Memory Leak? CUDA memory finishes quickly when using more than 50 pre-annotated data

[brr53]

Even when using 150 pre annotated images to label 10 additional images, CUDA memory finishes on a 3090 GPU.

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 31.36 GiB (GPU 0; 24.00 GiB total capacity; 49.21 GiB already allocated; 0 bytes free; 49.35 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

It stops working after about 50 pre annotated images due to this error.