Run mlflow project on multiple remote servers?

user_7034613 0
Viewed 221

Can MLflow be used to dispatch projects to multiple remote servers?(not aws,azure etc.) from a local tracking server?
I have the following scenario-
Multiple servers, where I would like to dispatch the mlflow project to all with different parameters, and let them "report" back to the current tracking server:

for ip in servers_ips:
    start_remote_mlflow(entry_point=GITHUBPATH,tracking_server=this_server_ip,hparams)

I see one can dispatch mlflow projects to aws or azure by specifying the ip or the remote machine. Can it be done with desktops as well?

mlflow remote-server python
1 Answers

No, MLflow is a centralized registry and tracking and is not responsible for distributred scheduling. You can use another frameworks and tools, like Dask or Luigi to accomplish this.

user_917140
Related