Update app.py

app.py

@@ -122,6 +122,7 @@ ROLE_PROMPTS = {}
 ROLE_PROMPTS["AI Assistant"]=system_message
 ##"You are an AI assistant with Zephyr model by Mistral and Hugging Face and speech from Coqui XTTS . User will you give you a task. Your goal is to complete the task as faithfully as you can. While performing the task think step-by-step and justify your steps, your answers should be clear and short sentences"
 
+LLM_STOP_WORDS= ["</s>","<|user|>","/s>"]
 
     
 ### WILL USE LOCAL MISTRAL OR ZEPHYR
@@ -191,7 +192,7 @@ def generate_local(
     temperature=0.8,
     max_tokens=256,
     top_p=0.95,
-    stop = ["</s>","<|user|>"]
+    stop = LLM_STOP_WORDS
 ):
     temperature = float(temperature)
     if temperature < 1e-2:
@@ -225,7 +226,7 @@ def generate_local(
                 return
                 
             
-            output += response["choices"][0]["text"].replace("<|assistant|>","").replace("<|user|>","")
+            output += response["choices"][0]["text"].replace("<|assistant|>","").replace("<|user|>","").replace("/s>","")
             yield output
 
     except Exception as e: