Benchmarking

The multievolve_hyperparameter_tuning.py script in notebooks/benchmark can be used to train the models for benchmarking MULTI-evolve.

"dataset_summary.csv" provides details on the individual datasets used for benchmarking. The table was derived and modified from ProteinGym.

Data

Large datasets are not included in this repository due to size constraints. Please download DMS dataset files from Zenodo (10.5281/zenodo.17620759) and place in:

data/benchmark/datasets/