Last modified: Jun 20, 2025 By Alexander Williams
How to Uninstall Keras in Python
Keras is a popular deep learning library. Sometimes, you may need to uninstall it. This guide will help you remove Keras cleanly.
Table Of Contents
Why Uninstall Keras?
You might want to uninstall Keras for several reasons. Maybe you need a different version. Or you're cleaning up your Python environment.
Other times, conflicts with libraries like TensorFlow or PyTorch may require removal.
Check If Keras Is Installed
First, verify if Keras is installed. Run this command in your terminal:
pip show keras
If installed, you'll see package info. If not, you'll get a "not found" message.
Uninstall Keras Using Pip
The easiest way is using pip
. Run this command:
pip uninstall keras
Confirm the removal when prompted. This will remove the main Keras package.
Remove Keras Dependencies
Keras may have dependencies. To remove them too, use:
pip uninstall keras tensorflow
This removes both Keras and TensorFlow, its common backend.
Uninstall Keras in Conda
For Anaconda users, use this command:
conda remove keras
Conda will handle dependencies automatically. Confirm the action when asked.
Verify Uninstallation
After uninstalling, check if Keras is gone:
import keras
You should get an ImportError. This confirms successful removal.
Clean Up Residual Files
Sometimes, files remain after uninstall. To clean them:
pip cache purge
This clears pip's cache. For conda, use conda clean --all
.
Reinstall Keras If Needed
To reinstall Keras later, use:
pip install keras
Or for conda:
conda install keras
Troubleshooting
If you face issues, try these steps:
1. Use pip install --upgrade pip
first
2. Check for multiple Python versions
3. Use virtual environments
For other library uninstalls, see our guide on uninstalling scikit-learn.
Conclusion
Uninstalling Keras is simple with pip or conda. Always verify removal and clean residual files. This keeps your Python environment tidy.
Remember to handle dependencies properly. This prevents conflicts with other ML libraries.