ChengsenWang/TinyDNABERT-PretrainData-V1

Here is the preprocessed pretraining data for TinyDNABERT-20M-V1. During the preprocessing stage, we remove all “N” markers from the human reference genome GRCh38.p14 and randomly split the remaining sequences into segments ranging from 256 to 1024 base pairs, resulting in a total of 4,904,969 training samples.

For more details, please refer to the TinyDNABERT repository.