ignite
High-level library to help with training and evaluating neural networks in PyTorch flexibly and transparently.
What is ignite?
High-level library to help with training and evaluating neural networks in PyTorch flexibly and transparently.
Overview of ignite
What it is
Ignite is a high-level library to help with training and evaluating neural networks in PyTorch flexibly and transparently. It provides an engine and event system, out-of-the-box metrics, and built-in handlers to compose training pipelines.
Key features
- Less code than pure PyTorch while ensuring maximum control and simplicity
- Library approach with no program's control inversion - use Ignite where and when needed
- Extensible API for metrics, experiment managers, and other components
- Extremely simple engine and event system with built-in events filtering and custom events
- Out-of-the-box metrics for various tasks: Precision, Recall, Accuracy, Confusion Matrix, IoU, ~20 regression metrics
- Built-in handlers to compose training pipeline, save artifacts, and log parameters and metrics
Use cases
Training and evaluating neural networks in PyTorch, with simplified training/validation loops, event-driven customizations, and easy metric computation.
Getting started
Install from pip:bashpip install pytorch-igniteFrom conda:bashconda install ignite -c pytorchFrom source:bashpip install git+https://github.com/pytorch/igniteFor nightly releases:bashpip install --pre pytorch-igniteSee the repository README for more details.
License and pricing
Licensed under BSD-3-Clause. Open source, no pricing information stated.
Based on the project's public README and repository metadata.