Flag to enable use of GPU?

[lvora]

Does uniGRADICON CLI require a flag to use the system's GPU for out of box inference?

When I run the command provided on the home page (CLI) w/o IO (and also w IO); the gpu resources don't seem to used. It is still the CPU being used. Is there an explicit flag?

[lvora]

@HastingsGreer

[HastingsGreer]

Hi! It is supposed to detect the GPU and use it automatically.

@lvora

As a debugging step, could you

1. activate the virtual environment where you installed unigradicon
2. python
3. import torch
4. torch.tensor([123]).cuda()

and see if it throws an error?

[lvora]

Good point, thank you. I guess I will have to install a pyTorch version that does have CUDA support since I get an error.

torch.tensor([123]).cuda()
Traceback (most recent call last):
File "", line 1, in
File "...\torch\cuda_init_.py", line 310, in _lazy_init
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled