Version compatibility issue

I find myself in this situation:

System: Lenovo Thinkpad P16 - 32 GB Ram + NVidia RTX3500.

OS: Ubuntu Linux 24.04.3

`nvidia-smi
Fri Oct 10 11:49:15 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.163.01 Driver Version: 550.163.01 CUDA Version: 12.4

...`

I am using a python package that is kind of bleeding edge in development. It is an astrophotography package called SetAstroSuitePro. It has many functions that can be greatly enhanced with the GPU. Previous versions of SAS Pro were compatible with my machine and did well utilizing CUDA version 12.4.

Now though, the developers have introduced a new version which uses a later version of pytorch which must be installed. However, as I understand your webpage, version 12.4 is no longer supported in this pytorch version. 12.6 is the lowest version supported.

Would it be possible for me to install this package following the pattern of the commands supplied on this page:

pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu124

Or will that fail?

If that will not work, what is the best way to get v12.6 onto my system? Can it be done without upgrading the driver (which the Software Updates app will not allow due to version mismatch issues).

The posted command should work but will install the last PyTorch version built with CUDA 12.4, which is torch==2.6.0.

The PyTorch binaries ship with their own CUDA runtime dependencies and your locally installed CUDA toolkit won’t be used. Your Ada Lovelace GPU is supported in all of our builds and you would only need to properly install an NVIDIA driver (corresponding to the desired major CUDA version).

Okay thanks.

Properly install an NVIDIA driver. Driver 550.163.01 is the most recent that can be installed on the system using normal means (Software Updates app). Higher offered versions like 570 and 580 fail at installation due to unresolved dependencies. I guess I should try the command and see if the software works.

NVIDIA driver 550.x will work with our CUDA 12.x builds as CUDA’s minor version compatibility will be used in this case as described here.

OK, I’ve made no progress/

I tried the following commands:

pip3 install torch torchvision --index-url ``https://download.pytorch.org/whl/cu124

pip3 install torch torchvision --index-url ``https://download.pytorch.org/whl/cu129

even though 12.4 is my current version. These both installed torch 2.8

Following the lead of https://pytorch.org/get-started/previous-versions/ I got the command

./venv/bin/python -m pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --index-url ``https://download.pytorch.org/whl/cu124

None of these produced an installation that the application I was running was satisfied with.

I am not the developer of this application. I am an end-user. I need help with what I can do to get this application running.

Should I have been running pip3?