Sample data fix

[Ellmen]

Thanks for the very interesting code.

As discussed in #1, the create_datasets.py and cfg.py don't work with the data provided. In particular, there is no solubility data, the data labels are mismatched, and the truncated datasets are missing values which cause issues when filtering datasets. This minimally updates those files to be able to run the sample code with:

cd data_processing
python create_datasets.py
cd ..
./run.sh

I also added Mac support on my fork if that would be of interest.

[albertdannie]

Hi, author! Your project is pretty charming! But I also have a question about the sample_pipeline.py and the static_eval.py. When I ran the sample_pipeline.py. It showed me the "AssertionError: need dumped states, run static_eval first". But after I ran the static_eval.py, it still doesn't work. I guess maybe there are some mistakes in the static_eval. Can you help me to address that?