- Add Multi-GPU support using Dask-cuDF
- Add support for reading datasets from S3, GCS and HDFS
- Add 11 new operators: ColumnSimilarity, Dropna, Filter, FillMedian, HashBucket, JoinGroupBy, JoinExternal, LambdaOp, NormalizeMinMax, TargetEncoding and DifferenceLag
- Add HugeCTR integration and an example ntoebook showing an end to end workflow
- Signicantly faster dataloaders featuring a unified backend between TensorFlow and PyTorch
- Switch to using the release version of cudf 0.14
- Fix PyTorch dataloader for compatability with deep learning examples
- Fix FillMissing operator with constant fill
- Fix missing yaml dependency on conda install
- Fix get_emb_sz off-by-one error
- Initial public release of NVTabular