## Describe the bug Running `ot.solve` with tensors on the CPU allows memory on the GPU (thisis documented in `get_backend_list`) but also seems to use the GPU, as the Watts are increasing. See attached screencast : [Screencast from 08-03-2024 11:24:44.webm](https://github.com/PythonOT/POT/assets/8993218/ac6ba4b6-8529-436a-987d-9c0a661fee4e) Is it normal? Script ```python import torch import ot n_samples = 5_000 x = torch.randn(n_samples, 2) y = torch.randn(n_samples, 2) a = torch.rand(n_samples) a /= a.sum() b = torch.rand(n_samples) b /= b.sum() M = ot.dist(x, y) res = ot.solve(M, a, b, reg=0.1, reg_type="entropy") ```