[0.0.2] (10 Nov 2022)

Release of version 0.0.2 is a minor release which introduces several performance and usability improvements to Dorado. In particular, we are happy to announce the inclusion of multi-GPU base modification calling and 5mC models.

Major changes

• 1636153 - Major upgrade to modified base calling speed.
• fea9181 61a7bc0 - Improve convolutional layer performance for increase in basecalling speed.
• fb7cb24 - Reduction in memory requirement for M1 basecalling, allows running on larger number of M1 devices.
• f5371d1 - Add signal move tables to SAM output.
• 621cc4d - Fix CPU basecalling
• 9c4bf71 - Add dorado namespace
• acbca36 - Improvements to Dorado portability - Dorado archive now bundles more dependencies.

Minor changes

• 16cb5cb - Fix downloading models in Windows
• 3202db8 - Basecaller node timeout, enables computation of non-full batches after timeout elapses.
• 1a40836 - Add duration tag to SAM.
• 0b3a336 - Add verbose flag to Dorado.
• d75303d - Solve segfault-causing race condition
• ac0a4d5 - Adjust overlap and chunk size to be a multiple of the model stride
• 626137a - Improve the way we look up GPU core count on Apple
• b2864db - Added spdlog logging framework
• e9c7866 - Reduced Startup cost on multi-GPU systems
• fe71021 - Migrate archive distributions from Box to CDN
• 4b67720 - Resolved issue with SSL verification for model downloads.