Author: Amir Farbin

What is DLKit?

A library that makes deep learning easy...

Often it is difficult to wire up all the things necessary to start training with Deep Neural Networks on large datasets. In order to combat this barrier to entry, libraries such as Keras and TensorFlow, Theano, Scikit-Learn, etc., are used to do the heavy lifting for you. However, it seems that these libraries all leave out important pieces that really unlock the potential of your hardware. There is a demand for a framework that combines all of these great tools together and make it easier than ever to dive into scalable Deep Learning without the hassle of looking under the hood.

Components and Features

  • Model Persistence
  • Multi-Process Data Generators
  • Loss and Accuracy Libraries
  • Base DNN Models
  • Check pointing & Early Exiting
  • Sparse Tensors!
  • And More!

Data Preparation