Deep Pavlov

Deep Pavlov 1.0 is an open-source NLP framework built on Py Torch and transformers. Deep Pavlov 1.0 is created for modular and configuration-driven development of state-of-the-art NLP models and supports a wide range of NLP model applications.

The Deep Pavlov models are organized in separate configuration files under the config folder. A config file consists of five main sections: dataset reader, dataset iterator, chainer, train, and metadata. The dataset reader defines the dataset’s location and format. After loading, the data is split between the train, validation, and test sets according to the dataset iterator settings.

The chainer section of the configuration files consists of three subsections:

• the in and out sections define input and output to the chainer,
• the pipe section defines a pipeline of the required components to interact with the models,
• The metadata section describes the model requirements along with the model variables.

The transformer-based models consist of at least two major components:

• Preprocessor that encodes the input
• Transformer-based model itself.

The Deep Pavlov framework provides a scalable microservice architecture, enabling the integration and orchestration of diverse conversational skills. It supports incorporating default skills as well as custom skills developed by users, facilitating flexibility and ease of development.