Torch not compiled with CUDA enabled

[JP Ventura]

In [1]: import torch

In [2]: torch.cuda.is_available()
Out[2]: False

In [3]: torch.zeros(1).cuda()
---------------------------------------------------------------------------
AssertionError Traceback (most recent call last)
Cell In[3], line 1
----> 1 torch.zeros(1).cuda()

File /usr/lib/python3/dist-packages/torch/cuda/__init__.py:239, in _lazy_init()
    235 raise RuntimeError(
    236 "Cannot re-initialize CUDA in forked subprocess. To use CUDA with "
    237 "multiprocessing, you must use the 'spawn' start method")
    238 if not hasattr(torch._C, '_cuda_getDeviceCount'):
--> 239 raise AssertionError("Torch not compiled with CUDA enabled")
    240 if _cudart is None:
    241 raise AssertionError(
    242 "libcudart functions unavailable. It looks like you have a broken build?")

AssertionError: Torch not compiled with CUDA enabled

[JP Ventura (jpventura)]

Packages nvidia-cg-toolkit, nvidia-cuda-toolkit-doc, nvidia-cuda-toolkit-gcc, and libcudart12 were install from Canonical repositories. However the bug persists.

If I create a python or a conda environment, torch installed by pip work perfectly.