ntadimeti

This will allow for us to encode only non-zero data for both training and inference. Also makes the data more compact and self-contained.

Create AWS AMI that contains pre-installed atacworks. NOTE: This will only be a snap-shot version. User is encouraged to clone repo on their own if they want to do developmental...

Some of the package dependencies are pinned to pretty old versions. It would be good to update to the latest stable versions. - Re-train the pretrained models and make sure...

This is preventing us from reading the config files that were output by atacworks.

Currently the generated bigwig files only have chr1-22,X,Y. Re-generate the bw files to contain all the chroms in hg19.chrom.sizes.

enable automated docker image upload with new merges to master

1

Currently, pushing a new image to dockerhub is a manual process. This prevents users from having a seamless experience of using the latest image at all times.

Currently resume only reads the weights from an existing model and starts training with them instead of randomly initialized weights. This is only partially correct. Ideally, we would save all...