Commit Graph

1334 Commits

Author SHA1 Message Date
Yineng Zhang
4ab43cfb3e chore: bump v0.4.2 (#3180) 2025-01-27 21:42:05 +08:00
Yineng Zhang
2f79f58873 feat: use sgl-kernel 0.0.3 in sglang (#3179) 2025-01-27 21:39:52 +08:00
Lianmin Zheng
53cef81587 Improve weight loading and code style (#3174) 2025-01-27 03:00:41 -08:00
yigex
351a72d40b add dsv3 mi300 triton config for block scale (#3146) 2025-01-27 17:25:53 +08:00
Lianmin Zheng
52c03f16b9 Add activation parameters to fused_moe (#3170) 2025-01-27 00:23:37 -08:00
YAMY
b045841bae Feature/function calling update (#2700)
Co-authored-by: Mingyuan Ma <mamingyuan2001@berkeley.edu>
Co-authored-by: Chayenne <zhaochen20@outlook.com>
Co-authored-by: shuaills <shishuaiuoe@gmail.com>
2025-01-26 09:57:51 -08:00
Lianmin Zheng
1dda8c5e4c Return more infos for computing average acceptance length (#3152) 2025-01-26 04:51:54 -08:00
Yineng Zhang
7e0976133c udpate sgl-kernel version for srt (#3150) 2025-01-26 20:22:34 +08:00
Lianmin Zheng
d1a0863251 Add a test case for cached_tokens (#3145) 2025-01-26 01:39:28 -08:00
Hubert Lu
f8b28e461a Add CPU affinity setting to latency benchmark (#3085) 2025-01-25 23:52:05 -08:00
Lianmin Zheng
4f118a39d7 Fix repetition penalty (#3139) 2025-01-25 21:48:58 -08:00
yigex
66283dbc0c [Fix] Not skip NVML Check on AMD Platform (#3135) 2025-01-25 21:33:51 -08:00
Hui Liu
8e48ca8cc1 enable kv_scale for Gemma2 (#3113) 2025-01-25 18:29:14 -08:00
Lianmin Zheng
27acf63bbd Use torch.compile for scaling penalty (#3133) 2025-01-25 18:27:33 -08:00
Lianmin Zheng
ea535dc574 Revert "disable custom allreduce on HIP" (#3067) 2025-01-22 21:33:35 -08:00
Ke Wen
862bcff833 Support loading of larger models with on-the-fly quantization (#3061) 2025-01-22 21:33:17 -08:00
Lianmin Zheng
8b84e69f25 Fix tp token sync for dp attention (#3062) 2025-01-22 18:51:40 -08:00
Lianmin Zheng
022614d26e Add some flags to allow sync token ids across TP ranks (#3060) 2025-01-22 15:05:51 -08:00
lukec
b8ab989ff4 Fix the FP8 E4M3 parsing offline scales failure bug (#3045) 2025-01-22 14:19:33 -08:00
Hui Liu
ddc2001fb0 disable custom allreduce on HIP (#3058) 2025-01-22 13:57:22 -08:00
nstream-ai-devx
0d2148efaa fix rotary_embedding rope_scaling for phi (#3055) 2025-01-23 02:15:32 +08:00
Lianmin Zheng
3d8f1c9bcf Use int64 as indices for set_kv_buffer (#3039) 2025-01-21 19:46:09 -08:00
Lianmin Zheng
a4331cd260 Add accuracy and latency tests of eagle into CI (#3027) 2025-01-21 02:55:14 -08:00
Lianmin Zheng
287d07a669 Misc fixes for eagle (flush_cache, CPU overhead) (#3014) 2025-01-20 20:27:38 -08:00
Hui Liu
d2571dd5c7 Enable Cohere2 Models (#3018) 2025-01-20 19:21:41 -08:00
996_icu
b730aa6b9e [EAGLE] Fix some boundary situation when retract reqs and req's max token = 1 (#2939)
Co-authored-by: josephyou <josephyou@tencent.com>
2025-01-20 17:46:43 -08:00
Lianmin Zheng
60b2a44a80 Fix flaky tests in test_programs.py (#3022) 2025-01-20 16:50:39 -08:00
Hongpeng Guo
949b3fbfce [Doc] Update doc of custom logit processor (#3021)
Signed-off-by: Hongpeng Guo <hpguo@anyscale.com>
2025-01-20 16:50:25 -08:00
Hui Liu
da4e8b3892 enable kv_scale remap (#3017) 2025-01-20 14:40:45 -08:00
Enrique Shockwave
af6c5357d5 deepseek v3 and r1 chat template (#3015) 2025-01-20 14:40:12 -08:00
Yineng Zhang
e94fb7cb10 chore: bump v0.4.1.post7 (#3009) 2025-01-20 21:50:55 +08:00
Lianmin Zheng
73401fd016 Sync distributed package from vllm 0.6.4.post1 (#3010) 2025-01-20 04:57:14 -08:00
Lianmin Zheng
89cd923581 Roll back to use vllm custom allreduce (#3006) 2025-01-20 04:03:15 -08:00
Lianmin Zheng
dc1881326f Fix perf regression on small batch sizes (#3008) 2025-01-20 03:39:49 -08:00
Hongpeng Guo
583697cd71 [Enhancement] Custom Logit Processor Improvement (#2998)
Signed-off-by: Hongpeng Guo <hpguo@anyscale.com>
2025-01-20 02:00:35 -08:00
Lianmin Zheng
09bcbe0123 Update TypeBasedDispatcher and balance CI tests (#3001) 2025-01-19 23:37:27 -08:00
Lianmin Zheng
03464890e0 Separate two entry points: Engine and HTTP server (#2996)
Co-authored-by: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
2025-01-19 22:09:24 -08:00
Yineng Zhang
44a9669770 keep rotary_embedding only (#2997) 2025-01-20 13:21:36 +08:00
Chaitanya Sri Krishna Lolla
1a820e38a2 Remove dependency of pynvml on ROCm (#2995) 2025-01-20 13:00:35 +08:00
Lianmin Zheng
cd493b5afc Improve metrics, logging, and importing orders (#2992) 2025-01-19 18:36:59 -08:00
Lianmin Zheng
61f42b5732 Move sgl.Runtime under sglang/lang (#2990) 2025-01-19 17:10:29 -08:00
Hongpeng Guo
e403d23757 [Feature] Add sampler custom logits processor (#2396)
Signed-off-by: Hongpeng Guo <hpguo@anyscale.com>
2025-01-19 14:46:53 -08:00
Enrique Shockwave
3bcf5ecea7 support regex in xgrammar backend (#2983) 2025-01-20 04:34:41 +08:00
Yineng Zhang
2c05f81f15 fix custom op version compatibility (#2988) 2025-01-20 04:21:29 +08:00
Seungduk Kim
d77caa2b75 [#2812] Make the decode status dict capcity adjustable by a CLI param (#2839) 2025-01-19 11:36:53 -08:00
giorgiopiatti-dfinity
8b6a4486ec fix missing revision arg when loading tokenizer (#2982) 2025-01-19 11:36:07 -08:00
Yineng Zhang
6ada05d0ed feat: check for is_cuda for sgl_kernel import (#2984) 2025-01-19 23:33:04 +08:00
yizhang2077
24cafe3177 add config to swtich from vllm custom allreduce to sgl_kernel custom allreduce (#2981) 2025-01-19 22:30:38 +08:00
Yineng Zhang
5a176c92df fix deepseek v2 with cpu device (#2975) 2025-01-19 21:33:27 +08:00
Lianmin Zheng
23196d5254 Simplify logits processor (#2974)
Co-authored-by: SangBin Cho <rkooo567@gmail.com>
2025-01-18 23:03:49 -08:00