Created image not used and updated within ml.azure.com

[Katzmann1983]

Hello,

since about two weeks, we are trying to use our own custom image within this framework. We therefore updated the .pipelines/diabetes_regression-ci.yml but found out, that it does not influence the environment which is being used within ml.azure.com at all. Next we updated MLOpsPython/environment_setup/Dockerfile diabetes_regression/ci_dependencies.yml and noticed that the environment setup is only done once upon setup of the whole environment.

Is there a possibility to always use a new image created within the CI step? It seems kind of strange that within the CI step no image is produced.

[lokijota]

Can you check this issue and verify if the solution helps? https://github.com/microsoft/MLOpsPython/issues/382

[Katzmann1983]

Thanks @lokijota

We are actually using the environment from a Docker image. We therefore changed the environment to load a Docker image (including Parent image, username and password) within the UI. It is working nicely and also seems to automatically detect if the underlying Docker image has changed. On the other hand we were unable to setup a Docker connection correctly within the manage_environemnt.py to load a Docker image from an external repository.