RuntimeError: Cannot re-initialize CUDA in forked subprocess.

princewang1994 / TextSnake.pytorch

A PyTorch implementation of ECCV2018 Paper: TextSnake: A Flexible Representation for Detecting Text of Arbitrary Shapes

https://arxiv.org/abs/1807.01544

MIT License

437 stars 92 forks source link

RuntimeError: Cannot re-initialize CUDA in forked subprocess. #54

Open AmrutaAnalytics opened 3 years ago

AmrutaAnalytics commented 3 years ago

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 error2

yuanjiechen commented 3 years ago

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

AtefehBaniasadi commented 3 years ago

Hello dear,

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

Thank you.