How to generate answers from a loaded GPT model

[bridgearchway]

I've loaded my pretrained model. But when I was doing 'model.generate()' and 'tokenizer.batch_decode()', the output is just nonsense. Just get stuck here.

[ht-zhou]

Thanks for your feedback.The code we support for training in example is a demo.You need to get

1. pre-trained model params
2. high quality dataset
   to train for over 1 million episodes to get an actor with good performance.