search

giotto-ai / giotto-deep

Deep learning made topological.
Other
76 stars 11 forks source link

Misleading use of train_epoch #41

Closed raphaelreinauer closed 2 years ago

raphaelreinauer commented 3 years ago

self.train_epoch is increased by 1 for every batch. https://github.com/giotto-ai/giotto-deep/blob/2965f49a5653bc51790bc76c66e21f2e5ca3e65d/gdeep/pipeline/pipeline.py#L69 However, it should only be updated every epoch, and the average cross_entropy_loss across all batches should be added to the writer.

matteocao commented 2 years ago

indeed it has been modified with train_cycle : https://github.com/giotto-ai/giotto-deep/blob/4b2f10e37408e840d57823221eeeaea3d8a31705/gdeep/pipeline/pipeline.py#L86