Improvements * Added support for custom event shape for logistic head. * Added head for image outputs.
1.3.1
Change Log Minor performance improvements.
Improvements * Removed casting to tensor when building dataset. * Added mean absolute error metric for regression.
Bug Fixes * Removed reinitializing gaussian mixture layer on call.
1.3.0
Change Log Improved support for image and video processing.
Improvements * Added support for image and video predictors. * Better support for arbitrary regression shaped output.
Bug Fixes * Resolved issue of tensor inputs not feeding correctly with time series tasks.
1.2.0
Change Log Improved data pipeline. Removed generator and implemented records dataset for faster processing from graph mode.
Improvements * Speed up data processing and batching pipeline. * Removed TunerContainer class for easier supplying of tuner and parameters to model.train() method. * Added more samples.
Bug Fixes * Fixed issue with weighted loss not properly calculated for density estimation.
1.1.0
Change Log Added support for tensor input and outputs.
Improvements * Added support for vector, matrix, and arbitrary tensor inputs and response. * Speed up vocabulary adapt for large inputs.
1.0.2
Change Log Speed up and simplify processing of time-series data. Simplified naming convention for predictors to match convention in models and heads.