Install TensorFlow in Jupyter Notebook: Step-by-Step

This guide walks you through installing TensorFlow so you can use it directly inside Jupyter Notebook. You’ll set up a Python environment, install the TensorFlow package, and verify the installation by running a simple import and version check in a notebook cell. No prior TensorFlow experience is needed, but you should be comfortable with basic command-line operations.

What You’ll Need

Python 3.8–3.12 installed on your system (check with python --version).
pip (Python’s package installer) – usually comes with Python.
Jupyter Notebook or JupyterLab installed. If not, you’ll install it as part of the steps.
An internet connection to download packages.
Optional but recommended: a virtual environment tool like venv or conda to keep dependencies isolated.

Step-by-Step

Open your terminal or command prompt. On Windows, use Command Prompt or PowerShell; on macOS/Linux, use Terminal.
Create a virtual environment (recommended). This avoids conflicts with other Python projects.
```
python -m venv tf_env
```
Activate the environment.
- Windows: tf_env\Scripts\activate
- macOS/Linux: source tf_env/bin/activate
Install Jupyter Notebook if you don’t already have it in this environment:
```
pip install jupyter
```
Install TensorFlow. The standard CPU version is fine for most learning and development:
```
pip install tensorflow
```
For GPU support (requires compatible NVIDIA GPU and CUDA setup), use:
```
pip install tensorflow-gpu
```
(Note: As of TensorFlow 2.x, tensorflow-gpu is deprecated; just use tensorflow and install CUDA/cuDNN separately if needed.)
Launch Jupyter Notebook. While still in the activated environment, run:
```
jupyter notebook
```
This opens the Jupyter interface in your browser.
Create a new notebook. Click “New” → “Python 3” (or the kernel matching your environment).
Test the installation. In the first cell, type:
```
import tensorflow as tf
print(tf.__version__)
```
Run the cell (Shift+Enter). If you see a version number like 2.16.1 without errors, TensorFlow is ready to use.
Optional: Verify GPU availability (if you installed GPU support):
```
print(tf.config.list_physical_devices('GPU'))
```
An empty list means TensorFlow isn’t using the GPU; a list with GPU details means it is.

Common Mistakes

Installing TensorFlow in the wrong environment. If you activate a different environment later, TensorFlow won’t be available. Always activate the same environment before launching Jupyter.
Using an outdated Python version. TensorFlow requires Python 3.8–3.12. If you have Python 3.13 or older 3.7, pip may fail to find a compatible version.
Running pip install tensorflow outside the virtual environment. This can pollute your global Python and cause permission errors.
Forgetting to restart the kernel after installation. If you installed TensorFlow while Jupyter was already running, restart the kernel (Kernel → Restart) to pick up the new package.
Mixing pip and conda packages. If you use conda, prefer conda install -c conda-forge tensorflow to avoid dependency conflicts.

Tips

Use pip install --upgrade pip before installing TensorFlow to avoid resolver issues.
Keep your environment specific to TensorFlow projects. Create a new one for other projects.
If you have both CPU and GPU versions, uninstall one to avoid conflicts: pip uninstall tensorflow tensorflow-gpu then reinstall the one you need.
For faster installation, use a mirror if you’re in a region with slow PyPI access, e.g., pip install tensorflow -i https://pypi.tuna.tsinghua.edu.cn/simple (but verify the mirror’s trustworthiness).
Check TensorFlow’s official installation page for the latest version-specific instructions.

Troubleshooting

“No module named tensorflow” – This means the package isn’t installed in the environment your notebook is using. Check your kernel: in Jupyter, go to Kernel → Change Kernel and select the one that matches your virtual environment. Also, run pip list | grep tensorflow in the terminal to confirm it’s installed.

pip error: “Could not find a version that satisfies the requirement tensorflow” – Your Python version is likely unsupported. Check python --version. If it’s 3.13 or higher, install Python 3.12 or use a conda environment with Python 3.10.

ImportError: DLL load failed (Windows) – This often happens due to missing Microsoft Visual C++ Redistributable. Download and install the latest “vc_redist.x64.exe” from Microsoft’s official site.

Kernel dies or notebook crashes when importing TensorFlow – This can be due to insufficient RAM or an old CPU without AVX support. Close other applications, or try installing a CPU-only build with pip install tensorflow-cpu (if available for your version).

GPU not detected – Ensure you have the correct NVIDIA drivers, CUDA Toolkit, and cuDNN matching the TensorFlow version. Check TensorFlow’s GPU guide for the exact versions. You can also run nvidia-smi to see your driver version.

Permission errors during install – On macOS/Linux, use pip install --user tensorflow if you don’t want to use a virtual environment, but better to create one. On Windows, run the terminal as Administrator if you’re not using a venv.

Frequently Asked Questions

Q: Do I need to install Jupyter separately if I already have Anaconda?
No, Anaconda comes with Jupyter Notebook and conda. You can create a conda environment and install TensorFlow with conda install -c conda-forge tensorflow.

Q: Can I install TensorFlow in a notebook cell directly?
Yes, you can run !pip install tensorflow in a cell. However, it’s better to install in the terminal to avoid path issues and to keep the notebook clean.

Q: What is the difference between TensorFlow and TensorFlow-GPU?
TensorFlow-GPU is the legacy package for GPU support. Since TensorFlow 2.1, tensorflow includes GPU support if CUDA and cuDNN are installed. For most users, just install tensorflow.

Q: How do I know if TensorFlow is using my GPU?
Run tf.config.list_physical_devices('GPU'). If it returns a non-empty list, GPU is being used. You can also add tf.debugging.set_log_device_placement(True) to see device assignments in logs.

Q: Can I use TensorFlow with JupyterLab?
Yes, JupyterLab works the same way. Install JupyterLab with pip install jupyterlab, launch it, and create a notebook. The steps are identical.

Q: Why does my installation take so long?
TensorFlow is a large package (hundreds of MB). The download time depends on your internet speed. Using a faster mirror or a wired connection can help.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.

More cards in this drawer

The whole Numbers and Powers drawer →