ElemwiseRetro

Implementation of ElemwiseRetro developed by Prof. Yousung Jung group at Seoul National University

(contact: [email protected])

( Predicting Synthesis Recipes of Inorganic Crystal Materials using Elementwise Template Formulation )

DOI : 10.1039/D3SC03538G

Developer

Seongmin Kim ([email protected])

Python Dependencies

• Python (version == 3.8.13)
• Numpy (version == 1.22.3)
• PyTorch (version == 1.11.0)
• Pymatgen (version == 2022.9.21)

How to use

First, take zip files from https://zenodo.org/record/8123145 (from Ceder's group, textmined inorganic synthesis dataset), put them in "./dataset" folder, and execute the below codes

---

Data.py ; Preprocessing the data

Train_P.py ; Target -> Precursors prdicting model training and Save (ElemwiseRetro)

Train_T.py ; Target + Precursors -> Temperature predicting model training and Save

baseline_Model.py ; Template popularity based baseline model and Save

Test_TP.py ; Load trained models and show the results