From vertex_ray in `google-cloud-aiplatform` package, it says that we have to use one of the following four base images to build our custom container for head/worker nodes.
https://github.com/googleapis/python-aiplatform/blob/main/google/cloud/aiplatform/vertex_ray/util/re...
We currently support Ray v2.9 and v2.33 and python v3.10.
The custom images must be extended from the following base images:
"{region}-docker.pkg.dev/vertex-ai/training/ray-cpu.2-9.py310:latest",
"{region}-docker.pkg.dev/vertex-ai/training/ray-gpu.2-9.py310:latest",
"{region}-docker.pkg.dev/vertex-ai/training/ray-cpu.2-33.py310:latest", or
"{region}-docker.pkg.dev/vertex-ai/training/ray-gpu.2-33.py310:latest". In
order to use custom images, need to specify both head and worker images.
However, this docker container is not publicly accessible. Could you let GCP team update the permission of these container images?
LinkedIn
Twitter
