How to Use Dashboard

Continuing from our Getting Started guide, you might be wondering how to monitor your AI’s training progress. While printing to the console works for simple examples, real-world projects often require more sophisticated monitoring.

The RLMatrix Dashboard provides real-time visualization of training metrics, helping you determine if your AI is learning effectively, has finished training, or is failing to make progress.

Setting Up the Dashboard

Prerequisites

Make sure you have the .NET Web SDK installed on your system (Microsoft.NET.Sdk.Web). This is required to run the dashboard application.

Option 1: Clone the Full Repository

If you want the full RLMatrix source code:

Clone the full RLMatrix repository

git clone https://github.com/asieradzk/RL_Matrix.git

Option 2: Get Just the Dashboard

If you only need the dashboard component:

Clone just the dashboard using Git sparse-checkout

git clone --filter=blob:none --sparse https://github.com/asieradzk/RL_Matrix.git
cd RL_Matrix
git sparse-checkout set src/RLMatrix.Dashboard

Running the Dashboard

Once you have the dashboard code, you can:

1. Open the project in your preferred editor (like Visual Studio or VS Code)
2. Build and run the project

Alternatively, you can build the binary and run that directly.

Since this is a Blazor Server application, you’ll see a console window with output similar to:

Dashboard Console Output

info: Microsoft.Hosting.Lifetime[14]
Now listening on: https://localhost:7126
info: Microsoft.Hosting.Lifetime[14]
Now listening on: http://localhost:5069
info: Microsoft.Hosting.Lifetime[0]
Application started. Press Ctrl+C to shut down.
info: Microsoft.Hosting.Lifetime[0]
Hosting environment: Development
info: Microsoft.Hosting.Lifetime[0]
Content root path: D:\path\to\your\dashboard\RLMatrix.Dashboard

If a browser window doesn’t open automatically, navigate to https://localhost:7126 in your browser.

Tip

When you first open the dashboard, the experiment dropdown in the top-left corner will be empty. This is normal - it will populate once training starts.

Connecting Your Project to the Dashboard

Let’s revisit our pattern matching example from the Getting Started guide. When you run your training code, you should see a message in the console:

Connected to dashboard

If you don’t see this message, something went wrong with the connection. Double-check that:

1. The dashboard is running
2. Your training code has the correct connection settings
3. No firewall is blocking the connection

Interpreting Dashboard Data

After starting your training, refresh the dashboard page. You should now see your experiment name in the top-left dropdown menu. Select it to view your training metrics:

The dashboard displays several important metrics:

Reward Over Time

This graph shows the reward received per episode. For our pattern matching example, you should see it award -1 less frequently as the AI learns the correct pattern.

Cumulative Reward Over Time

This shows the sum of rewards over time. The change of slope of this graph is particularly important - slope becomes steeper as your AI learns to maximize rewards.

What to Look For

A steadily increasing slope in the cumulative reward graph is one of the best indicators that your model is improving. Constant slope might mean the model is stuck and needs higher exploration stimulation.

Episode Length Over Time

This tracks how many steps each episode takes. In our simple pattern matching example, this will always be 1 since we take one step per episode. In more complex environments (like Cart-Pole), this metric can reveal if your agent is surviving longer over time.

Loss Over Time

This shows the neural network’s loss function value. Look for a generally decreasing trend - this indicates your model is fitting the data better.

Learning Rate Over Time

This displays changes in the neural network’s learning rate. By default, RLMatrix uses a scheduler that causes small oscillations during training.

Real-Time Updates

One of the most powerful features of the dashboard is its real-time updating. As your training progresses, the dashboard will update automatically to reflect the latest metrics.

In our pattern matching example, you can press Enter in your console window to continue training and watch the dashboard update in real-time.

Exporting Training Data

For further analysis or record-keeping, you can export your training data by clicking the “Export as CSV” button in the top-right corner of the dashboard.

This allows you to:

• Perform custom analysis in Excel or other tools
• Compare different training runs
• Share results with colleagues

Next Steps

Now that you understand how to monitor your training with the dashboard, you’re ready to explore more complex environments and action spaces.

Working with Continuous Actions Learn how to train agents with continuous action spaces for more realistic control problems.