Fix bug of lmdeploy.LMDeployServer (#132)

lagent/llms/lmdepoly_wrapper.py

@@ -290,6 +290,7 @@ class LMDeployServer(BaseModel):
                  serve_cfg=dict(),
                  **kwargs):
         super().__init__(path=path, **kwargs)
+        self.model_name = model_name
         # TODO get_logger issue in multi processing
         import lmdeploy
         self.client = lmdeploy.serve(
@@ -332,7 +333,7 @@ class LMDeployServer(BaseModel):
 
         resp = [''] * len(inputs)
         for text in self.client.completions_v1(
-                self.path,
+                self.model_name,
                 inputs,
                 session_id=session_id,
                 sequence_start=sequence_start,
@@ -382,7 +383,7 @@ class LMDeployServer(BaseModel):
         finished = False
         stop_words = self.gen_params.get('stop_words')
         for text in self.client.completions_v1(
-                self.path,
+                self.model_name,
                 prompt,
                 session_id=session_id,
                 sequence_start=sequence_start,