Closed aielawady closed 3 years ago
Hi, thanks for your interest in our work!
Retro* assumes a well-trained one-step model. It can be arbitrarily trained using your own data.
Or you can our provided data here: https://www.dropbox.com/sh/39a7wo0k26r5lld/AAAdineUgbu_NRSxEwaQwg7Ea?dl=0
Thank you!
Hello! Thank you for your work and for sharing the code.
I can't find the train/dev/test sets used in the paper. Is there a link to the splits or a script to reproduce the same sets to train the single step model?