I've gotten this notice a few times over the last few months...

Python apps for GPU hosts needs 27668.09MB more disk space. You currently have 5710.52 MB available and it needs 33378.60 MB.

A little over 30Gig? That seems like an awful lot of disk to be using. I know "disk is cheap" and whatnot, but are the datasets truly that big? Anyone know why it needs so much space?

I've gotten this notice a few times over the last few months...

Python apps for GPU hosts needs 27668.09MB more disk space. You currently have 5710.52 MB available and it needs 33378.60 MB.

A little over 30Gig? That seems like an awful lot of disk to be using. I know "disk is cheap" and whatnot, but are the datasets truly that big? Anyone know why it needs so much space?

Because it DOES need that much space. The tasks do reinforcement learning using Pytorch and that framework infrastructure needs that much disk space for the examples.

On Ubuntu 20.04.4, my www.gpugrid.net project directory is 3.7GB, and the slot it is running in shows 9.1GB. So that is certainly less than 30GB.

But my SSD is 250GB, and I have 170GB remaining, so I have not hit the minimum space requirements yet.

(I don't know what the maximum it needs is; it could be a lot more.)

and the slot it is running in shows 9.1GB. So that is certainly less than 30GB.

Probably WUs may vary in their volume and/or occupied volume may grow during the execution of the task.
My current WU consumes almost 16 GiB(slot).

Although of course it was interesting to know exactly how the values of the required space are calculated.
Based on the fact that different values appear in such messages, this is some kind of automated algorithm, and not manually set values for WUs.

deleted my post - sorry wrong thread! :)