Early stopping is a form of regularization used to avoid overfitting on the training dataset. Early stopping keeps track of the validation loss, if the loss stops decreasing for several epochs in a row the training stops. The EarlyStopping class in early_stopping_pytorch/early_stopping.py is used to create an object to keep track of the validation loss while training a PyTorch model. It will save a checkpoint of the model each time the validation loss decrease. We set the patience argument in the EarlyStopping class to how many epochs we want to wait after the last time the validation loss improved before breaking the training loop. There is a simple example of how to use the EarlyStopping class in the MNIST_Early_Stopping_example notebook.
Underneath is a plot from the example notebook, which shows the last checkpoint made by the EarlyStopping object, right before the model started to overfit. It had patience set to 20.
git clone https://github.com/your_username/early-stopping-pytorch.git
cd early-stopping-pytorchRun the setup script to create a virtual environment and install all necessary dependencies.
./setup_dev_env.shActivate the virtual environment:
source dev-venv/bin/activateInstall the package locally in editable mode so you can use it immediately:
pip install -e .You can now import and use the package in your Python code:
from early_stopping_pytorch import EarlyStopping-
Clone the repository:
git clone https://github.com/your_username/early-stopping-pytorch.git -
Set up the environment:
./setup_dev_env.sh -
Activate the environment:
source dev-venv/bin/activate -
Install the package in editable mode:
pip install -e . -
Optional: Build the package for distribution:
./build.sh
The EarlyStopping class in early_stopping_pytorch/early_stopping.py is inspired by the ignite EarlyStopping class.