Colab Pro getting disconnected automatically after 12 hours of Training Deep Learning Model using TPU Runtime

Viewed 279

I am using CoLab Pro for few days. I was trying to Train a Deep Learning Model using TPU Runtime in CoLab Pro. But the Runtime is getting automatically disconnected after 12 hours of Training. I have configured 70 epochs for Training and it stops at around 43 - 52 epochs. I have used Keras/Tensorflow for developing my model. For the last three days I am Training the same model but I am no able to complete the Training Process. It is frustrating that even after opting for CoLab Pro, we are not able to train the model fully.

1 Answers

When I tried with Google Chrome, I was able to perform the Training for around 11 hours in Colab Pro. So I think in Google Chrome, automatic disconnection may not happen even after 12 hours. Automatic disconnection will happen if we are using Mozilla. I have tested several times in Mozilla but it is getting disconnected after 8 or 12 hours.

Related