Description
Description
Following the guidance in this comment from an old closed issue, I successfully deployed a Ray cluster on Azure. However, the process feels quite hacky and is challenging to reproduce and share with my team. Specifically, I had to:
pip install ray
- Locate the template files stored locally
- Modify the templates in place
While this approach works on my machine, it is difficult to replicate and share these modifications with others.
Instead of using the CLI in this manner and manually hacking the template in the source code:
ray up example-full.yaml
It would be great if we could do something like this:
ray up example-full.yaml -t arm-template.json
This would allow us to replace the template at the command launch time, making the process more streamlined and reproducible.
If you think this feature is worthwhile, I'm willing to dedicate some time to it and potentially even make a pull request.
Use case
As a developer deploying Ray clusters on Azure, I need a more streamlined and reproducible way to manage and share deployment templates.
Activity