Update app.py

app.py

@@ -230,19 +230,19 @@ def generate(
     
     # Construct the input prompt using the functions from the construct_input_prompt function 
     input_prompt = construct_input_prompt(chat_history, message)
-
+    
     # Move the condition here after the assignment
     if input_prompt:
         conversation.append({"role": "system", "content": input_prompt})
+
+    # Convert input prompt to tensor
+    input_ids = tokenizer(input_prompt, return_tensors="pt").to(model.device)
+    
     for user, assistant in chat_history:
         conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
     conversation.append({"role": "user", "content": message})
 
-    # Convert input prompt to tensor
-    # input_ids = tokenizer(input_prompt, return_tensors="pt").to(model.device)
-    # Original from HuggingFace Llama2 Chatbot: input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
-    
-    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt").to(model.device)
+    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
     if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
         input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
         gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")