Commits · ffe9e7db82e199644dada4ea2260768052afb59b · OpenDAS / vllm_cscc

"vscode:/vscode.git/clone" did not exist on "5774b0a1da93492e60cc446e6e783a43da89e105"

07 Aug, 2025 3 commits
- [feat]支持mtp模型full_cuda_graph · ffe9e7db
  zhuwenwen authored Aug 07, 2025
  
  ffe9e7db
- [feat]支持mtp模型full_cuda_graph · f2218895
  zhuwenwen authored Aug 07, 2025
  
  f2218895
- update rocm Stream · 236f11df
  zhuwenwen authored Aug 07, 2025
  
  236f11df
06 Aug, 2025 7 commits
- update VLLM_FLASH_ATTN_V1 to VLLM_USE_FLASH_ATTN_PA · 77a29416
  zhuwenwen authored Aug 06, 2025
  
  77a29416
- update benchmark_throughput.py · 687b0ad7
  zhuwenwen authored Aug 06, 2025
  
  687b0ad7
- update warmup_sampling_params · 0896c8ef
  zhuwenwen authored Aug 06, 2025
  
  0896c8ef
- Revert "Revert "[feat]优化mtp相关函数返回类型"" · c17574bd
  zhuwenwen authored Aug 06, 2025
```
This reverts commit c34fa0bf.
```
  c17574bd
- Revert "Revert "[feat]1.支持mtp模型 full_cuda_graph; 2.优化mtp拒绝采样"" · 467490e6
  zhuwenwen authored Aug 06, 2025
```
This reverts commit 33485749.
```
  467490e6
- Revert "[feat]1.支持mtp模型 full_cuda_graph; 2.优化mtp拒绝采样" · 33485749
  zhuwenwen authored Aug 06, 2025
```
This reverts commit 93fae6b1.
```
  33485749
- Revert "[feat]优化mtp相关函数返回类型" · c34fa0bf
  zhuwenwen authored Aug 06, 2025
```
This reverts commit edc45625.
```
  c34fa0bf
05 Aug, 2025 7 commits
- [feat]优化mtp相关函数返回类型 · edc45625
  zhuwenwen authored Aug 05, 2025
  
  edc45625
- [feat]1.支持mtp模型 full_cuda_graph; 2.优化mtp拒绝采样 · 93fae6b1
  zhuwenwen authored Aug 05, 2025
  
  93fae6b1
- add tbo on v1 engine · 3d57a0d3
  zhuwenwen authored Aug 05, 2025
  
  3d57a0d3
- update lmslim import · 87d03653
  zhuwenwen authored Aug 05, 2025
  
  87d03653
- add glm4.5 k100-ai config · 92160378
  zhuwenwen authored Aug 05, 2025
  
  92160378
- add step3-vl k100-ai config · 375fdfb0
  zhuwenwen authored Aug 05, 2025
  
  375fdfb0
- when using VLLM_FLASH_ATTN_V1, set block_size to 64 · f54639f9
  zhuwenwen authored Aug 05, 2025
  
  f54639f9
04 Aug, 2025 3 commits
- add step3-vl tuning · 33d6c159
  zhuwenwen authored Aug 04, 2025
  
  33d6c159
- add step3-vl tuning · 4e2ff6b5
  zhuwenwen authored Aug 04, 2025
  
  4e2ff6b5
- update conv layout · 71e4b817
  zhuwenwen authored Aug 04, 2025
  
  71e4b817
02 Aug, 2025 1 commit
- add glm4.5 config · 54b9592c
  zhuwenwen authored Aug 02, 2025
  
  54b9592c
01 Aug, 2025 10 commits
- set default block_size to 16 · d8a060dd
  zhuwenwen authored Aug 01, 2025
  
  d8a060dd
- update N to N1 · 66092265
  zhuwenwen authored Aug 01, 2025
  
  66092265
- 增加w4a8相关支持修改 · 531c6304
  zhuwenwen authored Aug 01, 2025
  
  531c6304
- update rocm.py · 386e5c47
  zhuwenwen authored Aug 01, 2025
  
  386e5c47
- update num_rejected_tokens and num_speculative_tokens · e3cacb3b
  zhuwenwen authored Aug 01, 2025
  
  e3cacb3b
- update linear.py · 7bb28d54
  zhuwenwen authored Aug 01, 2025
  
  7bb28d54
- remove num_speculative_tokens init val · 54ab90dd
  zhuwenwen authored Aug 01, 2025
  
  54ab90dd
- remove pos_encoding_tgi_kernels.cu · b8b722eb
  zhuwenwen authored Aug 01, 2025
  
  b8b722eb
- [Model] Add step3 vl · 9f7a7815
  zhuwenwen authored Aug 01, 2025
  
  9f7a7815
- [feat]支持v1 engine mtp cudagraph · 00e13357
  zhuwenwen authored Aug 01, 2025
  
  00e13357
31 Jul, 2025 8 commits
- update unused code · 3de379de
  zhuwenwen authored Jul 31, 2025
  
  3de379de
- 去除多余的w4a8参数 · 5ad884ee
  zhuwenwen authored Jul 31, 2025
```
增加fused moe文件中w4a8的相关修改
fix: 修复W8A8读config路径错误，删除int8_utils.py文件
fix: 修复W8A8INT8读config问题
修改W4A8 以及W8A8量化量化092接口
```
  5ad884ee
- remove unused code · 84dfdb17
  zhuwenwen authored Jul 31, 2025
  
  84dfdb17
- update List[int] and update num_rejected_tokens · f137e58c
  zhuwenwen authored Jul 31, 2025
  
  f137e58c
- remove unused code · 1b78ef29
  zhuwenwen authored Jul 31, 2025
  
  1b78ef29
- update common.py · 0628e4b4
  zhuwenwen authored Jul 31, 2025
  
  0628e4b4
- update arg_utils.py · 2ae3fdba
  zhuwenwen authored Jul 31, 2025
  
  2ae3fdba
- skip conch-triton-kernels · b2d6c815
  zhuwenwen authored Jul 31, 2025
  
  b2d6c815
30 Jul, 2025 1 commit
- [Bugfix][P/D]skip fusedmoe layer for start_load_kv · 05c3c3bb
  zhuwenwen authored Jul 30, 2025
```
[Bugfix][P/D]Slove the problem that attn_medadata is not MLACommonMetadata
```
  05c3c3bb