7eb47b0f3d [CPU] [BF16] Call fused_experts_cpu, weight_packed_linear and bmm_cpu kernel in DeepSeek model (#6641)
Co-authored-by: Thien Tran <gau.nernst@yahoo.com.sg>
2025-06-25 01:43:33 -07:00
DangKai
bc2e5645c4 fix: force synchronization between TP workers when update_weights (#6626)
Co-authored-by: dangkai.dk <dangkai.dk@alibaba-inc.com>
2025-06-25 01:35:59 -07:00
Simo Lin
3abc30364d [ci] add router benchmark script and CI (#7498) 2025-06-25 01:28:25 -07:00
linzhuo
afeed46530 clean duplicate code (#7512) 2025-06-25 01:22:20 -07:00
yilian49
587b4c6e92 EPLB support for MTP (#7510) 2025-06-25 01:16:56 -07:00
Shangming Cai
7b9a174a7a [PD][Spec] Fix hidden state transfer for spec decode (#7516)
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
2025-06-25 00:42:07 -07:00
ybyang
03c039c48e [OAI] patch origin request_id logic (#7508) 2025-06-24 20:09:38 -07:00
Ke Bao
57ab776910 Fuse sorted_token_ids padding to moe_align_block_size kernel (#7437) 2025-06-24 17:44:27 -07:00
Chang Su
112b496a6c misc: Improvement to serving_chat.py and add more ut (#7489) 2025-06-24 17:19:51 -07:00
Chang Su
3562256bb2 fix: Add --model as an alias for --model-path in server_args (#7505) 2025-06-24 12:08:08 -07:00
Trevor Morris
5f527834a8 [PD] NIXL: Register kv args in advance and cleanup finished requests (#6717) 2025-06-24 11:26:09 -07:00
xianzhiT
9f1787fa60 Support multi-thread model weight loading (#7277) 2025-06-24 10:39:10 -07:00
Xiaoyu Zhang
8ecad0b16f [benchmark] fbgemm benchmark support bandwidth report and support fbgemm_cutlass_gmm (#7422) 2025-06-24 09:44:55 -07:00
Ke Bao
7151194bb2 Remove cumsum_buffer initilization (#7439) 2025-06-24 02:22:16 -07:00
Francis
2ed68d7a6c [PD Disaggregation] replace transfer with batch transfer for better performance (#7236) 2025-06-24 02:12:04 -07:00
valarLip
e984d5073b enable aiter_biased_grouped_topk kernel (#7423) 2025-06-24 02:09:42 -07:00
Alex Sun
755f314785 [AMD] add aiter fused moe in DeepEP path (#7268) 2025-06-24 09:05:47 +00:00
Yineng Zhang
7c3a12c000 chore: bump v0.4.8 (#7493) 2025-06-23 23:14:22 -07:00
Yineng Zhang
e846d95ef6 chore: bump sgl-kernel v0.2.0 (#7490) 2025-06-23 22:29:50 -07:00
Charles Chen
15f3401343 Fix MTP with Deepseek R1 Fp4 (#7376) 2025-06-23 21:38:07 -07:00
Chang Su
d04163b3fa Fix RequestValidationError response format (#7487) 2025-06-23 20:35:11 -07:00
huangtingwei
7732bbe458 bugfix: Prevent global mutation of conv.stop_str across requests (#7347)
Co-authored-by: Chang Su <chang.s.su@oracle.com>
2025-06-23 19:36:23 -07:00
u4lr451
ed0a0b692c Perormance: Enable cuda graph for dp idle batch (#7269)
Co-authored-by: austindeng <austindeng@tencent.com>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
Co-authored-by: ch-wan <cwan39@gatech.edu>
2025-06-23 17:34:13 -07:00
Chang Su
fa42e41962 ci: Revert openai_server related tests in AMD suites (#7449) 2025-06-23 15:28:22 -07:00
Yuhong Guo
e5afb88b1c Support weight loading without mmap (#7469) 2025-06-23 15:13:59 -07:00
Charles Chen
e5ddeb04d5 Quick fix for DeepGemm requant to also cover MTP. (#7378) 2025-06-23 12:08:54 -07:00
JieXin Liang
bdbb8d009a [perf] slightly imporve DeepSeek-R1-FP4 TP8 (#7481) 2025-06-23 12:05:30 -07:00
Zhiqiang Xie
34c3f9b2d3 kvcache io kernels and test case (#7382) 2025-06-23 11:58:59 -07:00
Liangsheng Yin
76139bfba0 update mooncake in dockerfile (#7480) 2025-06-24 02:29:30 +08:00
Cheng Wan
f8d48fd311 Fix dtype for idle input in spec decoding (#7456) 2025-06-23 11:23:25 -07:00
Chang Su
34b6b8426f feat(func_call): Add more check in BaseFormatDetector.parse_streaming_increment (#7479) 2025-06-23 11:15:47 -07:00
Liangsheng Yin
25549433e8 Fix prefill OOM due to wrong token calculation when page > 1 (#7397) 2025-06-24 02:12:29 +08:00
Shangming Cai
d6dddc19ff [CI] Upgrade mooncake to 0.3.4.post1 to fix 8 gpu tests (#7472)
Signed-off-by: Shangming Cai <caishangming@linux.alibaba.com>
2025-06-24 02:10:50 +08:00
Lianmin Zheng
55e03b10c4 Fix a bug in BatchTokenIDOut & Misc style and dependency updates (#7457) 2025-06-23 06:20:39 -07:00
kk
8aa68ed5c4 Solve docker build failed in the virtual machine (#7290)
Co-authored-by: wunhuang <wunhuang@amd.com>
Co-authored-by: Sai Enduri <saimanas.enduri@amd.com>
Co-authored-by: HAI <hixiao@gmail.com>
2025-06-23 09:10:30 +00:00
xutizhou
506c4928f5 feat: integrate deepgemm into EPMoE (#6821)
Co-authored-by: tianqilin.99 <tianqilin.99@bytedance.com>
Co-authored-by: TianQiLin666666 <1834987979@qq.com>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
2025-06-23 01:38:58 -07:00
Lianmin Zheng
30ceccc74a Update hyperparameter_tuning.md (#7454) 2025-06-22 22:42:55 -07:00
Cheng Wan
ac5010e0ba Fix CUDA Graph Check under Deepep with DP FFN (#7451) 2025-06-22 20:35:58 -07:00
Yi Zhang
3cee035e99 add fused moe config for qwen3 in triton3.3.1 (#7445) 2025-06-22 18:29:25 -07:00
Simo Lin
30f2a44a96 [misc] Add PD service discovery support in router (#7361) 2025-06-22 17:54:14 -07:00
kk
bd4f581896 Fix torch compile run (#7391)
Co-authored-by: wunhuang <wunhuang@amd.com>
Co-authored-by: Sai Enduri <saimanas.enduri@amd.com>
2025-06-22 15:33:09 -07:00
Ke Bao
50f1b6d6b1 Remove copy after bmm (#7441) 2025-06-22 15:07:44 -07:00
Trevor Morris
5962e70d8d FlashInfer NVFP4 MoE with EP & 2-stream shared expert (#7327)
Co-authored-by: JieXin Liang <Alcanderian@users.noreply.github.com>
Co-authored-by: alcanderian <alcanderian@gmail.com>
2025-06-22 13:38:47 -07:00