This version brings a number of major improvements, including batch training and compatibility with CUDA-enabled GPUs through RAPIDS. The naming of functions and variables has been changed and is now closer to PEP8 standards. This means the new API is not compatible with v1.
Make sure to read the README and the API documentation before moving to this new release.
A general code clean-up and a few bug fixes have been added.
Please contact me through Issues and pull requests if you are encountering any issues.