Running vgg16 on birds data

I am trying to train vgg-16 model on birds-200 data. the input is size [3, 448, 448] and there are 200 classes. My gpu is empty but when i try to run the code, i get the following error.

RuntimeError: CUDA out of memory. Tried to allocate 392.00 MiB (GPU 0; 7.93 GiB total capacity; 7.28 GiB already allocated; 111.25 MiB free; 21.83 MiB cached)

I am currently using batch size 32 an decreasing the batch size did not help. If i change the data to cifar and use vgg16, the code runs without any problem. Any idea what can be wrong?

Could you try to lower the batch size even further or resize the spatial size of your input images to e.g. 224x224?