Closed Wallace-222 closed 1 year ago
if the clip architecture uses transformer, do we still need to use model.eval() to fix all the norm layers in the training code? Thanks a lot.
Yes, we still use model.eval() in all experiments even if the backbone architecture is Transformer.
model.eval()
if the clip architecture uses transformer, do we still need to use model.eval() to fix all the norm layers in the training code? Thanks a lot.