confusion about train_set and dev_set about honk HOT 3 CLOSED

try setting --dev_pct 0
can you tell us why you need to combine the two?

from honk.

Thank you for your reply.
I know setting --dev_pct 0 can work. But I want to split the data into three set, and then combine the training set and validation set (dev_set) as a new training set and train another model with the whole new training data. Do you have any better idea that can recommend to me? Looking forward to your reply.

from honk.

That explains what you want to do, not why you want to do it.

A better way to combine the two splits would be to extend torch.utils.Dataset and rewrite the dataset splitting logic for the Speech Commands dataset. You can take a look at the (admittedly messy) utils/model.py for a general idea.

from honk.