PyText is a deep-learning based NLP modeling framework built on PyTorch. PyText addresses the often-conflicting requirements between enabling rapid experimentation for NLP models and serving these models at scale. It achieves this by providing simple and extensible interfaces and abstractions for the different model components, and by using PyTorch to export models for inference via the optimized Caffe2 execution engine. PyText is used at Facebook to iterate quickly on new modeling ideas and then seamlessly deploy them at scale.
PyText allows researchers and engineers to use efficient distributed training to train custom models for different NLP tasks on their own datasets and then export them to Caffe2 via ONNX. Examples of the tasks that are currently supported in PyText include:
PyText is designed and built to be easily extensible to support new tasks and use-cases, and allows the reuse of prebuilt components from other tasks.
Install PyText pip install pytext-nlp
Learn how to write a model in PyText, train and evaluate it against a benchmark and deploy the model at scale in AWS
Follow an in-depth tutorial on how to create a new task and model in PyText
Review the PyText documentation to familiarize yourself with its different components and abstractions