Storage is currently taken care of by the AI4OS Nextcloud instance.
You have currently two ways of accessing your Nextcloud files from your deployment:
you can copy them locally to your deployment disk using rclone.
you can access them in the
/storagepath in your deployment. This is a virtual filesystem that lets you access the Nextcloud storage via the network.
For these options to work, you need to configure your RCLONE credentials when launching your deployment (under Storage options).
Reading files from disk (1) is faster than reading them via the network (2). But as deployments have a limited disk capacity (couple tens of GBs), sometimes you have no other option than going with (2) if you want to train with a really big dataset.
With option (2):
File read errors can happen if the network is slow.
If you see the following error message when trying to access
ls: reading directory '/storage/': Input/output error
this means you probably submitted the wrong RCLONE credentials during deployment. So please try to make a new deployment with the correct credentials.
To free some space in your deployment, always make sure to delete files in the Trash (
In parallel, we are working on expanding disk capacity in the deployments!