Firstly please unzip all *.7z to get *.txt raw data file.

Then run
python token_utils.py --input training_raw.txt --max_len 64 --split 10 --save_path .
python token_utils.py --input test_raw.txt --max_len 64 --split 1 --save_path .
to get preprocessed data.

More details of data can be found at: https://github.com/molecularsets/moses