How to save a checkpoint after training a model for deployment ?

[sharabhshukla]

How to save a checkpoint after training a model for deployment ?

I am able to run the scripts on ray cluster with multiple worker nodes for training an RLLIb model. However. I would want to deploy the model and the model can be queried for optimal action, given a state as an input.

I tried using the save_checkoint method, however, I don't see any model being saved anywhere, I gave the file path the same as the script folder. However I also don't see any errors when running the script

[james-tn]

There's a default folder name ./outputs that is accessible everewhere in the cluster. Save the checkpoint to that.
Another option is to mount storage to the cluster. Check the Readme, there's guideline on how to mount a folder.