[ENH] Ensure CPU-only PyTorch Installation by Updating pyproject.toml and Dockerfile

[sumana-2705]

Fixes #9

Description

This PR updates the pyproject.toml and Dockerfile to ensure that only the CPU version of PyTorch is installed, preventing unintended CUDA dependencies.

Changes Made

pyproject.toml

• Explicitly specified the CPU version of PyTorch.
• Adjusted dependency constraints to avoid GPU-related versions.

Dockerfile

• Added --index-strategy unsafe-best-match to uv sync to resolve dependency resolution issues.
• As torch is installed from trusted package sources (PyPI), I think it should be fine to use --index-strategy unsafe-best-match

How Has This Been Tested?

I have tested installation by developing docker image in my local machine, I have attached the screenshot of it

Checklist:

• My code follows OCF's coding style guidelines
• I have performed a self-review of my own code
• I have made corresponding changes to the documentation
• I have added tests that prove my fix is effective or that my feature works
• I have checked my code and corrected any misspellings

[sumana-2705]

Hello @peterdudfield

Please review the PR, and suggest me if any changes required.

[peterdudfield]

Looks great @dfulu you ok to check this?

[dfulu]

@sumana-2705 Thanks for these changes. About the testing you did via command line, you would get the same outputs

 torch.backends.mps.is_available()
>> False
 torch.cuda.is_available()

if you had a GPU version of torch installed but only had a CPU available on the machine you were testing this code.

So does the machine you tested this on have a GPU?

[sumana-2705]

Hello @dfulu

Earlier I checked it in a pc which does not have GPU but I have now tried it in the GPU supported pc, it worked correctly as shown in attached picture

[dfulu]

Thanks @sumana-2705