An open-source neural network builder for creating deep neural networks (DNNs) by combining different predefined layers together. The builder currently can construct only DNNs for solving classification problems using supervised learning.
- Install CNTK 2.4.0 on your machine
- Create new class, that implements AbstractModel
- Override methods
Before
,GetNetworkDescriptor
,GetLayers
- Pass a path to a directory where you want to store network's checkpoints as a constructor parameter
You can find examples of different network models here