search

modern-fortran / neural-fortran

A parallel framework for deep learning
MIT License
395 stars 82 forks source link

Add missing update for `conv2d_layer` #141

Closed milancurcic closed 1 year ago

milancurcic commented 1 year ago

Conv2d layers were previously not getting their parameters updated during training. cnn_mnist now converges.

milancurcic commented 1 year ago

cnn_mnist converges with or without this fix but to a relatively low accuracy (~93% in 10 epochs), whereas it should easily get to >98%. While the fix introduced in this PR is necessary for updating conv2d layers to work, there is another bug elsewhere that is causing bias and weight gradients to remain zero during training.