Thermodynamic modelling in Python: simple access and fast performance

Thermodynamic modelling in Python: simple access and fast performance Authors: Paul Asimow, Mathew Gleason, Penny Wieser, and Ben Thyer We present two complementary aspects to our recent work expanding the ease of access and overall performance of thermodynamic modeling in Python. We will present PetThermoTools, an integrated and intuitive Python interface for thermodynamic calculations using either the alphaMELTS or MAGEMin packages. Well-documented functions and installation instructions allow PetThermoTools to be used by users with limited coding experience, while also providing performance advantages for more experienced users through the use of parallel processing. Nevertheless, the speed at which thermodynamic calculations can be performed remains limited and further advances in computational efficiency require different approaches. As such, we will introduce nMELTS, a neural network powered MELTS and HeFESTO emulator that offers a ~1000X speedup on GPUs compared to traditional thermodynamic modeling.