CHANGELOG.md

Taiyaki

Version numbers: major.minor.patch

• Major version bump indicates a substantial change, e.g. file formats or removal of functionality.
• Minor version bump indicates a change in functionality that may affect users.
• Patch version bump indicates bug-fixes or minor improvements not expected to affect users.

v5.3.0

• Based on pytorch version 1.5
• Acceleration loss function lead to signifcant performance improvement
• Many bug fixes

v5.0.0

• Based on pytorch version 1.2
• Improved training stability: gradient capping and warm-up
• Merged mod-base and canonical entry points
  • Custom model definitions should now take an alphabet_info argument rather than outsize
• Improved RNA support: tools can reverse references and basecalls
• Basecaller changes:
  • chunk size argument now matches guppy
  • CPU calling enabled
  • lower memory usage
• Multi-GPU training enabled
• Bug fixes

v4.1.0

• Ab initio ("bootstrap") training of models

v4.0.0

• Modified base training and basecalling
• Minor changes to input format to trainer, use misc/upgrade_mapped_signal.py to upgrade old data

v3.1.0

• Basecaller script that uses GPU
• Training walk-through
• Tweaks to optimisation parameters

v3.0.2

• Improved training parameters
• Use orthonormal initialisation of starting weights

v3.0.1

• Bug fix: package version did not work in github's source releases

v3.0.0

Initial release:

• Prepare data for training basecallers by remapping signal to reference sequence
• Train neural networks for flip-flop basecalling and squiggle prediction
• Export basecaller models for use in Guppy