So much difference for different model

[whyydsforever]

Hello, I believe this work is a great contribution to the material science.
When I use a different pretrained model to draw the potential energy following the same procedure as you provided in embedding.ipynb, I got a totally different result.

The energy is only -1.4 eV. I am confused why could this happen, because I used the Gemnet all for OC20. it should get the similar result compared to the providing result.

[jkitchin]

It is hard to say what the reason is since you don't say what model you used. The different models are for different tasks, and sometimes give adsorption energy and sometimes give total energy.