Fine-tuned Model

[jiangix-paper]

Hello, Thanks a lot for your great works. I want to know the principle of fine-tuned model. For example, if i want to fine-tune the pretrained model on supervised cnn/dailymail dataset, is the model seq2seq? And i only load the pretrained word embedding based on pegasus, and set it as the input of seq2seq?

[JingqingZ]

If i want to fine-tune the pretrained model on supervised cnn/dailymail dataset, is the model seq2seq?

The model uses the encoder-decoder Transformer framework so yes it is seq2seq.

And i only load the pretrained word embedding based on pegasus, and set it as the input of seq2seq?

Sorry, I am not sure if I understand this question. Could you elaborate on your question?

[jiangix-paper]

If i want to fine-tune the pretrained model on supervised cnn/dailymail dataset, is the model seq2seq?

The model uses the encoder-decoder Transformer framework so yes it is seq2seq.

And i only load the pretrained word embedding based on pegasus, and set it as the input of seq2seq?

Sorry, I am not sure if I understand this question. Could you elaborate on your question?

[jiangix-paper]

Thanks for your reply. When I fine-tune the model with encoder-decoder transformer framework on cnn/dailymail dataset, whether the model structure and initial model parameter is the same as pretrained model. The model parameter of transformer encoder and decoder is updated during fine-tuning.

I want to sure whether the fine-tuned model are as follows: