VS Code Remote Debugging

Attach the VS Code debugger to your Valohai execution for interactive debugging with breakpoints, variable inspection, and step-through execution.

What you'll accomplish:

• Set breakpoints in VS Code on code running in Valohai

• Inspect and modify variables in real-time

• Step through execution line-by-line on cloud infrastructure

Prerequisites

Before starting, ensure you have:

1. SSH access configured - Follow SSH Overview to set up keys and firewall rules

2. VS Code installed with Python extension

3. SSH private key from the SSH setup

Add debugpy to Your Code

debugpy implements the Debug Adapter Protocol for Python. Install it and add wait logic to your training script.

Python Code

import debugpy
import numpy as np
import time

# Listen on port 5678 inside the container
debugpy.listen(5678)

# Halt execution until VS Code attaches
debugpy.wait_for_client()

# Your training code runs after debugger connects
for x in range(1, 20):
    print(f"Doing computation {x}")
    data = np.random.random((50, 50))
    sum = np.sum(data)
    time.sleep(2)

Key points:

• debugpy.listen(5678) - Container port for debug connection

• debugpy.wait_for_client() - Pauses until VS Code attaches

valohai.yaml

- step:
    name: train
    image: python:3.9
    command:
      - pip install numpy debugpy
      - python train.py

The execution will pause at wait_for_client() until you attach VS Code. This keeps it running without needing sleep commands.

Start Execution with SSH

Launch your execution with SSH debugging enabled:

Web UI:

1. Enable Run with SSH

2. Paste public key or auto-generate

3. Start execution

Command Line:

vh exec run --adhoc \
  --debug-key-file=/path/to/key.pub \
  --debug-port 2222 \
  train

Wait for the IP address in execution logs:

SSH connection available at: 52.214.159.193:2222

Configure VS Code

1. Create Debug Configuration

Create or edit .vscode/launch.json in your project:

{
  "version": "0.2.0",
  "configurations": [
    {
      "name": "Python: Remote Attach",
      "type": "python",
      "request": "attach",
      "connect": {
        "host": "localhost",
        "port": 5678
      },
      "pathMappings": [
        {
          "localRoot": "${workspaceFolder}",
          "remoteRoot": "/valohai/repository"
        }
      ]
    }
  ]
}

Configuration details:

• host: localhost - We'll tunnel through SSH

• port: 5678 - Must match debugpy.listen() port

• remoteRoot: /valohai/repository - Where your code runs in the container

2. Open SSH Tunnel

Create a tunnel from your local machine to the execution:

ssh -i /path/to/private-key <IP-FROM-LOGS> -p 2222 -L 5678:127.0.0.1:5678

What this does:

• Connects to Valohai execution via SSH

• Forwards local port 5678 → container port 5678

• Keeps connection open for debugging

💡 Keep this terminal window open while debugging.

3. Attach Debugger

1. Open your train.py in VS Code

2. Set breakpoints by clicking left of line numbers

3. Open Run and Debug panel (Ctrl+Shift+D)

4. Select Python: Remote Attach

5. Click Start Debugging (F5)

Your code will resume from debugpy.wait_for_client() and hit your breakpoints.

---

Debug Your Code

Once attached, you can:

• Inspect variables - Hover over variables or use Variables panel

• Evaluate expressions - Use Debug Console to run Python expressions

• Step through code - F10 (step over), F11 (step into), Shift+F11 (step out)

• Modify variables - Edit values in Variables panel during execution

---

Common Issues

VS Code can't connect?

• Verify SSH tunnel is running (check terminal)

• Ensure execution hasn't timed out

• Confirm debugpy.listen(5678) matches your tunnel port

Path mapping errors?

• Update remoteRoot in launch.json if your code is in a different container path

• VS Code may auto-detect - click "Auto-detect" if prompted

Execution exits before attaching?

• Ensure debugpy.wait_for_client() is called before your training code

• Check execution logs for errors during pip install debugpy

---

Next Steps

• PyCharm Remote Debugging - Alternative IDE debugging

• SSH Overview - Learn about direct SSH access and tunneling