Update app.py

app.py

@@ -49,8 +49,14 @@ def predict(message, history):
     # messages =  system_prompt + sft_end_token.join([sft_end_token.join([f"\n{sft_start_token}{user_role}\n" + item[0], f"\n{sft_start_token}{assistant_role}\n" + item[1]])
     #                     for item in history_transformer_format])
 
-    messages = [{user_role: item[0], assistant_role: item[1]} for item in history_transformer_format]
-    model_inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(device)
+    model_messages = []
+    for item in history_transformer_format:
+        model_messages.append({"role": user_role, "content": item[0]})
+        model_messages.append({"role": assistant_role, "content": item[1]})
+
+    print(f'model_messages: {model_messages}')
+        
+    model_inputs = tokenizer.apply_chat_template(model_messages, add_generation_prompt=True, return_tensors="pt").to(device)
     # model_inputs = tokenizer([messages], return_tensors="pt").to(device)
     
     streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)