CUDA out of memory

[Backlonhw2468]

I encountered an Error saying "CUDA out of memory. Tried to allocate 2.93 GiB (GPU 0; 23.80 GiB total capacity; 12.04 GiB already allocated; 2.82 GiB free; 19.69 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".

The error above occurs when training a GAT model. Is there any way to change the "max_split_size_mb " setting in torch?

The torch version is 0.13.0. The CUDA version is 11.7.