Commits · b98cf398666614133050feb57a735c8746c65d03 · change / sglang · GitLab

14 Oct, 2025 17 commits
- [Auto Sync] Update collector.py (20251014) (#11625) · b98cf398
  Lianmin Zheng authored Oct 14, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Byron Hsu <byronhsu1230@gmail.com>
```
  b98cf398
- [Auto Sync] Update scheduler.py, server_args.py (20251014) (#11623) · 27d71045
  Lianmin Zheng authored Oct 14, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
```
  27d71045
- Fix log for chunked prefix cache (#11624) · c224a4c6
  Baizhou Zhang authored Oct 14, 2025
  
  c224a4c6
- use non_blocking h2d in ForwardBatch.prepare_mlp_sync_batch. (#11605) · 94d26d85
  strgrb authored Oct 15, 2025
  
  94d26d85
- Reduce one step decode for draft model. (#11561) · 5ea96ac7
  Liangsheng Yin authored Oct 14, 2025
  
  5ea96ac7
- move eagle draft post process to cuda graph (#11434) · 56222658
  yinghui authored Oct 14, 2025
```
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>
```
  56222658
- make radix cache deterministic (#10721) · dc965db0
  Alex Chi Z authored Oct 14, 2025
```
Signed-off-by: Alex Chi Z <iskyzh@gmail.com>
```
  dc965db0
- Refactor spec decoding metrics calculation into separate `TokenizerManager`... · 817e46f4
  Scott Lee authored Oct 14, 2025
```
Refactor spec decoding metrics calculation into separate `TokenizerManager` utility function (#11586)
```
  817e46f4
- Optimize Triton Draft Backend (#11556) · 5a33c3aa
  Liangsheng Yin authored Oct 14, 2025
  
  5a33c3aa
- Add fused_moe_triton config: triton_3_4_0/E=256,N=256,device_name=NVIDIA_B200.json (#11587) · e4358a45
  Qiaolin Yu authored Oct 13, 2025
  
  e4358a45
- [Auto Sync] Update model_config.py (20251014) (#11580) · ba2ce28f
  Lianmin Zheng authored Oct 13, 2025
```
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Hanming Lu <69857889+hanming-lu@users.noreply.github.com>
```
  ba2ce28f
- Implement LRU eviction policy for LoRA adapters (#11041) · 28f80b12
  Chenxi Li authored Oct 13, 2025
  
  28f80b12
- bench_serving support PD Disaggregation (#11542) · 88a6f9da
  Xiaoyu Zhang authored Oct 14, 2025
  
  88a6f9da
- Make DeepEP combine recv do not overlap (#11535) · cb8ed2c0
  fzyzcjy authored Oct 14, 2025
  
  cb8ed2c0
- [DSv32] Use torch.compile for _get_logits_head_gate (#11565) · 38473363
  Trevor Morris authored Oct 13, 2025
  
  38473363
- [FEATURE] Add Profile Trace Merger for Distributed Traces (#11413) · aaf7af1b
  Neelabh Sinha authored Oct 13, 2025
  
  aaf7af1b
- Compilation Folder Reset (#11539) · 932e2637
  Yuwei An authored Oct 13, 2025
```
Signed-off-by: Oasis-Git <ayw.sirius19@gmail.com>
```
  932e2637
13 Oct, 2025 22 commits
- Fix accept rate in speculative decoding metrics (#11572) · 43f80884
  Qiaolin Yu authored Oct 13, 2025
  
  43f80884
- [router][grpc] Add `serve_grpc` to `launch_server` and log id for HealthCheck (#11564) · 887c2b45
  Chang Su authored Oct 13, 2025
  
  887c2b45
- Tiny cleanup fp4 gemm calls (#11537) · 065ce815
  fzyzcjy authored Oct 14, 2025
  
  065ce815
- [NVIDIA] update pyproject.toml to support cu130 option (#11521) · cb8f3d90
  Johnny authored Oct 13, 2025
  
  cb8f3d90
- Fix DeepSeek-v3.2 default config (ValueError: not enough values to unpack... · c9cff2b9
  Trevor Morris authored Oct 13, 2025
```
Fix DeepSeek-v3.2 default config (ValueError: not enough values to unpack (expected 4, got 3)) (#11557)
```
  c9cff2b9
- Add metrics for speculative decoding (acceptance rate, average acceptance length) (#11441) · b6fb5d76
  Scott Lee authored Oct 13, 2025
  
  b6fb5d76
- Minor: improve sampler & remove unused fields from model_config.py (#11531) · 5e3f7e7f
  Lianmin Zheng authored Oct 13, 2025
  
  5e3f7e7f
- Move deep gemm related arguments to `sglang.srt.environ` (#11547) · acc2327b
  Liangsheng Yin authored Oct 14, 2025
  
  acc2327b
- Adjust overlap event loop (#11507) · bfadb5ea
  Liangsheng Yin authored Oct 14, 2025
  
  bfadb5ea
- [router][Fix] Include grpc reflection runtime dependency (#11419) · 9cc1e065
  ai-jz authored Oct 13, 2025
```
Co-authored-by: Chang Su <chang.s.su@oracle.com>
```
  9cc1e065
- fix: fix video input for qwen3-vl (#11442) · f35f120d
  Mick authored Oct 14, 2025
  
  f35f120d
- Remove `tp_worker.worker` (#11548) · 54a46a26
  Liangsheng Yin authored Oct 13, 2025
  
  54a46a26
- [Fix] Add per_channel_quant parameter to MoE config functions (#11201) · c7867b67
  Mohammad Miadh Angkad authored Oct 13, 2025
  
  c7867b67
- Depreate `global_server_args_dict` (#11528) · 516738b0
  Liangsheng Yin authored Oct 13, 2025
  
  516738b0
- [Reland] perf: optimize qwen-vl with symm mem allreduce (#11457) · 0b6f535f
  Yuan Luo authored Oct 13, 2025
```
Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>
```
  0b6f535f
- [HICache]: Support 3FS-Store with page_first_direct layout (#11460) · 318424e2
  hzh0425 authored Oct 13, 2025
  
  318424e2
- ci: improve nightly-ci (#11385) · 0c0779d6
  Mick authored Oct 13, 2025
  
  0c0779d6
- [Feature] Support mamba radix cache v0 (#11214) · a55cf530
  Yi Zhang authored Oct 13, 2025
```
Co-authored-by: hanming-lu <hanming@x.ai>
Co-authored-by: hzh0425 <hzh0425@apache.org>
Co-authored-by: thalahors <ericalcaide1@gmail.com>
```
  a55cf530
- [Fix]: add missing device attribute to ChunkCache (#11493) · 19ba16aa
  Yuanhang Sun authored Oct 13, 2025
  
  19ba16aa
- Update DeepSeek-R1-FP4 default config on blackwell (#11512) · a2b3d9b9
  Qiaolin Yu authored Oct 12, 2025
  
  a2b3d9b9
- Improve dp attention port assignment scheme (#5889) · a20e7df8
  Yongtong Wu authored Oct 13, 2025
```
Co-authored-by: Cheng Wan <cwan@x.ai>
```
  a20e7df8
- Revert "Deprecate `global_server_args_dict`" (#11520) · 1bdd0102
  Cheng Wan authored Oct 12, 2025
  
  1bdd0102
12 Oct, 2025 1 commit
- Sync changes on io_struct.py and deterministic ops (#11498) · 2ac46e94
  Lianmin Zheng authored Oct 12, 2025
  
  2ac46e94