Luca Ballore
Results
2
comments of
Luca Ballore
Same problem here :) @rahiakela did you manage to solve this in some way? As a workaround, I actually set the `batch_size` to 20 again, so that `steps_per_epoch * batch_size`...
Nice! But you could obtain better performance by setting `steps_per_epoch=100`, `batch_size=20` and `epochs=150`. This should make your curves less noisy. Look at my example here: https://github.com/lucone83/deep-learning-with-python/blob/master/notebooks/chapter_05/02%20-%20Using%20convnets%20with%20small%20datasets.ipynb