Difference between benchmark.py and model.benchmark

[Mlzzzzz]

Hi MiCaM team,

I noticed that your source code includes a file called benchmark.py, where the distribution learning benchmark by GuacaMol is utilized. However, in train.py, the model is benchmarked using the model.benchmark(train_path=paths.train_path) method, which seems to take longer than the approach in benchmark.py. Could you explain the main differences between these two benchmarking methods?

Thank you!