RL Training Dashboard

Real-time dashboard for watching RL training metrics. Built with Streamlit and Plotly because I needed to see what was happening during training without tailing log files.

What it does

• Shows current step, average reward, and loss
• Updates every 2 seconds while training runs
• Plots reward and loss curves
• Dark theme (easier on the eyes during long training runs)

Setup

pip install -r requirements.txt
streamlit run app.py

Dashboard opens at http://localhost:8501

Use it in your training loop

from train_logger import TrainingLogger

logger = TrainingLogger()

for step in range(1000):
    # your training code here
    reward = compute_reward()
    loss = compute_loss()
    
    logger.log_step(step, reward, loss)

Metrics get written to metrics.json and the dashboard reads from there.

Files

• app.py - Streamlit dashboard
• train_logger.py - Logs metrics to JSON
• requirements.txt - Dependencies

Built this to monitor my parkour RL agent training.