BentoML

BentoML currently has no strong opinions on how the project is structured. As long as a `bentofile.yaml` is present, BentoML can serve and build bento. Standard repos like Git or AzureRepos integrates well with Azure Pipelines. If you structure your repository with a `bentofile.yaml` at the root, I expect everything to just work. :slightly_smiling_face:

But How bentoml build &amp; BentoML containerize would work?
Azure pipelines doesn’t have provision for this or it would create docker image on its own based on bentofile.yaml

Azure Pipelines requires dockerfile to create dockerimage and push it to registry.

We do the following to build a docker image:
1. `bentoml export iris_classifier:latest saved/iris_classifier.tar.gz`
2. `tar -xf saved/iris_classifier.tar.gz -C saved/`
3. `docker build -t app_name:version -f saved/env/docker/Dockerfile --build-arg saved`
4. `docker push app_name:version`
The `--build-arg` in step 3 is to define build context because the Dockerfile is not at the top level in the saved folder.

I hope that helps!

Do you have any doc for Azure Pipeline integration?

I don't. We use Google Cloud; Kubernetes + Gitlab CI/CD.