You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I' m working on using adapters to learn domain specific knowledge. Since my previous work is based on Pytorch Lightning, I tried to train adapters with it too. However, the document says that " AdapterTrainer is only intended for training adapters. The Trainer class should still be used to fully fine-tune models. ". I thought it means if I use Trainer, all the parameters will be finetuned.
But Pytorch Lightning's Trainer doesn't have an "Adapter" version. Is it possible I use AdapterHub with Pytorch Lightning?
The text was updated successfully, but these errors were encountered:
You can use the lightning Trainer as usual with your adapter model. Just do model.train_adapter(adapter_id) before passing it to trainer.fit, which disables training of all weights outside the task adapter.
I' m working on using adapters to learn domain specific knowledge. Since my previous work is based on Pytorch Lightning, I tried to train adapters with it too. However, the document says that "
AdapterTrainer
is only intended for training adapters. TheTrainer
class should still be used to fully fine-tune models. ". I thought it means if I useTrainer
, all the parameters will be finetuned.But Pytorch Lightning's Trainer doesn't have an "Adapter" version. Is it possible I use AdapterHub with Pytorch Lightning?
The text was updated successfully, but these errors were encountered: