A library of classes and functions for working with PyTorch.
Project description
firekit
Firekit is a library of classes and functions for training and evaluating PyTorch models. The main focus of the library is a Trainer
class that performs the standard training and evaluation loops, reports the training and evaluation loss and the evaluation performance on user-defined metrics, saves the model state when performance improves, and reloads the best model at the end of training. The Trainer
class also provides a simple interface to make predictions with the model on a new PyTorch Dataset
.
This project exists to support my work. It is in active development and the API is not stable. You are welcome to use it if it helps but you shouldn't rely on it.
Installation
Install with pip
or pipenv
in the normal way.
pip install firekit
Use the --extra-index-url
argument to install PyTorch for CUDA as a dependency. For example, use the following to get PyTorch with CUDA 11.3.
pip install firekit --extra-index-url https://download.pytorch.org/whl/cu113
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.