0.8.0

• As of this release Python 2 is no longer supported

0.7.2

• SSH Tunnel Workflow

0.7.1

• Sklearn Pipeline patch
• Changed default log level to INFO

0.7

• Thread-safe Keras Sequence dataset splits
• Additional Seq2Seq models
• Bastion tunneling support for SSH db connections
• Explicit modules for constants and imports
• Additional base classes for database connections (plain and alembic)
• Database independent sqlachemy types
• Switched pickle library from dill to cloudpickle
• SQLite support
• Changed default DB connection to SQLite

0.6

• Full database initialization with alembic
• DB schema validation on start
• Main configuration file for all credentials
• Drop official support for python 3.4
• Automatic handling of no data operations
• Remaining cloud provider support
• Feature metadata for classification models
• Runtime environment validation
• Add Split and SplitContainer objects
• Simplejson dependency
• Pipeline generator support
• Library specific model base classes
• Generalized database connection classes

0.5

• Default identity pipeline
• Alembic integration for database migration
• Standardized model inheritance pattern
• Condensed pandas split dataframes into single df
• Remaining classification metrics
• Updated schema with hash datatype
• Updated hash to use joblib code, consistent across initializations
• Generator pipeline and fitted kwarg
• Dropped base prefixes
• Moved composed subclasses to inits
• Unified datasets and pipelines

0.4

• Keras Seq2Seq support
• Keras model support
• Minimized required installation dependencies
• Abstract base classes
• Complex object JSON serialization
• Python 3 compatibility
• Travis and tox for CI/testing

0.3

• Save mixins for external files
• Keras sequential model support
• hdf5 binary saving mechanism
• numpy hashing support
• Some tests

0.2

Fully functioning alpha for basic classification tasks