BentoML docker image size

[mat-berna]

Hi,

I create a simple ML service with BentoML using the sample described on BentoML quickstart (simple IrisClassifer with a random forest algorithm).
I want to containerize the BentoML bundle.
Everything works correctly but I noticed that the size of the generated image is more or less 500-600 MB.
I think the image is bigger than expected.

Is it correct?
Is there a slim version image?
If it can help I can also provide all the source code.

Thanks.

[parano]

@Berna4 yes there's a slim image you can use, see documentation here: https://docs.bentoml.org/en/latest/concepts.html?highlight=slim#using-other-docker-base-images

Unlike the default debian based docker image, you will need to use your python version and bentoml version to find the right slim docker image to use, e.g.:

@env(docker_base_image="bentoml/model-server:0.12.0-slim-py37")
@artifacts([SklearnModelArtifact('model')])
class ExamplePredictionService(BentoService):
  ...```

[mat-berna]

Hi @parano, thank you for the answer.
I compare created images using slim or "normal" bentoml base image as you can see below:

Thank you for all.
Best regards.

Matteo