Getting your data in

The portal visualizes data that your training code logs to Metrana. Before runs and metrics appear here, your code needs to send them using the Metrana client library.

Note

This page is about getting data into the portal. For the full logging API (installing the client, init, log, the reinforcement learning helpers, and close), follow the Quickstart — open it from the button in the portal header, or from the Quickstart link in the Workspaces hub.

The short version

1. Get an API key. In your workspace, open Settings → API keys and create a key. Copy it somewhere safe — it’s shown only once.
2. Point your code at Metrana. Initialize the client with your API key and the workspace, project, and run names. New projects and runs are created automatically the first time you log to them.
3. Log metrics from your training loop, then finish the run.

Follow the Quickstart for the exact code.

What shows up where

Once your code is logging, here’s where each kind of data appears in the portal:

What you log	Where it appears
The run itself (name, config)	The runs table, with config as attributes
Hyperparameters / metadata (config)	A run’s Overview attributes
Training metrics	A run’s Charts and Dashboard
Evaluation metrics	Same charts, distinguished from training data
Reinforcement learning per-episode / per-step data	The run’s Environments views

To understand how these map to Metrana’s underlying entities, see Core concepts and Metrics, series & attributes.

Next steps

• Explore the runs table
• Visualize metrics with charts
• Compare multiple runs