A PyTorch-based project for image classification experiments, featuring rapid model prototyping through dynamic configuration.
Create and activate the environment:
conda env create -f conda.yaml
conda activate image-classification
Ensure a virtual environment (Python 3.9) is activated:
pip install -r requirements.txt
This project uses Hydra for configuration managment. Modify main configuration file conf/config.yaml
for custom experiment settings.
From project root, execute the main script with optional configuration overrides:
python src/main.py training.n_epochs=5 experiment.use_gpu=False
Launch MLflow's UI for tracking experiments. Use optional --port
for a custom port (default access at http://localhost:5000
):
mlflow ui --port 8080