Commits · 389ebcf734ec1217d152a0adcc49ab7b5d26a504 · OpenDAS / vllm_cscc

24 Mar, 2025 2 commits
- [fix]修复fused_moe.py中fused_moe接口未初始化moe_ep_size导致的deekseek等模型报错 · 389ebcf7
  zhuwenwen authored Mar 24, 2025
  
  389ebcf7
- add mla tuning script and configs · 18c2244b
  zhuwenwen authored Mar 24, 2025
  
  18c2244b
18 Mar, 2025 2 commits
- 解决PA部分size计算错误的问题 · a5b976df
  zhuwenwen authored Mar 18, 2025
```
优化bf16精度
解决bf16精度问题，解决cudagraph精度问题
调整pa tc和非tc调用关系
```
  a5b976df
- update benchmarks · 10ce38cc
  zhuwenwen authored Mar 18, 2025
  
  10ce38cc
17 Mar, 2025 1 commit
- update fused_moe.py · cbeaf0b5
  zhuwenwen authored Mar 17, 2025
  
  cbeaf0b5
15 Mar, 2025 1 commit
- 支持fusemoe对int4的scale合zero合并读取操作 · 2b47d4fa
  zhuwenwen authored Mar 15, 2025
  
  2b47d4fa
14 Mar, 2025 2 commits
- 增加deepseek v3 block-int8量化支持 · ba1ff372
  zhuwenwen authored Mar 14, 2025
```
增加线性int8 gemm配置
```
  ba1ff372
- [feat]add gen_mla_pa_tables.py and optimize triton config configuration · 6188d468
  zhuwenwen authored Mar 14, 2025
  
  6188d468
13 Mar, 2025 1 commit
- update mla to obtain the optimal configuration from config · eee6148a
  zhuwenwen authored Mar 13, 2025
  
  eee6148a
12 Mar, 2025 1 commit
- [feat]添加BW deepseek_v3 fused_moe configs · abac3adc
  zhuwenwen authored Mar 12, 2025
  
  abac3adc
11 Mar, 2025 2 commits
- add VLLM_USE_TRITON_OPT_MLA to use optimized MLA attention · 1ee40c4d
  zhuwenwen authored Mar 11, 2025
```
update mla optest
```
  1ee40c4d
- [feat]优化fused_moe tuning，上传K100_AI的configs · e3a0d6bb
  zhuwenwen authored Mar 11, 2025
  
  e3a0d6bb
28 Feb, 2025 1 commit
- 1. 修复w8a8找不到对应config的bug · baf7d06a
  zhuwenwen authored Feb 28, 2025
```
2. 删除vllm的w8a8 config
```
  baf7d06a
27 Feb, 2025 1 commit
- update mtp layout · c5888d31
  zhuwenwen authored Feb 27, 2025
  
  c5888d31
26 Feb, 2025 6 commits
- 解决cudagraph计算错误的问题 · d365012a
  zhangshao authored Feb 26, 2025
  
  d365012a
- update deepseek-v2 · 96802ca0
  zhuwenwen authored Feb 26, 2025
  
  96802ca0
- add cudagraph · 4d4c6fe3
  zhuwenwen authored Feb 26, 2025
  
  4d4c6fe3
- add cudagraph · d969c150
  zhuwenwen authored Feb 26, 2025
  
  d969c150
- allow to set --enforce-eager · 537e2d9c
  zhuwenwen authored Feb 26, 2025
  
  537e2d9c
- allow to set --enforce-eager · fbfb40ad
  zhuwenwen authored Feb 26, 2025
  
  fbfb40ad
25 Feb, 2025 2 commits
- update rocm.py · 601057c9
  zhuwenwen authored Feb 25, 2025
  
  601057c9
- update version · 76621631
  zhuwenwen authored Feb 25, 2025
  
  76621631
22 Feb, 2025 3 commits
- update qwen2_vl nn lyaout · 437e6aef
  zhuwenwen authored Feb 22, 2025
  
  437e6aef
- set enforce eager · 7f881746
  zhuwenwen authored Feb 22, 2025
  
  7f881746
- support qwen2_5-vl · 0a89b8a3
  zhuwenwen authored Feb 22, 2025
  
  0a89b8a3
20 Feb, 2025 14 commits
- [Bugfix] Fix deepseekv3 grouped topk error (#13474) · ed6e9075
  chenxiaobing authored Feb 20, 2025
```
Signed-off-by: Chen-XiaoBing <chenxb002@whu.edu.cn>
```
  ed6e9075
- [Kernel] LoRA - Refactor sgmv kernels (#13110) · b69692a2
  Varun Sundar Rabindranath authored Feb 20, 2025
  
  b69692a2
- 适配deepseekv3\v2 moe awq的推理支持 · 47bd229c
  yangql authored Feb 20, 2025
  
  47bd229c
- [API Server] Add port number range validation (#13506) · 3738e6fa
  Yuan Tang authored Feb 20, 2025
```
Signed-off-by: Yuan Tang <terrytangyuan@gmail.com>
```
  3738e6fa
- [ROCm] MI300A compile targets deprecation (#13560) · 0023cd2b
  Gregory Shtrasberg authored Feb 20, 2025
  
  0023cd2b
- [Misc] add mm_processor_kwargs to extra_body for Qwen2.5-VL (#13533) · 041e2947
  燃 authored Feb 20, 2025
  
  041e2947
- [Misc] Warn if the vLLM version can't be retrieved (#13501) · 96216678
  Alex Brooks authored Feb 19, 2025
```
Signed-off-by: Alex-Brooks <Alex.brooks@ibm.com>
```
  96216678
- skip fp8 fusion · 4a734b9d
  zhuwenwen authored Feb 20, 2025
  
  4a734b9d
- [bugfix] spec decode worker get tp group only when initialized (#13578) · 8c755c3b
  Simon Mo authored Feb 19, 2025
  
  8c755c3b
- [core] add sleep and wake up endpoint and v1 support (#12987) · ba811639
  youkaichao authored Feb 20, 2025
```
Signed-off-by: youkaichao <youkaichao@gmail.com>
Signed-off-by: cennn <2523403608@qq.com>
Co-authored-by: cennn <2523403608@qq.com>
```
  ba811639
- [ROCm][MoE] mi300 mixtral8x7B perf for specific BS (#13577) · 0d243f2a
  Divakar Verma authored Feb 19, 2025
```
Signed-off-by: Divakar Verma <divakar.verma@amd.com>
```
  0d243f2a
- [Misc] Qwen2.5 VL support LoRA (#13261) · 512368e3
  Jee Jee Li authored Feb 20, 2025
  
  512368e3
- [3/n][CI] Load Quantization test models with S3 (#13570) · 473f51cf
  Kevin H. Luu authored Feb 19, 2025
```
Signed-off-by: <>
Co-authored-by: EC2 Default User <ec2-user@ip-172-31-20-117.us-west-2.compute.internal>
```
  473f51cf
- [BugFix] Avoid error traceback in logs when V1 `LLM` terminates (#13565) · a4c402a7
  Nick Hill authored Feb 19, 2025
```
Signed-off-by: Nick Hill <nhill@redhat.com>
```
  a4c402a7
19 Feb, 2025 1 commit
- [Misc] Avoid calling unnecessary `hf_list_repo_files` for local model path (#13348) · 550d97eb
  Isotr0py authored Feb 20, 2025
```
Signed-off-by: isotr0py <2037008807@qq.com>
```
  550d97eb