Customizing training data for fine-tuning the model

[sumin5784]

Hello,

I'm trying to use my own dataset to fine-tune DNABERT-5/6.
I have several questions about this.

1. All input sequence length should be the same? Or input sequences can be different length?
2. Does label should be 1/0? Basically, does classification class can be more than two classes?

Any feedback would be appreciated.
Thank you!

[Peng_Lee]

Excuse me, do you understand these two questions？