latent-diffusion
latent-diffusion copied to clipboard
RuntimeError: DataLoader worker (pid 13772) is killed by signal: Killed.
when I run the code with DDP (8 Tesla V100 ), this problem will arise. Setting num_workers = 0 may solve the problem , however, the utilization of GPU will decrease a lot. Does anyone have a better solution, thank you
when I run the code with DDP (8 Tesla V100 ), this problem will arise. Setting num_workers = 0 may solve the problem , however, the utilization of GPU will decrease a lot. Does anyone have a better solution, thank you
https://github.com/CompVis/latent-diffusion/issues/17