Update app.py

app.py

@@ -53,13 +53,14 @@ def predict(message, history):
     print(f'history: {history}')
     for i, item in enumerate(history):
         model_messages.append({"role": user_role, "content": item[0]})
-        if i < len(history) - 1:
-            model_messages.append({"role": assistant_role, "content": item[1]})
+        model_messages.append({"role": assistant_role, "content": item[1]})
 
+    model_messages.append({"role": user_role, "content": message})
+    
     print(f'model_messages: {model_messages}')
-        
+
+    print(f'model_final_inputs: {tokenizer.apply_chat_template(model_messages, add_generation_prompt=True, tokenize=False)}, flash=True')
     model_inputs = tokenizer.apply_chat_template(model_messages, add_generation_prompt=True, return_tensors="pt").to(device)
-    print(f'model_final_inputs: {tokenizer.apply_chat_template(model_messages, add_generation_prompt=True, tokenize=False)}')
     # model_inputs = tokenizer([messages], return_tensors="pt").to(device)
     
     streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)