IndexError with bonito train about bonito HOT 3 CLOSED

Hello @Serka-M

Bonito does store model checkpoints after every epoch but unfortunately you have run into an issue just prior to that.

The traceback suggests that your converted references.npy file has more encoded bases in that it’s expecting.

Can you load the file and check for the highest index?

>>> import numpy as np
>>> r = np.load("path/to/references.npy")
>>> np.max(r)
4

The max index expected is 4 - where the indexes {0, 1, 2, 3, 4} map to the labels {N, A, C, G, T}.

Chris.

from bonito.

Hey Chris,

The converted Taiyaki dataset is for modified base detection, so it has one extra nucleotide letter.

Is there a possibility for a "quick fix" by editing some lines in Bonito code to make it all work?

from bonito.

Okay that explains it - the labels for the model are define in the config so the quick fix would be to add the extra label, this might not be the most effective way to handle modified base however.

from bonito.