Remote dataset storage

Is it possible to work with remote storage, which contains all the dataset corpus, instead
of storing it locally on the computing cluster?

Should it be linked with a mounted disk that points to a network drive? SSH?

Hey, we solely encourage peoples to work with local datasets as networking might become a bottleneck. To overcome this issue, increasing the num_workers usually is sufficient, unless your networking is very bad. As long as your path is mounted on your compute node, you can do it.