Quick Start Guide
This guide will walk you through running the included quick start examples for easy-lightning. Two example scripts are provided:
quick_start_rec.py— A quick start using therecbackend.quick_start_torch.py— A quick start using the PyTorch backend.
Before You Begin
Make sure you have easy-lightning installed and have initialized your project:
pip install easy-lightning
easy-lightning-init
Running the rec Quick Start
The quick_start_rec.py script demonstrates a simple experiment using the rec backend.
To launch it, run:
python quick_start_rec.py
This will execute the experiment using the preconfigured settings in the script. Check the console output and generated logs for training progress and results.
Running the PyTorch Quick Start
The quick_start_torch.py script shows how to set up and run a quick experiment using PyTorch.
Run it with:
python quick_start_torch.py
Like the rec example, this will run with a minimal setup so you can quickly verify your environment and get results.
Experiment Directory Structure
The base directory for saving experiments is defined in the global configuration using the exp_name field.
This corresponds to a top-level folder that will contain all related experiment runs.
Within this folder, EasyLightning automatically organizes output files into the following structure:
out/log/— Contains all logs and metric outputs, including training and evaluation statistics.out/exp/— Stores the full YAML configuration files used for each experiment.out/models/— Contains the saved model weights (checkpoints) generated during training.
For more details on how to configure the experiment directory and related settings, see the Configuration page in the documentation.
Next Steps
After running these quick start scripts:
Review the code in each script to understand the configuration and workflow.
Modify the parameters to match your dataset or model.
Explore advanced usage in the full documentation.