RuntimeError: CUDA error: out of memory

[peng666]

I train the network through the following command：
python train.py --gpu 0,,1 --workdir log --model msn

When I execute the command, the program stops as follows：

THCudaCheck FAIL file=/opt/conda/conda-bld/pytorch_1616554800319/work/aten/src/THC/THCCachingHostAllocator.cpp line=278 error=2 : out of memory
Traceback (most recent call last):
File "train.py", line 83, in
main()
File "train.py", line 79, in main
model.runner()
File "/home/pjw/projects5/SpareNet/runners/base_runner.py", line 338, in runner
self.val()
File "/home/pjw/projects5/SpareNet/runners/base_runner.py", line 209, in val
self.val_step(items)
File "/home/pjw/projects5/SpareNet/runners/msn_runner.py", line 54, in val_step
, (, _, _, data) = items
File "/home/pjw/projects5/SpareNet/utils/misc.py", line 18, in var_or_cuda
x = x.cuda(non_blocking=True)
RuntimeError: CUDA error: out of memory

[peng666]

The problem has been solved, thank you

[HackHarry]

The problem has been solved, thank you

hello, I have this problem too. How do you solve this?