v0.5.2

Bug fix release

• Fix bug leading to dropping of pileup chunks during loading

v0.5.1

Development and performance release

• Refactor batch queuing in preparation to using keras Sequence
• Asynchrounous feature loading during inference
• Pin version of h5py to work around intermittent errors in saving models

v0.5.0

• Large refactor of training code
  • Resolve hanging at the end of training #15
  • Switch to CuDNN for GRU layers
  • Improved storage and retrieval of features for better IO
  • Training speed improved >10X
• Resolve issue with contained chunks during stitching #20
• Changes to medaka_consensus
  • checks presence of minimap2 and samtools
  • provides more feedback on error #16

v0.4.3

• Add support for R9.4.1 flip-flop basecaller

v0.4.1

Development release

• Adds build infrastructure for source distributions and manylinux wheels.

v0.4.0

• Large refactoring of feature and sample generation (#10)
  Fixes many small bugs and edge cases
• Resize models for small contigs (#9)
• Faster Generation of inference features
• Model updates
• Remove redundant samtools tview code
• Ability to handle multiple read types
• Limit CPU usage when running without a GPU

v0.4.0 Release Candidate 5

• Accelerate multi-datatype pileup creation.
• Fix some bugs in training.

v0.4.0 Release Candidate 4

• Fix bug in medaka stitch when --region is specified.

v0.4.0 Release Candidate 3

• Large refactoring of feature and sample generation
• Resize models for small contigs

v0.4.0 Release Candidate 2

• Fix import error after switching to local imports of keras/tensorflow