Summary
The user is trying to run a diabetes classification example but is facing issues with the split_traitest_dataset task failing due to the missing pandas module in the Flyte default image. They are asking if there are prebuilt images with pandas and other dependencies or if they need to create a custom image. Additionally, they mention encountering another error related to pulling a container, which they believe is connected to their M1 MacBook, referencing a potential exec format error. They also needed “registry=localhost:30000.” The user concludes that it works now.
david.espejo
oh sorry for the struggles, feel free to contribute to the example by pointing to these missing pieces
denis.shvetsov
Hi!
I’m trying out <https://docs.flyte.org/en/latest/flytesnacks/examples/pima_diabetes/diabetes.html|diabates classification example>, but with remote execution the split_traitest_dataset task fails because there is no pandas module
I think it is because flyte default image doesn’t contain this module. Is there any prebuilt images that I can use with pandas and other dependencies, or I have to build it myself?
denis.shvetsov
Also I needed “registry=localhost:30000”
It works now, thank you!
david.espejo
sorry, try with platform = "linux/arm64"
denis.shvetsov
Thank you, this doesn’t work me as well, now on a building step
ERROR: failed to solve: <http://ghcr.io/astral-sh/uv:0.2.37|ghcr.io/astral-sh/uv:0.2.37>: failed to resolve source metadata for <http://ghcr.io/astral-sh/uv:0.2.37|ghcr.io/astral-sh/uv:0.2.37>: no match for platform in manifest: not found
david.espejo
just in case you're getting the exec formaterror
david.espejo
if you're running it locally, please add platform=darwin/arm64 to your ImageSpec config
denis.shvetsov
Works, thank you!
Or at least I’ve got another error with pulling a container. but I think it is connected with M1 macbook
david.espejo
<@U0662JS8EV8> you should be able to use https://docs.flyte.org/en/latest/user_guide/customizing_dependencies/imagespec.html#image-spec-example|ImageSpec to declare the custom image with the required dependencies. In that way you don't have to write a Dockerfile