Tensorflow not found using pip

Python and system compatibility

Before proceeding, verify that your Python version is compatible and you have a 64-bit installation:

# Do "64-bit" check (no, it's not a weight check)
import struct
print(struct.calcsize("P") * 8)


Regrettably, TensorFlow has a "not-so-32-bit-friendly" policy. You may grab a 64-bit Python from here, and remember, for Windows, Python 3.5.x to 3.8.x plays nicely with TensorFlow. If post-installation you meet unexpected errors, a system reboot could be a quick fix.

Direct package installation via URL

In some cases, installing TensorFlow directly from a wheel URL may be the answer. Here's the list of fresh URLs from TensorFlow's kitchen: tensorflow.org/install/pip#package-location

To install TensorFlow from a URL:

# Install from URL (like ordering pizza online, but for TensorFlow)
python3 -m pip install --upgrade URL_TO_TENSORFLOW_WHEEL


Note: Replace URL_TO_TENSORFLOW_WHEEL with the URL of your choice.

Alternate distribution routes

Say you're on Windows and the standard TensorFlow distribution frowns at you. Well, Anaconda might be your helpful friend, specifically versions 3-5.2.0. But caution, version 3-5.3.0 on Windows can be a thorn in the side!

Contriving virtual environments

Among the best practices is setting up a virtual environment. It’s the cordial host for different project dependencies, preventing a messy "dependency soup". Here's how to cook one up:

# Create a virtual environment (your neighbor-free zone!)
python -m venv myenv
# Activate your new home
source myenv/bin/activate # Unix-like
myenv\Scripts\activate # Windows


Now, install TensorFlow within this clean environment for a peaceful coexistence with global packages.

Use pip like a pro

To ensure that you are doing it right, use this module form of pip:

# Install TensorFlow right (and grab a cup of coffee)
python -m pip install tensorflow


Before you go, don't forget TensorFlow demands its butler, pip, to be at least version 8.1. So keep pip up-to-date:

# Upgrade pip to the latest (but remember, age is just a number!)
pip install --upgrade pip


After installation woes

If you're stranded on a 'cannot find TensorFlow' island after installation, you might be facing a PATH issue. Make sure Python binaries aren't playing hide 'n seek with your system. Modify your shell profile on Unix, or adjust your environmental variables on Windows.

Check your TensorFlow installation by boldly attempting to import it in a Python session. If this fails, do a Sherlock Holmes with verbose logging (-vvv):

# Install TensorFlow with verbose logging (Get ready for some deep details!)
python3 -m pip install tensorflow -vvv
