Closed cyrta closed 1 year ago
Some things have changed with Ray 2.0 https://docs.ray.io/en/latest/cluster/kubernetes/index.html
They introduce KubeRay Operator is the recommended way to do so.
Thanks for the input @cyrta
As we have no other demand for this, and multiple solutions for ML at OVHcloud, we do not plan to work on this. However, you are more than welcome to offer our Kubernetes user a documentation to deploy Ray on our managed services.
We have multiple partners that have done so like CloudCasa or https://docs.ovh.com/gb/en/kubernetes/backing-up-cluster-with-velero/ .
Ray.io on demand service
It is possible to use kubernetes + terraform https://docs.ray.io/en/latest/cluster/cloud.html#cluster-cloud
However, I believe it would be valueble to make deployment easier, by just one call or button click
Rferences:
See this blog post for a step by step guide to using the Ray Cluster Launcher.
To learn about deploying Ray on an existing Kubernetes cluster, refer to the guide here.
It is very easy to prepare massive, large dataset for AI model training https://speakerdeck.com/anyscale/ray-datasets-scalable-data-preprocessing-for-distributed-ml