fix: added ray in the minimal in the pyproject.toml#5979
fix: added ray in the minimal in the pyproject.toml#5979patelchaitany wants to merge 5 commits intofeast-dev:masterfrom
Conversation
Signed-off-by: Chaitany patel <patelchaitany93@gmail.com>
|
@patelchaitany Please run make lock-python-dependencies-all and commit the updated requirements files |
Signed-off-by: Chaitany patel <patelchaitany93@gmail.com>
franciscojavierarceo
left a comment
There was a problem hiding this comment.
I don't think Ray should be in minimal
|
@franciscojavierarceo, the issue is that feast-operator uses the minimal image, so using the Ray offline store requires us to maintain a custom Docker image. I believe adding Ray to the minimal build would solve this and improve the developer experience. Do you think that is a valid approach? |
@franciscojavierarceo We will need these compute engines in some image so that users do not need to build their custom images. We can either include in ray/spark in minimal or have new feast[server] group to build image, but having separate group means having 2 different images as well. Some slack context - https://feastopensource.slack.com/archives/C01M2GYP0UC/p1768568651039479 |
|
@ntkathole i think the custom image path is the right option or we deploy images for each support compute engine (e.g., Ray, Spark, etc.). |
What this PR does / why we need it:
This PR adds the ray in the minimal in the pyproject.toml
Which issue(s) this PR fixes:
Add Ray dependencies to Feast Operator and feature server images to support Ray Offline Store and Ray Compute Engine
Feast now supports Ray but upon configuring services with Ray getting runtime error
ModuleNotFoundError: No module named 'ray'
FeastModuleImportError: Could not import module 'feast.infra.offline_stores.contrib.ray_offline_store.ray'
Misc