-
Notifications
You must be signed in to change notification settings - Fork 459
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Client-side computation of cached steps #3068
Conversation
06341ec
to
661869f
Compare
E2E template updates in |
Classification template updates in |
…zenml-io/zenml into feature/PRD-657-precompute-cached-steps
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
🦭
docs/book/user-guide/starter-guide/cache-previous-executions.md
Outdated
Show resolved
Hide resolved
Co-authored-by: Alexej Penner <thealexejpenner@gmail.com>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Left a few comments, but nothing critical - well done!
data.pop("name", None) | ||
data.pop("version", None) | ||
data.pop("model", None) | ||
return data |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It looks like older pipelines might fail silently downstream because of this.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In which case would they fail?
I'm guessing this would only be possible if upgrading the server to the new release while still having an old scheduled pipeline running that uses a Docker image of an older version? The server would then remove those attributes before sending the response.
Would that case be fixed if I just don't remove this data here and add extra=allow
in the pydantic config?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
To me, it would be better just to give some instructions on how to fix this and write that deprecation clearly in the new release notes. I don't ask for any technical solution here - these guys should have been gone for a while already.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The problem here is that these attributes will already be removed on the server side, which means the warning/instructions will be very hidden. I instead added some instructions on the runtime error below that happens if the ID is missing.
Co-authored-by: Andrei Vishniakov <31008759+avishniakov@users.noreply.github.com>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
🦭
Describe changes
This PR adds functionality to compute cached steps client-side. This reduces the time/money spent waiting for remote orchestrators, which can often take a long time to spin up a step container only for it to immediately shut down because the step was cached.
Some things were removed as part of this PR to make the model linking more consistent for cached and non-cached steps:
ArtifactConfig
Pre-requisites
Please ensure you have done the following:
develop
and the open PR is targetingdevelop
. If your branch wasn't based on develop read Contribution guide on rebasing branch to develop.Types of changes