Home

TorchAssistant is a deep learning framework built on top of PyTorch. It provides a set of tools to automate the training and evaluation of models. It also reduces the amount of trivial code one usually needs to write.

Just create a specification file configuring the training session and let the framework do everything else for you.

Main features:

• scripts for training, evaluation, and inference
• automatically calculating metrics
• automatically saving training session
• resuming the interrupted training session
• automatically saving metrics history to CSV files
• nicely formatted information about the training session: epoch, iteration, loss, metrics, etc.
• highly flexible and customizable
• support for building complex training pipelines