v0.5.3

Highlights

• c0f83f3 POD5 support - https://github.com/nanoporetech/pod5-file-format
• 27f8468 Kit 14 models dna_r10.4.1_e8.2_{fast, hac, sup}@v3.5.1 added.
• e6282a2 Upgrade to remora 1.1 (kit 14 mod bases).

Bugfixes

• 4585b74 fix for handling stitching of short reads (read < chunksize).
• 9a4f98a fix for overly confidant qscores in repeat regions.
• 3187198 scaling protection for short reads.

Misc

• d57a658 training validation times improved.

Installation

The default version on PyTorch in PyPI supports Volta and below (SM70) and can be installed like so -

$ pip install ont-bonito

For newer GPUs (Turing, Ampere) please use -

$ pip install -f https://download.pytorch.org/whl/torch_stable.html ont-bonito-cuda113