When I try to train PB and PF models, I see the need to train 2 stages for each model, stage 2 (e2e) will take the checkpoint of the warp model trained from stage 1. But, I don't see you using validate when training the warp model (stage 1). How can you choose the best warp checkpoint for e2e training?
When I try to train PB and PF models, I see the need to train 2 stages for each model, stage 2 (e2e) will take the checkpoint of the warp model trained from stage 1. But, I don't see you using validate when training the warp model (stage 1). How can you choose the best warp checkpoint for e2e training?