esm
esm copied to clipboard
Allow inverse folding model to score sequences on GPU
I modified the get_sequence_loss
function to use whichever device the model is currently on to also be the device that the batch of sequences are put on.