Practice lab C3W3 not working locally - many dependencies issues

I am trying to get the lab to work locally and run into dependencies problems.

Please, can you provide guidance.

Francis



What I tried getting mad:

    My projects on vscode didn't work well (on a poetry environment)
    I noticed tensorflow requires python 3.9 and Gym is only installable using conda so i created a conda virtual environment with Python 3.9.18
    
    I read this post [https://community.deeplearning.ai/t/error-on-jupyter-lab/24052](https://error-on-jup) so created a new venv Python 3.7.16
    
    I started reinstalling all and checked all installed versions on online Labs.
    [In]
    import pandas as pd
    pd.__version__
    [Out]
    '1.0.3'

To work on the reinforcement learning lab locally, you will need to have the following specific versions of Python and the mentioned libraries installed:

    Python Version: Python 3.7 or higher
    
    Library Versions:
    
        OpenAI Gym: Version 0.18.0
        NumPy: Version 1.21.2
        Matplotlib: Version 3.4.3
        TensorFlow: Version 2.6.0
        Keras: Version 2.6.0
    
    
    conda activate course_3_7
    (course_3_7) me@me:~$ conda install pandas==1.0.3
    '1.18.4'
    
    conda install ipykernel
    pip install gym==0.24.0
    
    pip install Pillow==7.1.2
    conda install tensorflow==2.1.0 > issues
    
    DeprecationWarning: `np.bool` is a deprecated alias for the builtin `bool`.
    
    Warning: Gym version v0.24.0 has a number of critical issues with `gym.make` such that the `reset` and `step` functions are called before returning the environment. It is recommend to downgrading to v0.23.1 or upgrading to v0.25.1
    
    pip install gym==0.23.1
    pip install imageio (Ok)
    
    pip install pyvirtualdisplay==3.0
    
    sudo apt-get install xvfb
    No module named 'tensorflow_core.estimator'

    conda install -c conda-forge matplotlib

… then I stopped try to make it work by myself and asked you for your help.


My machine:
Debian12 with this GPU: VGA compatible controller: NVIDIA Corporation GA106M [GeForce RTX 3060 Mobile / Max-Q] (rev a1)
On my system root python is Python 3.11.2
Architecture : x86_64
Processor(s) : 16
Identifiant constructeur : AuthenticAMD
Nom de modèle : AMD Ryzen 7 5800H with Radeon Graphi

Start by installing Anaconda for Linux:

That should cover most of the dependencies, unless you are referring to support (.py) files.

Hello Nevermnd,
I have just installed anaconda-navigator right now.

Created an environment with python 3.7 but cannot install gym 0.24.0 using
anaconda.
gymnasium 0.28.1 requires python >=3.9


Anaconda installation required 9GB of memory, plus an additional 4GB for the virtual environment and didn’t resolve the problem of ‘gym’ obsolescence.

I had to open the hood and dirty my hands but eventually I got it working.

Francis

Gym is a problem, as it is no longer supported.

Search online for its status, there is a successor package.

Yes TMosh thank you :smile:

Special thanks to Dibya Chakravorty. With his help, I successfully navigated through the issue and safely landed my moon lander. Hooray!
Youtube link [https://www.youtube.com/watch?v=e3DyCg0fgx0](https:// Dibya Chakravorty)

Before I forget, another tricky issue:

# Set the random seed for TensorFlow
# tf.random.set_seed(utils.SEED)
# tf.random.set_random_seed(utils.SEED) deprecated at home
tf.compat.v1.random.set_random_seed(utils.SEED) # use this instead

If you are interested in getting the lab working at home, you can ask me for details.

Francis

1 Like

Great news!