RuntimeError: CUDA error: out of memoryCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.For debugging consider passing CUDA_LAUNCH_BLOCKING=1.Compile with `TORCH_USE_CUDA_DSA` to enab..

문제

GPU 상의 cache를 정리해주기 위해 torch.cuda.empty_cache() 사용 시 에러

torch.cuda.empty_cache()

>> RuntimeError: CUDA error: out of memoryCUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.For debugging consider passing CUDA_LAUNCH_BLOCKING=1.Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

 

해결

with torch.cuda.device('cuda:0'):
	torch.cuda.empty_cache()

 

참고

Out of memory when I use torch.cuda.empty_cache - PyTorch Forums