GPU out of memory error

[ajati]

https://github.com/thuml/SimMTM/blob/main/SimMTM_Forecasting/scripts/pretrain/ECL_script/ECL.sh fails with GPU out of memory error in a A100 80GB.

RuntimeError: CUDA out of memory. Tried to allocate 13.26 GiB (GPU 0; 79.15 GiB total capacity; 60.37 GiB already allocated; 7.43 GiB free; 70.33 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[linxi20]

Hello, I have also encountered this issue. Have you resolved it now?

[iDestro]

I also have this problem, why?

Hardware: 3090

Script: bash ./scripts/pretrain/ECL_script/ECL.sh

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 207.37 GiB (GPU 0; 23.70 GiB total capacity; 8.14 GiB already allocated; 13.37 GiB free; 9.09 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF