potiuk commented on code in PR #53182:
URL: https://github.com/apache/airflow/pull/53182#discussion_r2201084800
##########
providers/google/src/airflow/providers/google/cloud/operators/vertex_ai/ray.py:
##########
@@ -136,9 +138,9 @@ class CreateRayClusterOperator(RayBaseOperator):
def __init__(
self,
+ python_version: str,
+ ray_version: Literal["2.9.3", "2.33", "2.42"],
Review Comment:
But maybe I am guessing what **really** the problem you have is wrong.. If
your problem is the "list" itself - that sounds like property of the vertexai
service and unless we can get a constant from vertex ai, I guess we have no
other choice but to have the hard-coded list.
Unless you have other idea what to do here? What exactly is your proposal
(other than linking to docs from the other comment). I searched for docs and
there is no mentioning of the supported ray versions - so link to github is the
best for now.
Or do you propose not have the list at all and let the users find out that
they put the wrong version after the ray job is already submitted ? We
generally folow the pattern tha we do validation as early as possible -
additionally this is just MyPy type check - which means that's it mostly a hint
that you can ignore.
So I think - when I think of the options we have - the current proposal
where we link to available information, have an explanatory comment about where
it comes from, where there is a mypy hint that does not limit - but only
"warns" and only when you run your DAGs through MyPY is likely the best
approach I can think of.
What woudl be a better proposal ?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
