I am trying to train model for kaggle submission. I am using all of the training data that is provided. While loading the data using from_paths, I specify the test folder. But what should I specify for val_name (valid folder). Since I am using all training data, I don’t have any valid folder. I tried specifying val_name=None, but it throws error indicating None is not allowed. If there is no valid folder in the PATH, then also it throws error indicating that no valid folder was found. How to handle this?
I’m really talking across models here (I mostly use columnar and language models so far) but in most cases I believe you can have a test_df or test_name (if available) as none while the validation set is usually a part of your training set. Again, not sure this applies in this case, hope it can help.