use onnxruntime-gpu for large-model-test

[wanliAlex]

use onnxruntime-gpu for large-model-test

[vicilliar]

i don't think tox envs were meant to use deps like this, why not instead do this:

• don't use pip install
• install deps with

deps =
  {[testenv]deps}
  -r{toxinidir}{/}temp{/}marqo{/}requirements.txt

• define onnxruntime-gpu==1.13.1 in requirements.txt

But if it's not possible to change this in the requirements file itself, your solution could be a good workaround :)

[wanliAlex]

Yes, the requirements.txt is using onnxruntime in general. We only need onnxruntime-gpu for the large model test. any good idea?

[vicilliar]

I see. note that the solution you put here will affect both largemodel_unit_test_CI.yml and unit_test_CI.yml. Have you tried running regular unit tests with this?

[wanliAlex]

i missed this point. we should use the onnxruntime in unittest as the instance does not have GPU. do you reckon creating a new test env will solve the problem?

[wanliAlex]

correctly uninstall onnxruntime and install onnxruntime-gpu