RuntimeError: Cannot re-initialize CUDA in forked subprocess.

[AmrutaAnalytics]

Getting RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method. working with colab. Here pytorch version is 1.7.0

[yuanjiechen]

Use python multiprocessing without torch.multiprocessing, and write multiprocessing.set_start_method('spawn') in main process

[AtefehBaniasadi]

Hello dear,

If I want to run this repo without multiprocessing what should I do?

Thank you.