Install tensorflow-gcs-config from the python-tensorflow-whl - #788
mcollins42 merged 2 commits into
Conversation
Sorry, something went wrong.
…s-config from the wheel
rosbo
left a comment
There was a problem hiding this comment.
Can you add a test to ensure we don't have regression for the tensorflow-gcs-config package?
- You can add a new file at tests/test_tensorflow_gcs_config.py.
- Wait until ci built has built this image.
- Run docker pull gcr.io/kaggle-images/update-tensorflow-wheel-py37-3-staging
- Iterate until you have a good test that passes: ./test -i gcr.io/kaggle-images/update-tensorflow-wheel-py37-3-staging -p tests/test_tensorflow_gcs_config.py
Sorry, something went wrong.
|
I can add a test, but beyond checking if the module loads (this was failing with the binary distribution) and that a call to the main entry point doesn't fail, it's not clear how we would check whether the package is actually working in a test environment. Any thoughts? I'm not sure how far we've gone with some of the other packages. |
Sorry, something went wrong.
|
Checking that the module can load would be a good start. You could add it here: I don't know the tensorflow-gcs-config package much but here are some examples of advanced test:
I wouldn't go as far as running a GCS emulator for this simple test. |
Sorry, something went wrong.
|
I tried adding a test to import the module and call the main entry point configure_gcs, but sadly even that call immediately makes a request to the Google OAuth server to validate the credential passed in, so add it to test_imports.py seems to be the most we can do without some serious hacking. |
Sorry, something went wrong.
Update the version of python-tensorflow-whl used in the build and install tensorflow-gcs-config from the wheel.
http://b/152051681