Update projected_gradient_descent_pytorch.py [Solve non-writable NumPy array and device mismatch issues]

[salomonhotegni]

Description

This pull request addresses two issues in the ProjectedGradientDescent attack of the Adversarial Robustness Toolbox (ART):

1. A UserWarning indicating that a given NumPy array is not writable, leading to potential undefined behavior when converting it to a PyTorch tensor.
2. A RuntimeError caused by tensors being on different devices (CPU and CUDA) during the _projection function.

The solution involves copying the non-writable NumPy array to ensure it is writable and modifying the tensor conversion in the _projection function to ensure all tensors are on the same device.

Fixes # (issue)

Type of change

Please check all relevant options.

• Improvement (non-breaking)
• Bug fix (non-breaking)
• New feature (non-breaking)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Testing

I have performed the following tests to verify the changes:

• Verified that the UserWarning about non-writable NumPy arrays no longer appears.
• Confirmed that the RuntimeError related to device mismatch is resolved.

Test Configuration:

• OS: Linux-5.15.0-91-generic-x86_64-with-glibc2.35
• Python version: 3.9.7
• ART version: 1.18.0
• PyTorch version: 1.13.1+cu117

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• My changes have been tested using both CPU and GPU devices

---

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.70%. Comparing base (a20c78f) to head (d79e663).

Additional details and impacted files

@@             Coverage Diff             @@
##           dev_1.18.1    #2455   +/-   ##
===========================================
  Coverage       85.69%   85.70%           
===========================================
  Files             329      329           
  Lines           30075    30075           
  Branches         5559     5559           
===========================================
+ Hits            25773    25775    +2     
+ Misses           2871     2870    -1     
+ Partials         1431     1430    -1     

Files	Coverage Δ
...ient_descent/projected_gradient_descent_pytorch.py	95.86% <100.00%> (ø)

... and 1 file with indirect coverage changes

[beat-buesser]

Hi @salomonhotegni Could you please fix the DCO check?

[beat-buesser]

Hi @salomonhotegni Could you please also make the same change in line 500?

[beat-buesser]

Please run Python formatter black with version black==21.12b0 using command black --line-length 120 --check art/ to fix the failing style checks.

[salomonhotegni]

Hi @salomonhotegni Could you please also make the same change in line 500?

Sure!

[beat-buesser]

Hi @salomonhotegni Thank you very much. To pass the DCO check, could you please add your signed-off-by line to every commit in this branch using the following commands:

• In your local branch, run: git rebase HEAD~2 --signoff
• Force push your changes to overwrite the branch: git push --force-with-lease origin my-feature-branch

[beat-buesser]

@salomonhotegni Could you please also take a look at updating the code to pass the style checks with the instructions above?

[beat-buesser]

Hi @salomonhotegni Could you please run command black --line-length 120 --check art/ with version black==21.12b0 and commit/push the changes to make the style check pass? Please let me know if you need help from me.

[beat-buesser]

Thank you for fixing DCO.

[beat-buesser]

Hi @salomonhotegni Thank you very much for your first contribution to ART!