about 'cuda out of memory'

yijingru / BBAVectors-Oriented-Object-Detection

[WACV2021] Oriented Object Detection in Aerial Images with Box Boundary-Aware Vectors

MIT License

462 stars 87 forks source link

about 'cuda out of memory' #113

Open Amberrferr opened 2 years ago

Amberrferr commented 2 years ago

Hello! Wenn I run this programm, there is a RuntimeError: CUDA out of memory. Tried to allocate 182.00 MiB (GPU 0; 15.90 GiB total capacity; 14.60 GiB already allocated; 75.75 MiB free; 15.12 GiB reserved in total by PyTorch). I would like to ask, besides reducing Batch_size, is there any other way to solve this problem? Because when I reduce the Batch_size to 4, hm Loss is always nan.

yijingru commented 2 years ago

Hi, you might crop input images to smaller size

Amberrferr commented 2 years ago

Hi, you might crop input images to smaller size

Thanks for your answers! I tried to change the Input size to 400x400, gap=80, but on my computer, it can only run when batch size=4. (PS: the programm can run on batch size=6, when Input size is 600x600). The GPU I use is TU102 (Geforce RTX 2080 Ti, GPU memory 10GB) It's not working :(