✅ Full Guide: Running Jupyter Notebook on GPUs¶

This guide helps you set up Jupyter Notebook with GPU support using Anaconda, CUDA, cuDNN, and deep learning libraries like PyTorch or TensorFlow.

⚙️ Step 1: Install Anaconda¶

• Download and install from:
  👉 https://www.anaconda.com/products/distribution

• After installation, launch Jupyter Notebook via:

jupyter notebook

This opens Jupyter Notebook in your browser.

⚙️ Step 2: Install CUDA Toolkit¶

CUDA enables your Python libraries (e.g., TensorFlow, PyTorch) to run on NVIDIA GPUs.

• Download from:
  👉 https://developer.nvidia.com/cuda-downloads

• Choose the version that matches:

• Your GPU model

• Your Operating System

• To verify your GPU details:

nvidia-smi

⚙️ Step 3: Install cuDNN Library¶

cuDNN accelerates deep learning on GPUs.

• Download from:
  👉 https://developer.nvidia.com/cudnn

• Match cuDNN version with your installed CUDA version.

After downloading: - Extract the files. - Copy bin/, lib/, and include/ folders into your CUDA installation directory (usually /usr/local/cuda/ on Linux).

⚠️ cuDNN must be manually installed — not via conda/pip.

⚙️ Step 4: Create a Conda Environment (Python 3.8)¶

conda create --name gpu_env python=3.8
conda activate gpu_env

⚙️ Step 5: Install Required Packages¶

Choose one of the following options depending on your framework preference:

🧠 Option A: TensorFlow + Keras (GPU-enabled)¶

conda install -c anaconda tensorflow-gpu keras-gpu

🔥 Option B: PyTorch (GPU-enabled, recommended)¶

conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia

⚠️ Replace 11.8 with your actual CUDA version.

➕ Optional: Add data science packages¶

conda install jupyter numpy pandas matplotlib scikit-learn

⚙️ Step 6: Configure Jupyter to Use GPU Environment¶

python -m ipykernel install --user --name gpu_env --display-name "Python (GPU)"

✅ This registers the environment as "Python (GPU)" in the Jupyter kernel list.

🚀 Step 7: Launch Jupyter Notebook¶

jupyter notebook

• Click New Notebook
• Choose the "Python (GPU)" kernel

✅ Step 8: Verify GPU is Being Used¶

Run this in a notebook cell:

import torch
torch.cuda.is_available()

Expected Output:

True

If True, your environment is GPU-enabled. 🎉

🧠 Quick Troubleshooting¶

Once you configure the versions of CUDA, cudnn and jupyter, you go over steps 7 and 8 again and you should find your notebook running on GPU.