Commits · 66324895c6925c86c2b8c811ebb6dfb93ae42356 · change / sglang · GitLab

20 May, 2025 15 commits
- [docs] Fix torch version (#6472) · 66324895
  Wenxuan Tan authored May 20, 2025
  
  66324895
- Fix master CI for DeepSeek (#6447) · 13feffd0
  fzyzcjy authored May 20, 2025
  
  13feffd0
- Support dispatching logical to physical experts (#6385) · e98afbe0
  fzyzcjy authored May 20, 2025
  
  e98afbe0
- [doc] add note for get_num_kv_splits in triton_backend (#6444) · 69af3ec3
  JieXin Liang authored May 20, 2025
  
  69af3ec3
- Update extend/decode attention kernel for CPU in sgl-kernel and add UTs (#6405) · 32cc66ef
  YanbingJiang authored May 20, 2025
```
Co-authored-by: mingfeima <mingfei.ma@intel.com>
```
  32cc66ef
- [QuickFix] fix gptq model initialize (#6429) · 83f2d9d4
  PGFLMG authored May 20, 2025
  
  83f2d9d4
- Address performance regression: disable multiple streams on ROCm (#6412) · 6317c5c6
  HAI authored May 19, 2025
  
  6317c5c6
- Support DeepSeek EPLB algorithm with static distributions (#6387) · cba1cdbc
  fzyzcjy authored May 20, 2025
  
  cba1cdbc
- Support loading weights when physical experts are different from logical experts (#6386) · c471d39e
  fzyzcjy authored May 20, 2025
  
  c471d39e
- Refactor DeepSeek logic into atomic operations (#6326) · d0443275
  fzyzcjy authored May 20, 2025
  
  d0443275
- Remove `Cargo.lock`, add it into .gitignore (#6438) · 17d080b7
  Liangsheng Yin authored May 20, 2025
  
  17d080b7
- Refactor communication logic of DeepSeek for extensibility and understandability (#6321) · 1b19df4b
  fzyzcjy authored May 20, 2025
  
  1b19df4b
- Expert distribution recording without overhead for EPLB (#4957) · f0653886
  fzyzcjy authored May 20, 2025
  
  f0653886
- Revert "Implement `return_hidden_states` for the OpenAI API (#6137)" (#6440) · b1465557
  Yineng Zhang authored May 19, 2025
  
  b1465557
- [BUG] fix stop_profile crash (#6431) · b06215da
  Yi Zhang authored May 20, 2025
  
  b06215da
19 May, 2025 6 commits
- [Metrics] Add KV events publishing (#6098) · 7adf245b
  Trevor Morris authored May 19, 2025
  
  7adf245b
- Fix throughput threshold for amd ci test (#6414) · 299fd22f
  Baizhou Zhang authored May 19, 2025
  
  299fd22f
- Improve supported models doc (#6430) · 506e5de8
  simveit authored May 19, 2025
  
  506e5de8
- Fix nodeepgemm init (#6417) · 844e2f22
  lukec authored May 19, 2025
  
  844e2f22
- Implement `return_hidden_states` for the OpenAI API (#6137) · 4f39bcf7
  kyle-pena-kuzco authored May 19, 2025
  
  4f39bcf7
- Fix request id error (#6401) · 31c9569b
  fzyzcjy authored May 19, 2025
  
  31c9569b
18 May, 2025 19 commits
- [Bugfix] Fix field error in v1_embedding_request (#6400) · 1be6956d
  Chang Su authored May 18, 2025
  
  1be6956d
- vlm: tensor hash kernel (#5974) · 626ccb7d
  Mick authored May 19, 2025
  
  626ccb7d
- Refactor DeepSeek MoE layer to unify the two forward branches (#6325) · 72bfb0ba
  fzyzcjy authored May 19, 2025
  
  72bfb0ba
- refactor: Extract repeated member variables in KVCache subclasses to base class. (#6323) · 15521495
  wangxiyu191 authored May 19, 2025
  
  15521495
- [Misc] Implement RankZeroFilter for rank-specific logging in model_runner.py (#6333) · ebe58d54
  Chang Su authored May 18, 2025
  
  ebe58d54
- [OAI] Add rid tracing for v1/embeddings and fix rid type in Chat (#6397) · 066cf445
  Chang Su authored May 18, 2025
  
  066cf445
- Add missing model to doc (#6396) · 6dc6b306
  applesaucethebun authored May 18, 2025
```
Co-authored-by: Brayden Zhong <b8zhong@uwaterloo.ca>
```
  6dc6b306
- [fix] fix fa3 forward_decode with spec_decode (#6395) · 1f30c05d
  JieXin Liang authored May 19, 2025
```
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
```
  1f30c05d
- Add fp8 shared_expert kernel for CPU in sgl-kernel and add UT (#6339) · 5dd62c3a
  Chunyuan WU authored May 19, 2025
```
Co-authored-by: Jiang, Yanbing <yanbing.jiang@intel.com>
Co-authored-by: mingfeima <mingfei.ma@intel.com>
```
  5dd62c3a
- Add 4-GPU runner tests and split existing tests (#6383) · f11481b9
  fzyzcjy authored May 19, 2025
  
  f11481b9
- chore: tiny remove duplicated code (#6392) · 9d24c3ff
  doujiang24 authored May 18, 2025
```
Signed-off-by: doujiang24 <doujiang24@gmail.com>
```
  9d24c3ff
- The Gemma template is missing a newline after the user role. (#6331) · 24161c59
  Yury Sulsky authored May 18, 2025
```
Co-authored-by: Yury Sulsky <ysulsky@tesla.com>
```
  24161c59
- feat: add long context example (#6391) · eabcf82a
  Yineng Zhang authored May 18, 2025
  
  eabcf82a
- Clean up AMD CI (#6365) · c47a51db
  Sai Enduri authored May 18, 2025
  
  c47a51db
- Add pipeline parallelism for Qwen2 and Qwen3 Model (#6250) · 11553c1a
  libra authored May 18, 2025
  
  11553c1a
- refactor: minor refactors regarding multimodal processing (#6187) · 01dd39ba
  Mick authored May 18, 2025
  
  01dd39ba
- Do not use FA3 for mistral (#6379) · b3f3d610
  Lianmin Zheng authored May 17, 2025
  
  b3f3d610
- chore: upgrade sgl-kernel v0.1.3 (#6377) · f07c6a00
  Yineng Zhang authored May 17, 2025
  
  f07c6a00
- Fix CI tests (#6362) · 4bb816d4
  Lianmin Zheng authored May 17, 2025
  
  4bb816d4