How does num_training_steps relate to epochs and the size of the dataset?

TooTouch / MemSeg

Unofficial re-implementation of MemSeg for Anomaly Detection

MIT License

184 stars 31 forks source link

Closed TorAP closed 1 year ago

TooTouch commented 1 year ago

Hi, @TorAP

num_training_steps is an argument indicating how many times to repeat the training by the batch size.

TorAP commented 1 year ago

Yes, I do know this.. but is num_training_steps: 1000 equal 1 epoch or what is the relation?