Reputation: 467
I do wonder what is the parameter 'eval_steps' in learn.experiment in tensorflow ? Why would you run over the evaluation set several times every time you want to evaluate your model ?
Thanks!
Upvotes: 2
Views: 1212
Reputation: 5808
If your evaluation is a whole epoch, you're right that it doesn't make much sense. eval_steps
is more for the case when you're doing mini-batch evaluation and want to evaluate on multiple mini-batches.
Upvotes: 2