are pytorch and tensorflow really using the GPU?

[BrunoLiegiBastonLiegi]

Both the backends set the GPU as default device, if available:

qiboml/src/qiboml/backends/pytorch.py

Line 55 in 96f7c47

self.device = self.np.device("cuda:0" if torch.cuda.is_available() else "cpu")

https://github.com/qiboteam/qiboml/blob/96f7c471286192ef0e43295e811878c57d3acafe/src/qiboml/backends/tensorflow.py#L66C1-L70C46

however, if you try something as simple as:

from qibo import set_backend
from qibo.models import QFT

set_backend(backend="qiboml", platform="pytorch")
c = QFT(5)
r = c()
r.probabilities().device
# device(type='cpu')
r.state().device
# device(type='cpu')

you quickly notice that the GPU was probably not used (the same happens for tf).
At this point we might want to double check that the selected device is used in the computation.