RundialDocs

Appearance

Quickstart ​

Start Rundial and view a live chart for your first training run.

Prerequisites ​

1. Start Rundial ​

bash
git clone https://github.com/rundial-dev/rundial
cd rundial && bash scripts/setup/open_core_up.sh

On a fresh interactive installation, press Enter to accept the default auth mode. If the web port is occupied, choose an available port. Accept the default public S3 URL and admin email, then enter and confirm a password of at least eight characters. Accept the workspace and project defaults. The script creates the admin account and project. The first run builds images and can take several minutes.

If setup fails, see Troubleshooting.

When setup finishes, it prints the web app URL, normally http://localhost:5173, and the API URL, normally http://127.0.0.1:8787. Use the printed web app URL if the port differs.

2. Sign in and create a key ​

Open the web app and sign in with your admin email and password. Open Default Project, then Settings → Ingestion keys → Create ingestion key. Enter an optional label and select Create ingestion key. Copy the token when it appears; it is shown once.

3. Install the SDK ​

bash
pip install rundial
export RUNDIAL_API_KEY='<your-api-key>'

Replace <your-api-key> with the token you copied. The SDK uses the local API at http://127.0.0.1:8787 by default.

4. Log a run ​

Save this as train.py and run python train.py. It logs one point per second for 100 seconds:

python
import time
import rundial as rd

with rd.init(
    workspace="default-workspace",
    project="default-project",
    name="resnet50-lr3e-4",
) as run:
    for step in range(100):
        run.log({"loss": 1 / (step + 1)}, step=step)
        time.sleep(1)

Open Default Project → Train and select resnet50-lr3e-4. Open Metrics to watch the loss chart update.

Next steps ​

Stop Rundial ​

Keep Rundial running to follow the other guides. To stop it without removing its data:

bash
bash scripts/setup/open_core_down.sh

Run the setup script again to start it with the retained data. The --volumes option deletes that data.