Update app.py

app.py

@@ -28,11 +28,10 @@ class OrcaChatBot:
         self.model = model
         self.tokenizer = tokenizer
         self.system_message = system_message
-        self.conversation_history = None
 
     def predict(self, user_message, temperature=0.4, max_new_tokens=70, top_p=0.99, repetition_penalty=1.9):
         # Prepare the prompt
-        prompt = f"<|im_start|>system\n{self.system_message}<|im_end|>\n<|im_start|>user\n{user_message}<|im_end|>\n<|im_start|>assistant" if self.conversation_history is None else self.conversation_history + f"<|im_end|>\n<|im_start|>user\n{user_message}<|im_end|>\n<|im_start|>assistant"
+        prompt = f"system\n{self.system_message}\nuser\n{user_message}\nassistant"
 
         # Encode the prompt
         inputs = self.tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
@@ -48,14 +47,10 @@ class OrcaChatBot:
             pad_token_id=self.tokenizer.eos_token_id,
             do_sample=True  # Enable sampling-based generation
         )
-        
 
         # Decode the generated response
         response = self.tokenizer.decode(output_ids[0], skip_special_tokens=True)
 
-        # Update conversation history
-        self.conversation_history = self.tokenizer.decode(output_ids[0], skip_special_tokens=False)
-
         return response
 
 Orca_bot = OrcaChatBot(model, tokenizer)