Access to full dataset of 2.4 million materials?

[hzone3898]

Hi,
you write in the SI:

"In particular, we find that with common thresholds of 25 or 50 meV/atom from the convex hull, the GNoME dataset provides a substantially large number of crystal structures. Even at the more conservative 25 meV/atom metric, there are over 2.4M unique compositions and associated structures in the GNoME dataset."

would it be possible to obtain these materials, or are you planning to release only the stable ones?