Commits · c555ce2ca20cd8a2fc87a0e048c39c181614388e · change / sglang · GitLab

25 Oct, 2024 7 commits
- Revert "Fix memory leak when doing chunked prefill" (#1797) · c555ce2c
  Lianmin Zheng authored Oct 25, 2024
  
  c555ce2c
- [Fix] Fix the log parsing in chunked prefill uni tests (#1794) · 40900bae
  Lianmin Zheng authored Oct 25, 2024
  
  40900bae
- Fix memory leak when doing chunked prefill (#1787) · a2f5e755
  Liangsheng Yin authored Oct 25, 2024
  
  a2f5e755
- Fix log parsing in the chunked prefill unit tests (#1793) · 2148914e
  Lianmin Zheng authored Oct 25, 2024
  
  2148914e
- Qwen2vl support cuda graph and disable radix cache (#1780) · def55bc8
  yizhang2077 authored Oct 25, 2024
  
  def55bc8
- [Fix] Fix seq_lens_sum for cuda graph runner in padded cases (#1789) · 86a2c473
  Lianmin Zheng authored Oct 24, 2024
  
  86a2c473
- Enhance the test case for chunked prefill (#1785) · 1701b0db
  Lianmin Zheng authored Oct 24, 2024
  
  1701b0db
24 Oct, 2024 6 commits
- Re-introduce `get_cuda_graph_seq_len_fill_value` (#1783) · 384d85ba
  Lianmin Zheng authored Oct 24, 2024
  
  384d85ba
- check user-specified model_max_len with hf derived max_model_len (#1778) · 60597219
  Xiaoyu Zhang authored Oct 25, 2024
  
  60597219
- [Fix] Fix cuda graph padding for triton attention backend (#1782) · fc82f5a7
  Lianmin Zheng authored Oct 24, 2024
  
  fc82f5a7
- [Fix] Fix NaN issues by fixing the cuda graph padding values for flashinfer (#1779) · 0089c4bc
  Lianmin Zheng authored Oct 24, 2024
  
  0089c4bc
- [Bug] Catch any errors caused by parsing json schema (#1776) · 72e7b57a
  zolinthecow authored Oct 24, 2024
  
  72e7b57a
- Fix MockTokenizer in the unit tests (#1774) · 87a7cfa0
  Lianmin Zheng authored Oct 23, 2024
  
  87a7cfa0
23 Oct, 2024 11 commits
- Fix the perf regression due to additional_stop_token_ids (#1773) · 8f8f96a6
  Lianmin Zheng authored Oct 23, 2024
  
  8f8f96a6
- Crash the server on warnings in CI (#1772) · 05b3bf5e
  Lianmin Zheng authored Oct 23, 2024
  
  05b3bf5e
- Fix out of memory message. (#1771) · 3f5ac88d
  Liangsheng Yin authored Oct 23, 2024
  
  3f5ac88d
- Fix missing additional_stop_token_ids (#1769) · 0d800090
  Lianmin Zheng authored Oct 23, 2024
  
  0d800090
- Update docs (#1768) · b7d05594
  Lianmin Zheng authored Oct 23, 2024
```
Co-authored-by: Chayenne Zhao <zhaochenyang20@gmail.com>
Co-authored-by: Chayenne <zhaochen20@outlook.com>
```
  b7d05594
- Fix stop condition for <|eom_id|> (#1766) · 80a90547
  Lianmin Zheng authored Oct 23, 2024
  
  80a90547
- [Fix] Fix abort in dp (#1767) · 9af7b88e
  Lianmin Zheng authored Oct 23, 2024
  
  9af7b88e
- Fix perf regression for set_kv_buffer (#1765) · fbcbb263
  Lianmin Zheng authored Oct 23, 2024
  
  fbcbb263
- [API] add get memory pool size (#1760) · 2fce449b
  Ying Sheng authored Oct 23, 2024
```
Co-authored-by: Byron Hsu <byronhsu1230@gmail.com>
```
  2fce449b
- Fuse more ops & Simplify token mapping (#1758) · ad4125d1
  Lianmin Zheng authored Oct 22, 2024
  
  ad4125d1
- Fix edge case for truncated (#1747) · 17536e7e
  Byron Hsu authored Oct 22, 2024
  
  17536e7e
22 Oct, 2024 1 commit
- Release v0.3.4.post1 (#1749) · 1f26e8b8
  Lianmin Zheng authored Oct 21, 2024
  
  1f26e8b8
21 Oct, 2024 11 commits
- Update `max_req_len` and `max_req_input_len` (#1748) · 5e1558f1
  Liangsheng Yin authored Oct 21, 2024
  
  5e1558f1
- Llama3.2 vision model support (#1551) · 94cde109
  Liangsheng Yin authored Oct 21, 2024
  
  94cde109
- Fix sliding window attention and gemma-2 unit tests in CI (#1746) · 00611286
  Lianmin Zheng authored Oct 21, 2024
  
  00611286
- misc: add CODEOWNERS (#1737) · e68b9e76
  Yineng Zhang authored Oct 21, 2024
  
  e68b9e76
- Faster overlap mode scheduler (#1738) · 7ce36068
  Lianmin Zheng authored Oct 21, 2024
  
  7ce36068
- Fix prefill oom (#1743) · efb099cd
  Liangsheng Yin authored Oct 21, 2024
  
  efb099cd
- Maintain seq_lens_sum to make more FlashInfer operations non-blocking (#1741) · 09603c6d
  Lianmin Zheng authored Oct 21, 2024
  
  09603c6d
- Make token mapping non-blocking in the overlapped mode (#1740) · cf470fea
  Lianmin Zheng authored Oct 20, 2024
  
  cf470fea
- Add GLM-4 TextGeneration Model support for SGLang (#1736) · 45d5af24
  sixgod authored Oct 21, 2024
  
  45d5af24
- Simplify batch result resolution (#1735) · b121bc03
  Lianmin Zheng authored Oct 20, 2024
  
  b121bc03
- Simplify the usage of device (#1734) · e12358dc
  Lianmin Zheng authored Oct 20, 2024
  
  e12358dc
20 Oct, 2024 4 commits
- [Bugfix] qwen2vl forward_extend (#1727) · 554fbf93
  yizhang2077 authored Oct 20, 2024
  
  554fbf93
- Split the overlapped version of TpModelWorkerClient into a separate file (#1726) · b48edff6
  Lianmin Zheng authored Oct 20, 2024
  
  b48edff6
- Temporarily skip this test_mixed_batch for QWen2VL (#1725) · 593b19f2
  Lianmin Zheng authored Oct 20, 2024
  
  593b19f2
- Unify the memory pool api and tp worker API (#1724) · 59cbf476
  Lianmin Zheng authored Oct 19, 2024
  
  59cbf476