While trying to install the latest version of Keras (3.4.1) with TensorFlow, I ran into some dependency issues that many developers might find frustrating. After a bit of troubleshooting, I was able to solve the problem and thought I’d share the solution here in case it helps anyone facing similar challenges.

When I tried to install Keras 3.4.1 using Conda with TensorFlow 2.17, I encountered the following error:

LibMambaUnsatisfiableError: Encountered problems while solving:
  - nothing provides bleach 1.5.0 needed by tensorboard-1.6.0-py27hf484d3e_0

The conflict involved dependencies between tensorboard and bleach, along with version issues around TensorFlow and Python. I was getting errors about incompatible versions of Python and some packages that didn’t seem to exist.

Here’s how I fixed the issue:

1. Update Conda: First, I ensured that Conda was up to date to avoid any versioning conflicts.

Do not forget to deactivate the running conda environment.

conda update conda
conda config --add channels conda-forge

2. Create a Clean Environment: I created a new Conda environment with a compatible version of Python to isolate the issue.

conda create -n tf_keras_env python=3.9
conda activate tf_keras_env

3. Install TensorFlow and Keras with Conda: Instead of using pip, I wanted to keep using
   conda to install TensorFlow and Keras, which handled the dependencies much better.

conda install tensorflow==2.17.0 keras==3.4.1

This avoided the issues I was having with tensorboard and bleach when using pip.

4. Pinning TensorBoard and bleach: In case you still run into issues, you can manually install the required versions of bleach and tensorboard before installing Keras and TensorFlow:

conda install bleach=4.1.0 tensorboard=2.17.0

Then, proceed with the TensorFlow and Keras installation.

Dependency conflicts like these can be tricky, but by updating Conda, creating a clean environment, and continuing with conda for the installation, I was able to solve the issue. Hopefully, this guide can save you some time if you’re stuck with similar errors.