Kokonut133

same question, differently formulated. Both dont seem to slow down training and GPU memory usage doesnt seem to have increased. Please let us know how to apply this. ``` def...

@pidajay I would be happy to see your implementation of this. If you could share a link, that would be great @davisyoshida So as I understand, I did apply it...

@davisyoshida So Itried implementing your solution and got to this: ``` def build_discriminator(self, input_shape): @checkpointable def discriminator_layer(input, filters, batch_norm, f_size=4, _checkpoint=True): d = Conv2D(filters, kernel_size=f_size, strides=2, padding="same", activation="relu")(input) return d...

Ok. So, I believe to understand more. Yet, I am uncertain exactly how to apply it. I have a Model (tf model) which i train with a custom train function....