We just open-sourced Egret-1, a new neural network potential for predicting molecular energies and forces at DFT accuracy — but several orders of magnitude faster.
It’s designed to cover a wide chemical space, from stable bio-organic molecules to challenging transition-state structures, and works out of the box with the ASE calculator interface.
We’re releasing three models:
Egret-1: general-purpose
Egret-1e: optimized for thermochemistry
Egret-1t: optimized for transition states
All models are MIT licensed and publicly available on GitHub.
If you’re doing anything with ML potentials, molecular dynamics, conformer generation, or quantum chemistry in general, would love to hear your thoughts or feedback.
We just open-sourced Egret-1, a new neural network potential for predicting molecular energies and forces at DFT accuracy — but several orders of magnitude faster.
It’s designed to cover a wide chemical space, from stable bio-organic molecules to challenging transition-state structures, and works out of the box with the ASE calculator interface.
We’re releasing three models:
Egret-1: general-purpose
Egret-1e: optimized for thermochemistry
Egret-1t: optimized for transition states
All models are MIT licensed and publicly available on GitHub.
GitHub: https://github.com/rowansci/egret-public
Use them on Rowan: http://rowansci.com
If you’re doing anything with ML potentials, molecular dynamics, conformer generation, or quantum chemistry in general, would love to hear your thoughts or feedback.