(Ah, didn’t see this is in CICD, this makes it less likely the folder is large.) The contents of the root folder are copied into a tarball and uploaded to blob storage. I have seen in the past that e.g. the

.git

folder is large or that ML engineers stored data within the python package the workflow belongs to. This data is then uploaded as part of the tarball everytime during fast-registration.

Yep. And that is huge, as it is the github runner homedir that checks out the whole repo which has other other ML model code that we don't want to package.

How do I tell it to not to that? 🙂