Can we run an experiment on a remote machine (without CML)?

[haimat]

Hello everyone, I have not used DVC so far, but it looks great so far from reading the docs and we want to give it a try. One question came up: We would like to work locally to write the ML training scripts etc., but execute the DVC experiments on a remote machine because of its GPUs. Do we need CML for that, or can we simply trigger a DVC experiment with a parameter or something to let it run on another machine?

[pmrowla]

This is not currently supported, the suggested way to run experiments remotely would be via CML

[haimat]

Thanks, I will have a closer look at CML then 👍

[courentin]

What about the remote executor feature, is it still "under active development"?

[pmrowla]

Unfortunately it is not currently being developed. Work on remote execution in DVC was tabled indefinitely in favor of focusing on CI scenarios (via CML) for the time being

[courentin]

Good to know thanks!