Devcontainer

[bhack]

Can we have a .devcontainer like in Keras?

https://github.com/keras-team/keras/tree/master/.devcontainer

[LukeWood]

Would this be useful to you personally?

If so, I can't see why not. Happy to do this to support your workflow. As long as we have someone using it and maintain it.

[bhack]

Can you double check this with the internally board before I start?

Honestly I had a not so good experience with another PR in this repo:
#11 (comment)
#14 (comment)

[LukeWood]

Understandable and still sorry about that. Cheers - I will double check tomorrow for you.

Does .devcontainer impact anyone else's workflow or only those who opt in to using them?

Is it used with VSCode?

[LukeWood]

Hey @bhack yeah happy to accept this. Just checked

[LukeWood]

Lets mirror keras-tuner's, but I don't think we want the linter as a pre-commit hook from what we have generally discussed as a team.

So our setup.sh will just have:

pip install -e ".[tests]"
pip install -e ".[examples]"

Cheers - if anyone wants to pick this up happy to accept

[bhack]

Do we want to depend on latest or nightly Tensorflow images?

https://github.com/keras-team/keras-tuner/blob/master/.devcontainer/Dockerfile#L1

See also out thread at:
#195 (comment)

[LukeWood]

we can use nightly for now, until the next tensorflow version comes out which will have BaseImageAugmentationLayer in the keras namespace.

[bhack]

we can use nightly for now, until the next tensorflow version comes out which will have BaseImageAugmentationLayer in the keras namespace.

Do you have a minimal date for nightly?

[LukeWood]

Do you have a minimal date for nightly?

minimal in what sense? A date when we can move to normal tf? I believe this can be done the release after 2.9, so 2.10. after 2.9 lands we can try changing the dependency to tensorflow from tf-nightly and see if it works.