Training become very slow after some epochs!

[duyhominhnguyen]

Hi authors,

Thank you for your interesting work! I'm running your provided example but the training step becomes very slow after epochs 40; especially at loss_i = self.loss(pred, Y[i]) and pred = self.predict(X[i]) in method.py. For e.g, times used for predict at epoch 49 is 4.12 secs, compute loss is 5.66 secs while at initial epochs they are just approximately 0.0025 and 0.0026 respectively. Do you have any ideas why these issues occur? Thank you.