added check to prevent CUDA driver activation if CUDA is not functional

[rssdev10]

This check is required when CUDA is imported somewhere but actual drivers are unavailable. Without that check, the package fails while attempting to CUDA libraries load.

[codecov-commenter]

Codecov Report

Base: 87.98% // Head: 87.98% // No change to project coverage 👍

Coverage data is based on head (0e3df1d) compared to base (f79d863).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #23   +/-   ##
=======================================
  Coverage   87.98%   87.98%           
=======================================
  Files           3        3           
  Lines         383      383           
=======================================
  Hits          337      337           
  Misses         46       46           

Impacted Files	Coverage Δ
src/ONNXRunTime.jl	100.00% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[jw3126]

Thanks a lot! Should we throw a warning if CUDA is not functional?

[rssdev10]

Actually, not sure that the warning message is required. The issue happened because it is hard to control what was loaded by 3-rd party components. We didn't use CUDA, but got that issues with drivers. CUDA.jl was included internally by Transformers.jl... The specifics of @require is unconditional activating of the code if the package was loaded somewhere before.