[RuntimeError: CUDA out of memory] when max dimensions of final image are increased

[VygintasC]

I am getting the following error when I increase the --max-dim parameter above 700 on a 6GiB GPU.

RuntimeError: CUDA out of memory. Tried to allocate 208.00 MiB (GPU 0; 6.00 GiB total capacity; 5.29 GiB already allocated; 0 bytes free; 5.33 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

I tried reducing the --input-depth to see if this would reduce the memory usage, but it didn't help.

Are there any other possible tweaks that would allow for processing of larger images?