[Feat]: Better instructions for CLI usage

[pinballelectronica]

Describe your use-case.

Hi. I love OT but I'd like to run it in Cloud for much larger datasets. I know there's a CLI but I'm having a hell of a time getting it working. I'm trying to get it running on Runpod w/ Conda Py 3.10.9. I install all requirements and copy all older files over from a UI run to continue it, edit the json to repoint paths to the new runpod /workspace structure. I stage the files in the same directory structure- tried to run create_train_files but was getting a Module not found error. I did get it to run once but it's complained about sd 1.5 not being present although my json file was pointing directly to a local sdxl base model.

Has anyone gotten this working satisfactorily on a cloud GPU provider? If you have, could you share the special sauce?
Thanks!

What would you like to see as a solution?

A shiny new update to the CLI usage.

Have you considered alternatives? List them here.

I guess I could hack together an X server and push it through webrtc but that's a bit of work.
Or maybe X11 forwarding via SSH tunnel but I'm guessing that won't work with Runpod as they are like pseudo docker containers.

[Arcitec]

My recent overhaul of the CLI and launcher scripts takes care of that: #477

It's been merged, so you can start using it now. The instructions are here:

https://github.com/Nerogar/OneTrainer/blob/master/LAUNCH-SCRIPTS.md

Look at the ./run-cmd.sh and its usage instructions to see how to launch CLI training or any of the other CLI tools. :)

There's also someone working on RunPod cloud support within OneTrainer, which will make RunPod itself much easier to use. I described that in another ticket, in a comment chain starting here:

#488 (comment)

Hope this helps! :)

[O-J1]

Issue has been solved, please reopen if it’s not sufficent