Commits · 02032631d608536ab2510f2137d105139947dfd2 · OpenDAS / vllm_cscc

31 Dec, 2025 3 commits
- Revert "增加mtp>1的pad" · d8c36dbf
  yangql authored Dec 31, 2025
```
This reverts commit 059b1961.
```
  d8c36dbf
- [feat]解决deepep auto模式 mtp >1 时卡住 · 37d2b35d
  王敏 authored Dec 31, 2025
  
  37d2b35d
- 增加mtp>1的pad · 059b1961
  yangql authored Dec 31, 2025
  
  059b1961
29 Dec, 2025 1 commit
- 增加mtp的pad · e60a51b4
  yangql authored Dec 29, 2025
  
  e60a51b4
24 Dec, 2025 2 commits
- [feat]解决EP开启pd_padding大输入卡住问题 · 7439b205
  王敏 authored Dec 24, 2025
  
  7439b205
- [fix]解决EP开启pd_padding大输入卡住问题 · 639a11d2
  王敏 authored Dec 24, 2025
  
  639a11d2
22 Dec, 2025 2 commits
- [fix]1.解决ep sequence parallel优化引入的mtp报错;2.解决共享专家无法和combine overlap问题 · dc01fce4
  王敏 authored Dec 22, 2025
  
  dc01fce4
- [feat]1.优化ep sequence parallel,区分主模型和mtp逻辑;2.ep sequence parallel添加cudagraph... · 62f05dde
  王敏 authored Dec 22, 2025
```
[feat]1.优化ep sequence parallel,区分主模型和mtp逻辑;2.ep sequence parallel添加cudagraph padding到tp_size;3.修复共享专家和deepep combine overlap
```
  62f05dde
18 Dec, 2025 2 commits
- [feat]合入基于deepep的大EP · 13130b89
  王敏 authored Dec 18, 2025
  
  13130b89
- [feat]优化dp attention，减少1次allgather耗时，高吞吐提升明显 · 3e386c3b
  王敏 authored Dec 18, 2025
  
  3e386c3b
02 Dec, 2025 1 commit
- [fix]解决dp模式mtp卡住问题 · 55ee9d72
  王敏 authored Dec 02, 2025
  
  55ee9d72
27 Nov, 2025 1 commit
- [fix]解决宽松mtp1报错 · deae0a22
  王敏 authored Nov 27, 2025
  
  deae0a22
26 Nov, 2025 1 commit
- [feat]支持宽松mtp · 02b4ec43
  王敏 authored Nov 26, 2025
  
  02b4ec43
27 Oct, 2025 1 commit
- [feat]w4a8适配deepep ht模式，解决开启dp时mtp>1时卡住问题 · 98b7432a
  王敏 authored Oct 27, 2025
  
  98b7432a
16 Sep, 2025 1 commit
- 添加mori ep · 1a56d6cb
  王敏 authored Sep 16, 2025
  
  1a56d6cb
14 Sep, 2025 1 commit
- [fix]修复eagle 创建cu_num_tokens类型错误问题 · b4610c06
  王敏 authored Sep 14, 2025
  
  b4610c06
04 Sep, 2025 1 commit
- 1.优化大EP，合入grouped gemm · d997afc4
  王敏 authored Sep 04, 2025
```
2.解决mtp >1 大EP推理all gather卡住问题
```
  d997afc4
13 Aug, 2025 1 commit
- [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
10 Aug, 2025 1 commit
- [fix]修复v1 mtp接受率低的问题 · 04b61f0e
  王敏 authored Aug 10, 2025
  
  04b61f0e
07 Aug, 2025 2 commits
- [feat]支持mtp模型full_cuda_graph · bd58c289
  王敏 authored Aug 07, 2025
  
  bd58c289
- [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 2 commits
- [feat]优化mtp相关函数返回类型 · 7e71c143
  王敏 authored Aug 05, 2025
  
  7e71c143
- [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
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
- [V1] Support any head size for FlexAttention backend (#20467) · 9fb52e52
  Cyrus Leung authored Jul 07, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  9fb52e52
30 Jun, 2025 1 commit
- [fix]解决v1 engine mtp只能设为1的问题 · 504c262e
  王敏 authored Jun 30, 2025
  
  504c262e
25 Jun, 2025 1 commit
- [V1][Speculative Decoding] Fix DeepSeek MTP (#20022) · 8359f4c8
  cjackal authored Jun 26, 2025
```
Signed-off-by: cjackal <44624812+cjackal@users.noreply.github.com>
```
  8359f4c8
13 Jun, 2025 3 commits
- [CUDA] Enable full cudagraph for FlashMLA (#18581) · 3597b06a
  Luka Govedič authored Jun 13, 2025
```
Signed-off-by: luka <luka@neuralmagic.com>
```
  3597b06a
- [torch.compile] reorganize the cache directory to support compiling multiple models (#19064) · d70bc7c0
  youkaichao authored Jun 13, 2025
```
Signed-off-by: youkaichao <youkaichao@gmail.com>
```
  d70bc7c0
- [Bugfix] Fix EAGLE vocab embedding for multimodal target model (#19570) · c68698b3
  qizixi authored Jun 12, 2025
```
Signed-off-by: qizixi <qizixi@meta.com>
```
  c68698b3
06 Jun, 2025 1 commit
- [Bugfix] Fix EAGLE vocab embedding construction for Llama 70B (#19033) · 3465b87e
  Benjamin Chislett authored Jun 05, 2025
```
Signed-off-by: Benjamin Chislett <benjamin.chislett@centml.ai>
```
  3465b87e
03 Jun, 2025 1 commit
- [Misc] Add SPDX-FileCopyrightText (#19100) · 02f0c7b2
  Simon Mo authored Jun 03, 2025
```
Signed-off-by: simon-mo <simon.mo@hey.com>
```
  02f0c7b2
28 May, 2025 1 commit
- [V1] [Bugfix] eagle bugfix and enable correct lm_head for multimodal (2) (#18781) · 5e13c07d
  RonaldBXu authored May 27, 2025
```
Signed-off-by: Ronald Xu <ronaldxu@amazon.com>
```
  5e13c07d
24 May, 2025 2 commits
- [V1][Spec Decode] Support multi-layer eagle draft model (#18030) · c1e4a405
  qizixi authored May 24, 2025
```
Signed-off-by: qizixi <qizixi@meta.com>
```
  c1e4a405
- [V1][Spec Decode] Small refactors to improve eagle bookkeeping performance (#18424) · d55e446d
  qizixi authored May 23, 2025
```
Signed-off-by: qizixi <qizixi@meta.com>
```
  d55e446d
23 May, 2025 1 commit

[V1] Support Deepseek MTP (#18435) · 2628a69e

Jiayi Yao authored May 23, 2025


Signed-off-by: Rui Qiao <ruisearch42@gmail.com>
Signed-off-by: YaoJiayi <120040070@link.cuhk.edu.cn>
Co-authored-by: Rui Qiao <ruisearch42@gmail.com>

2628a69e