DNNs

A repository for easy training of Neural Networks.

Present Features:

• Ability to declare and train models with less than 5 lines of code.
• Define dropout rates, batch norms, and more such features easily.
• Logging enabled - researchers can log their metrics in a clean pandas file without having to write any regex expressions and such.
• Use along with a transformer: A transformer (more precisely data modifier), can be anything from a pre-trained model to a data preparation pipeline. Additional support for pytorch models is added.
• Create ensembles with ease through ensemble.py