PyTorch does not using GPU on full power.
model and tensors are in GPU and PyTorch use GPU
print('The train is continue!') print(torch.cuda.get_device_name(0)) print('Memory Usage:') print('Allocated:', round(torch.cuda.memory_allocated(0) / 1024 ** 2, 1), 'MB') print('Cached: ', round(torch.cuda.memory_reserved(0) / 1024 ** 2, 1), 'MB') double_barrier_net = double_barrier_net.double().to(get_device())
This is the console logs
The train is continue!
NVIDIA GeForce RTX 2070 Super
Allocated: 14.5 MB
Cached: 22.0 MB
but training is going too slow and on my time_waste_test we could see that train on GPU even slowly than on CPU.
link to logs
I have no idea how to fix this