I want to use FNET model in a seq to seq model, This seq to seq model is an encoder decoder. I want to use it for the task of abstractive summarization.
I realized that FNET model forward method does not have encoder_hidden_states parameter. While Bert model has this. This causes a problem for me because I am trying to mimic what Bert is doing as these two models are really similar. I wanted to know what can I do and What modifications are needed for FNet model to have this parameter in its forward method?