RuntimeError: CUDA out of memory. Tried to allocate 32.00 MIB | view GPU memory

RuntimeError: CUDA out of memory. Tried to allocate 32.00 MiB (GPU 0; 6.00 GiB total capacity; 4.40 GiB already allocated; 0 bytes free; 4.43 GiB reserved in total by PyTorch)

Traceback (most recent call last):
  File "D:/Papers to read/2022.07/DeamNet-main/DeamNet-main/train.py", line 170, in <module>
    train(epoch)
  File "D:/Papers to read/2022.07/DeamNet-main/DeamNet-main/train.py", line 73, in train
    prediction = model(input)
................
 File "D:\ProgramData\Anaconda3\envs\python36\lib\site-packages\torch\nn\functional.py", line 1136, in relu
    result = torch.relu(input)
RuntimeError: CUDA out of memory. Tried to allocate 32.00 MiB (GPU 0; 6.00 GiB total capacity; 4.40 GiB already allocated; 0 bytes free; 4.43 GiB reserved in total by PyTorch)

Process finished with exit code 1

Method found

1. Check the computer GPU Memory stay cmd The input ：

nvidia-smi.exe -h

  And then you type in

nvidia-smi.exe -i 0

You can see the running memory of the computer ,

  About the found call GPU Methods , The source code contains , So it's not the problem

os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = "0,1"
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")  # modify ： Add this trip .  Judge CUDA Can I use , Can not be used CPU

2. Another way is to modify batchsize

batchsize Turned out to be 8, Change it to 1, It's ready to run

parser.add_argument('--batchSize', type=int, default=1, help='training batch size')