haotianteng / Xron

Xron - an omni basecaller for ONT reads.
GNU General Public License v3.0
14 stars 3 forks source link

Problems with training model #4

Open fokxon opened 4 months ago

fokxon commented 4 months ago

Hi, I want to train my own xron model. But it seems that it requires a 'move' dataset in the fast5 files. Do you know how to generate it?

haotianteng commented 2 months ago

Do you know how to generate it?

The move dataset would be in the fast5 files generated by Xron (with --fast5 option).

I want to train my own xron model. But it seems that it requires a 'move' dataset in the fast5 files.

Fast5 files is not required to train Xron model, as long as you can prepare the following files you can train the model with xron train -i chunks.npy --seq seqs.npy --seq_len seq_lens.npy -o YOUR_MODEL_OUTPUT_PATH: chunk.npy : this is a numpy file contains N signal chunks, where each signal should be padded to the same length (default is 4000) seqs.npy: this file contains the N corresponding sequences of the signal in chunks.npy _seqlens.npy: this file contains the length (int) of N sequences.