mckinziebrandon/DeepChatModels

From a user/developer standpoint, this project offers a cleaner interface for tinkering with sequence-to-sequence models. The ideal result is a chatbot API with the readability of Keras, but with a degree of flexibility closer to TensorFlow. On the'client' side, playing with model parameters and running them is as easy as making a configuration (yaml) file, opening a python interpreter, and issuing a handful of commands. This is just one way to interface with the project. For example, the user can also pass in parameters via command-line args, which will be merged with any config files they specify as well (precedence given to command-line args if conflict). You can also pass in the location of a previously saved chatbot to resume training it or start a conversation.