How to obtain the dataset for the first stage training of LLM

I'm from https://huggingface.co/datasets/U4R/GeoX-data The data and pretrain files have been downloaded from this website. May I ask which 100 million token dataset was used to train geoLLM in the first stage of the paper?