Preparing training files
=========================

`python prepare_dataset <type> <list_of_files_and_references> <output_directory>`

Where:
Type is one of temp/comp/2d.
List_of_files_and_references is a file which contains names of read files followed
by reference sequence (one pair on each line), e.g.:

file1.fast5 ACCGGGTTAACCC
file2.fast5 CCCGGTTAAACCCGGGG
file3.fast5 AAATTGGCCTTGGAAA

The reference sequence should cover whole read sequence and maybe overextend a bit
(like 20 bases on each side).

Output directory must exist before calling the script.

Training
===============

`OMP_NUM_THREADS=1 python train.py <starting network> <list_of_training_files>`

Use starting network from ../nets_data/ (map6temp/map6comp/map6-2d-big)

Then this creates a directory a starts training. Each 20th iteration it outputs network into that
directory with name dump-%s.npz (where %s is ID of iteration).
