Skip to main content
. 2024 Aug 12;22:756. doi: 10.1186/s12967-024-05567-z

Fig. 3.

Fig. 3

Multi-classification effects of sequences across different length ranges in the HERV dataset. (A) DNA_bert_6 model fine-tuning results; (B) human_gpt2-v1 model fine-tuning results. The top grey line is the percentage of sequences in different length ranges, the red line is the maximum number of tokens allowed by the model, and the RUN_Vote is the result of three independent runs of voting (RUN0, RUN1, RUN2)