Docker Integration 🐳

For maximum reproducibility and ease of deployment, Fishy Business can be run entirely within a Docker container. This ensures that all dependencies (including complex C-libraries for spectral analysis) are correctly configured.

Building the Image

From the root of the project directory, build the Docker image:

docker build -t fishy-business .

Running the Dashboard

The default behavior of the container is to launch the Streamlit dashboard on port 8501.

Basic Run

docker run -p 8501:8501 fishy-business

With Persistent Data

To ensure your training results and datasets persist outside the container, mount your local data and outputs directories:

docker run -p 8501:8501
  -v $(pwd)/data:/app/data
  -v $(pwd)/outputs:/app/outputs
  fishy-business

Using the CLI via Docker

You can use the container to run training experiments or manage data without installing Python locally.

Download Data

docker run -it -v $(pwd)/data:/app/data fishy-business
  fishy download-data --token <YOUR_GITHUB_TOKEN>

Run a Training Experiment

docker run -it
  -v $(pwd)/data:/app/data
  -v $(pwd)/outputs:/app/outputs
  fishy-business
  fishy train -m transformer -d species --epochs 50

Weights & Biases Integration

To log results to W&B from within Docker, pass your API key as an environment variable:

docker run -p 8501:8501
  -e WANDB_API_KEY=<your_key>
  -v $(pwd)/outputs:/app/outputs
  fishy-business

Summary

Using Docker is highly recommended for: * Server Deployment: Hosting the dashboard on a remote research server. * Reproducibility: Ensuring your thesis results aren’t affected by local library updates. * Quick Setup: Skipping the installation of complex dependencies like XGBoost.