1

I've been playing around with Tensorflow in Kaggle. It works fine until I enable multiprocessing, and then it deadlocks in the first epoch, just printing

Epoch 1/100

and then stalling.

There's a public MWE here. If you remove use_multiprocessing=True from the model.fit_generator on the last line, it works fine. I would be very grateful if anyone had any idea what was causing the deadlock & how I might be able to work around it.

Mohan
  • 7,302
  • 5
  • 32
  • 55
  • I've also faced it too. [Here](https://stackoverflow.com/a/54622603/9215780) is a discussion, probably you already encounter – Innat Dec 09 '20 at 09:27
  • Are you facing it only on the kaggle platform? Is it ok for you in colab or your local machine? – Innat Dec 09 '20 at 09:29
  • @Mohan, Can you please let us know on which platform are you facing issue? –  Dec 10 '20 at 15:41
  • I've only tried on kaggle, I'm afraid. My main PC died and I don't have a machine set up for running tensorflow on sensibly right now, so I'm restricted to browser-based code for a while. – Mohan Dec 10 '20 at 16:15

0 Answers0