Update app.py

app.py

@@ -58,8 +58,10 @@ def generate_response(transcription):
         tokenized_chat = f'<|system|>\n{sys_prompt}{eos_token}<|user|>\n{transcription}{eos_token}<|assistant|>\n'
 
         print(tokenized_chat)
+
+        tokenized_chat = tokenizer(tokenized_chat)
         
-        input_ids = tokenized_chat[0].to(device)
+        input_ids = tokenized_chat['input_ids'].to(device)
         if len(input_ids.shape) == 1:
             input_ids = input_ids.unsqueeze(0)
         with torch.no_grad():