MLCopilot

Real-Time ML Training Monitor for VS Code.

Features

• Automatic Detection: Automatically starts monitoring when you run Python ML scripts
• Live Dashboard: Real-time loss curves and gradient norm visualization
• Failure Detection: Detects exploding/vanishing gradients, loss divergence, NaN losses
• Root Cause Analysis: Explains why issues occur
• Actionable Recommendations: Code-level fixes for detected problems

Requirements

• Python 3.8+
• Python extension for VS Code
• Required Python packages:

  pip install fastapi uvicorn torch numpy websockets
  

Usage

1. Install the extension
2. Open any ML project
3. Run your Python training script
4. The dashboard opens automatically and streams live metrics

How It Works

When you execute a Python file that appears to be an ML training script:

1. The backend server starts automatically
2. Your browser opens to the live dashboard
3. Metrics stream in real-time via WebSocket
4. Issues are detected and recommendations provided

Commands

• MLCopilot: Open Dashboard - Manually open the monitoring dashboard

Extension Settings

This extension uses the Python extension's configured interpreter.

License

MIT