WIP: Pytorch version updated to 2.5.1

[tanwarsh]

Pytorch version updated to 2.5.1.

As we want to update PyTorch to version 2.5.1, we are proceeding with this PR along with a patch to fix the issue:
Type parameter +RV without a default follows type parameter with a default

[kta-intel]

Have we done any further investigation into why this redirection is necessary (i.e. why/when is pip._vendor.typing_extensions being called to enforce parameter ordering as opposed to typing_extensions)?

I know this was our originally proposed workaround, and I get that it might still be the only solution, but since we already needed to patch distutil in order run torch==2.4.x, I am very hesitant to introduce additional patches to get it working further.

My reason being that it is better for long term maintainability that we find a robust solution and avoid patching if we can.

[tanwarsh]

Thank you @kta-intel for the review. closing this PR for now.

[theakshaypant]

Please take a look at the pipelines, they have the parameter error.

[tanwarsh]

Please take a look at the pipelines, they have the parameter error.

Thank you for the review but this is still WIP.

[tanwarsh]

@teoparvanov, @kta-intel, @rahulga1.

After investigating this issue in detail, the only solution other than a patch is to fix this in the typing_extensions library by adding a more robust way to detect NoDefault from different modules. I will raise an issue in typing_extensions regarding this.