Can the fastai.distributed library be used with an RNN

I see the examples with a cnn learner, but could I train a language model on 8 GPU’s using the distributed library?


You’ll get a bug with the RNN, specifically the language model, since the dataloader behind the scenes isn’t a real dataloader and thus doesn’t take a custom sampler. I’ll try to fix it soon.

Sounds good and thanks for your support!