We tested BERT on IS2RE task of Matbench Discovery.
We described the structural information in natural language and then performed a regression from 768 to 1 on BERT.
Requirments | version |
---|---|
torch | 2.5.1 |
lightning | 2.4.0 |
transformers | 4.46.3 |
pymatgen | 2024.11.13 |
- download bert-base-cased on HuggingFace Web
- download IS2RE datasets
- Run load_md_data.py to exchange Structures to the input_ids and attention_mask of bert
- Train and predict