Run Jupyter Notebooks on a Cloud GPU Inside VS Code: Step-by-Step Guide

1) Prerequisites

• VS Code ≥ 1.80

• Thunder Compute account (includes $20 free GPU credit)

• Thunder Compute VS Code extension

• (Optional) Jupyter extension for VS Code

2) Install and Authenticate

1. Open Extensions (Ctrl + Shift + X) or click here.

2. Search “Thunder Compute” and click Install.

3. Press F1 → Thunder: Login.

4. Paste your Thunder Compute API token.

   • If you need one: Thunder: Get API Token opens the console.

3) Create a Cloud GPU Instance

1. Click the ⚡ Thunder Compute icon in the activity bar.

2. Press Thunder: Create Instance (plus icon)

3. Choose GPU type (A100, T4, etc.), vCPUs, and disk size and click "Create Instance."

4. Wait until the instance status reads Running.

4) Connect with Remote‑SSH

1. In the Thunder sidebar, right‑click the instance → Connect to Instance.

2. VS Code opens a new remote window rooted at ~/.

5) Open a Jupyter Notebook

1. VS Code detects .ipynb files and prompts to install the Jupyter extension inside the remote VM. Accept the prompt.

2. Create a new notebook: F1 → Jupyter: Create New Blank Notebook.

3. Select the kernel Python 3 (ipykernel).

6) Verify GPU Access (PyTorch)

Expected

7) Manage Instance State

8) Thunder Compute Extension Reference

• Simple Authentication: secure API‑token login

• Instance Management: list, start, stop, delete, refresh

• One‑Click Connect: SSH session launched automatically

• Cost Control: stop instances from VS Code to pause charges

9) Next Steps

• Benchmark your GPU cost with the A100 Pricing Showdown 2025 article.

• Check out the Thunder Compute documentation

• Automate environment setup with Thunder Compute snapshots.

Quick Shut‑Down Reminder

When you finish a notebook session, run Thunder: Stop Instance or press the ⏹ Stop button in the sidebar to avoid unnecessary charges.