Below are step by step guides on understanding the concepts and how-to tips, and best practices of PyTorch-Ignite.
Install PyTorch-Ignite from pip, conda, source or use pre-built docker images
In this guide, we will show how PyTorch code components can be converted into compact and flexible PyTorch-Ignite code.
Learn how to get the time breakdown for individual epochs during training, individual events, all handlers corresponding to an event, individual handlers, data loading and data processing using Engine’s State, BasicTimeProfiler and HandlersTimeProfiler.
This how-to guide demonstrates how we can leverage the
FastaiLRFinder handler to find an optimal learning rate to train our model on. We will compare the results produced with and without using the handler for better understanding.
To effectively increase the batch size on limited GPU resources, follow this simple best practice.
When the data provider for training or validation is an iterator (infinite or finite with known or unknown size), here are some basic examples of how to setup trainer or evaluator.
This how-to guide demonstrates how we can do Cross Validation using the k-fold technique with PyTorch-Ignite and save the best results.
Cross Validation is useful for tuning model parameters or when the available data is insufficient to properly test
Learn how to create custom events that depend on the loss calculated, backward pass, optimization step, etc.
Example on how to switch data during training after some number of iterations