This release contains the following changes:
- Significant (2-3x) speed improvement for scanning alignment files.
- Lifted the MD tag requirement for input bam files.
- More program options added for scan_alignment (e.g. filter duplicate reads,...)
- An option is added for normalizing MQ, BQ, and other alignment channels for input candidate matrices
- New pre-trained models based on SEQC-II dataset are provided
- Update to Pytorch 1.1.0, torchvision 0.3.0, and cudatoolkit 9.0
- GPU training speed improvement by 25%.
- Minor fixes