cyberes
/
local-llm-server
Archived
1
1
Fork 0
Code Issues Pull Requests Packages Projects Releases 1 Wiki Activity

g

This commit is contained in:
Cyberes 2023-10-04 12:59:19 -06:00
parent 364b795268
commit 77db34a6a7
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llm_server/routes/request_handler.py
View File

@ -36,11 +36,13 @@ class RequestHandler:
self.token_priority, self.token_simultaneous_ip = get_token_ratelimit(self.token) self.token_priority, self.token_simultaneous_ip = get_token_ratelimit(self.token)
self.backend_url = get_a_cluster_backend(selected_model) self.backend_url = get_a_cluster_backend(selected_model)
self.cluster_backend_info = cluster_config.get_backend(self.backend_url) self.cluster_backend_info = cluster_config.get_backend(self.backend_url)
self.selected_model = self.cluster_backend_info['model']
if not self.cluster_backend_info.get('mode'): if not self.cluster_backend_info.get('mode'):
print('keyerror: mode -', selected_model, self.backend_url, self.cluster_backend_info) print('keyerror: mode -', selected_model, self.backend_url, self.cluster_backend_info)
if not self.cluster_backend_info.get('model'):
print('keyerror: mode -', selected_model, self.backend_url, self.cluster_backend_info)
self.selected_model = self.cluster_backend_info['model']
self.backend = get_backend_handler(self.cluster_backend_info['mode'], self.backend_url) self.backend = get_backend_handler(self.cluster_backend_info['mode'], self.backend_url)
self.parameters = None self.parameters = None
self.used = False self.used = False