About training sets

[yuanpeikai]

Hi,I found that your training set has validation set data.Could you tell me if a validation set is used for training？

Problem code： ev_vl_data = {'index': combine(train_data, valid_data), 'eval': combine(test_data)} ev_tr_data = {'index': combine(valid_data, test_data), 'eval': combine(train_data)} tr_data = {'train': combine(train_data, valid_data), 'valid': ev_vl_data['eval']}

You use tr_data for training.

[migalkin]

Hi, yes we merge train + validation when training the final model with evaluating on the test set, eg. when USE_TEST True. When doing hyperparam optimization on the validation set, only the training set is used as seen in the code:

https://github.com/migalkin/StarE/blob/f40a5ee082d61851477e9870c21e991c7d91deb3/run.py#L298-L305

[yuanpeikai]

Thanks for your reply.