PACMT v1.0.0: code, model reproducibility materials and evaluation data

This record provides the reproducibility materials for PACMT, a DNABERT-2-based framework for viral sequence identification and hierarchical taxonomic classification. The code archive contains training and prediction scripts, environment specifications, documentation and a runnable example. The data archives contain derived training, validation and independent-test partitions, taxonomy-label mappings, benchmark manifests, machine-readable supplementary tables, source data for main figures and external-evaluation materials. The individual ZIP archives share the same top-level PACMT-v1.0.0-data directory and should be extracted into the same destination folder to reconstruct the complete data package. The actively maintained source repository is available at https://github.com/luanbei/PACMT. Pretrained model weights and tokenizer files are available at https://huggingface.co/luanbei/PACMT.