The benchmark should use the same random seed to split train, valid and test data

[shenwanxiang]

As we know, the random seed in splitting the dataset has a large impact on the final performance of test set, So all of these dataset should be evaluate as same as the test set in deepchem/molnet, namely, you should split your dataset as deepchem does! not just by yourself

[wengong-jin]

This repo is deprecated, we did compare with molnet on EXACTLY the same train/valid/test split in our new repo here: https://github.com/swansonk14/chemprop#results