Skip to content

GitLab

"docs/vscode:/vscode.git/clone" did not exist on "f230cc2ca6614dd4eecf3af9f12c3ddbcf83036e"
Commit 083b80ea authored by zhuwenwen's avatar zhuwenwen
Browse files

增加w8a8相关修改

parent 09428eec
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 37 additions and 6 deletions
vllm/model_executor/models/qwen2.py
View file @ 083b80ea
This diff is collapsed.
vllm/utils.py
View file @ 083b80ea
...@@ -1466,7 +1466,11 @@ class W8a8GetCacheJSON: ...@@ -1466,7 +1466,11 @@ class W8a8GetCacheJSON:
def _initialize(self): def _initialize(self):
current_folder_path = os.path.dirname(os.path.abspath(__file__)) current_folder_path = os.path.dirname(os.path.abspath(__file__))
self.triton_json_dir=(os.getenv('TRITON_JSON_DIR', current_folder_path+'/model_executor/layers/quantization/configs/w8a8')) json_folder_path=current_folder_path+'/../lmslim/configs/w8a8'
if not os.path.exists(json_folder_path):
json_folder_path=current_folder_path+'/model_executor/layers/quantization/configs/w8a8'
self.triton_json_dir=(os.getenv('TRITON_JSON_DIR', json_folder_path))
self.triton_json_dict=[] self.triton_json_dict=[]
def getspec_config(self,configs_dict,M,N,K): def getspec_config(self,configs_dict,M,N,K):
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment