pytorch-forecasting icon indicating copy to clipboard operation
pytorch-forecasting copied to clipboard

Published 20 hours ago • verified publisher icon jdb78

TFT Not running in Multiprocessing Pool.imap

Open thisisashukla opened this issue 3 years ago • 0 comments

  • PyTorch-Forecasting version: 0.8.4
  • PyTorch version: 1.12.1+cu102
  • Python version: 3.9.7
  • Operating System: Linux

Expected behavior

I am trying to train a TFT (pytorch forecasting) model in parallel with some other pytorch model. I am using multiprocessing to achieve this behaviour. The expected behaviour is that the model should train in multiprocessing.Pool.imap loop.

Actual behavior

The actual behaviour is that the training does not progress after printing the model summary. The progress bar for validation sanity check appears but does not progress.

Code to reproduce the problem

image image

These image have the snapshot of the issue. The training is running absolutely fine in ordinary usage (without multiprocessing).

thisisashukla avatar Aug 18 '22 04:08 thisisashukla