Commits · 5e1ca86dca002776432fffc5fcb122cab8a0fb8a · OpenDAS / vllm_cscc

06 Sep, 2025 1 commit
- fix precision issue in mtp · 2a4a2877
  lizhigong authored Sep 05, 2025
  
  2a4a2877
04 Sep, 2025 2 commits
- fix performance issues caused by enabling TBO · 424cccfe
  lizhigong authored Sep 04, 2025
  
  424cccfe
- 暂时去掉profilling标志位，避免影响其他模型 · 8d971060
  王敏 authored Sep 04, 2025
  
  8d971060
03 Sep, 2025 1 commit
- fix deepseek pp + mtp issue · bfffd59a
  lizhigong authored Sep 02, 2025
  
  bfffd59a
27 Aug, 2025 1 commit
- when use tbo, deepseek prefill use eager · 8702264b
  lizhigong authored Aug 27, 2025
  
  8702264b
19 Aug, 2025 1 commit
- update v1 fa layout · 864c718a
  zhuwenwen authored Aug 19, 2025
  
  864c718a
18 Aug, 2025 1 commit
- [fix]解决v1 deepseek cudagraph模式显存占用增长 · 81175fe4
  王敏 authored Aug 18, 2025
  
  81175fe4
15 Aug, 2025 1 commit
- add v1 engine + deepseek r1 mtp + zero-overhead scheduler · 5dcaac2f
  lizhigong authored Aug 15, 2025
  
  5dcaac2f
14 Aug, 2025 1 commit
- remove seq_lens_tensor of v1 fa · 693d5ed4
  zhuwenwen authored Aug 14, 2025
  
  693d5ed4
13 Aug, 2025 2 commits
- add VLLM_USE_PA_PRINT_PARAM to print v1 fa size · 8a1e7a3d
  zhuwenwen authored Aug 13, 2025
  
  8a1e7a3d
- [feat]优化mtp/eagle的计算逻辑，减少第1层并行解码的计算重复 · 619eb032
  王敏 authored Aug 13, 2025
  
  619eb032
11 Aug, 2025 1 commit
- add v1 engine zero overhead · 59e80222
  lizhigong authored Aug 11, 2025
  
  59e80222
08 Aug, 2025 1 commit
- update fa full_cuda_graph support · 513f17a4
  zhuwenwen authored Aug 08, 2025
  
  513f17a4
07 Aug, 2025 2 commits
- [feat]支持mtp模型full_cuda_graph · 89eecc55
  王敏 authored Aug 07, 2025
  
  89eecc55
- [feat]支持mtp模型full_cuda_graph · a1239b53
  王敏 authored Aug 07, 2025
  
  a1239b53
06 Aug, 2025 2 commits
- [feat]支持mtp模型full_cuda_graph · 9dd945c1
  王敏 authored Aug 06, 2025
  
  9dd945c1
- Revert "Merge remote-tracking branch 'origin/v0.9.2-dev-wm' into v0.9.2-dev" · 0c1cd0f5
  zhuwenwen authored Aug 06, 2025
```
This reverts merge request !169
```
  0c1cd0f5
05 Aug, 2025 3 commits
- [feat]优化mtp相关函数返回类型 · 7e71c143
  王敏 authored Aug 05, 2025
  
  7e71c143
- merge and debug tbo on 0.9.2 · 3f8b2afe
  lizhigong authored Aug 05, 2025
  
  3f8b2afe
- [feat]1.支持mtp模型 full_cuda_graph; 2.优化mtp拒绝采样 · 8e0ae19d
  王敏 authored Aug 05, 2025
  
  8e0ae19d
31 Jul, 2025 1 commit
- [feat]支持v1 engine mtp cudagraph · fe393be8
  王敏 authored Jul 31, 2025
  
  fe393be8
24 Jul, 2025 1 commit
- support full_cuda_graph · 8da1c576
  zhuwenwen authored Jul 24, 2025
  
  8da1c576
12 Jul, 2025 1 commit
- [fix]1.优化mtp代码，解决prefix-caching不兼容问题; 2.修复v1 engine从tokenizer config中读取max_model_len导致长文本输入报错 · b16f183c
  王敏 authored Jul 12, 2025
  
  b16f183c
07 Jul, 2025 1 commit
- [feat]支持v1 engine flashmla和mtp同时开启 · ac4dedb1
  王敏 authored Jul 07, 2025
  
  ac4dedb1
06 Jul, 2025 1 commit
- [BugFix][Spec Decode] Fix spec token ids in model runner (#20530) · 9528e3a0
  Woosuk Kwon authored Jul 06, 2025
```
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
```
  9528e3a0
04 Jul, 2025 1 commit

Enable V1 for Hybrid SSM/Attention Models (#20016) · 2f35a022

Thomas Parnell authored Jul 04, 2025


Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
Co-authored-by: Stanislaw Wozniak <stw@zurich.ibm.com>
Co-authored-by: Tyler Michael Smith <tysmith@redhat.com>
Co-authored-by: Chen Zhang <zhangch99@outlook.com>

2f35a022

02 Jul, 2025 1 commit

[V1] LogitsProcessor programming model (#16728) · 48fb076c

afeldman-nm authored Jul 02, 2025


Signed-off-by: Nick Hill <nhill@redhat.com>
Signed-off-by: Andrew Feldman <afeldman@neuralmagic.com>
Signed-off-by: Andrew Feldman <afeldman@redhat.com>
Co-authored-by: Nick Hill <nhill@redhat.com>

48fb076c

01 Jul, 2025 2 commits
- [Optimization] Cache sampled token ids in model runner (#20291) · 7f280d69
  Woosuk Kwon authored Jul 01, 2025
```
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
```
  7f280d69
- [V1] Only print cudagraph tqdm on rank 0 with `is_global_first_rank` (#19516) · be250bbc
  Michael Goin authored Jul 01, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  be250bbc
30 Jun, 2025 3 commits
- [Optimization] Use Shared `CachedRequestData` Instance Across All Requests (#20232) · 2863befc
  Woosuk Kwon authored Jun 30, 2025
```
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
```
  2863befc
- [Spec Decode] Refactor spec decoding into a separate function (#20238) · 2062c072
  Woosuk Kwon authored Jun 30, 2025
```
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
```
  2062c072
- [Misc] Fix import (#20233) · 19108ef3
  Woosuk Kwon authored Jun 29, 2025
```
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
```
  19108ef3
26 Jun, 2025 1 commit
- [Feature] Expert Parallelism Load Balancer (EPLB) (#18343) · e9fd658a
  Bowen Wang authored Jun 26, 2025
```
Signed-off-by: Bowen Wang <abmfy@icloud.com>
```
  e9fd658a
24 Jun, 2025 1 commit
- [PERF] Speedup of MRoPE prepare inputs (#19939) · 9a3b8832
  Vadim Gimpelson authored Jun 24, 2025
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@centml.ai>
```
  9a3b8832
23 Jun, 2025 1 commit
- [Bugfix][v1] Fix step pooler implementation and step pooling usage in v1 (#19956) · 61f4fc5d
  Isotr0py authored Jun 24, 2025
```
Signed-off-by: Isotr0py <2037008807@qq.com>
```
  61f4fc5d
20 Jun, 2025 1 commit
- Export NaNs in logits to scheduler_stats if output is corrupted (#18777) · 2e3e3c86
  Vlad Tiberiu Mihailescu authored Jun 20, 2025
```
Signed-off-by: Vlad Mihailescu <vtmihailescu@gmail.com>
```
  2e3e3c86
19 Jun, 2025 2 commits

Support embedding models in V1 (#16188) · 799397ee

Maximilien de Bayser authored Jun 19, 2025


Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Signed-off-by: Max de Bayser <maxdebayser@gmail.com>
Signed-off-by: 22quinn <33176974+22quinn@users.noreply.github.com>
Co-authored-by: 22quinn <33176974+22quinn@users.noreply.github.com>

799397ee

[BugFix] Fix use_cudagraph=False (#19612) · ed333497
Richard Zou authored Jun 18, 2025
```
Signed-off-by: Richard Zou <zou3519@gmail.com>
```
ed333497

18 Jun, 2025 1 commit
- [v1] Support mamba2 (#19327) · a89209b7
  Chen Zhang authored Jun 19, 2025
```
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
```
  a89209b7
13 Jun, 2025 1 commit
- [CUDA] Enable full cudagraph for FlashMLA (#18581) · 3597b06a
  Luka Govedič authored Jun 13, 2025
```
Signed-off-by: luka <luka@neuralmagic.com>
```
  3597b06a