Installation instructions for PyAutoGalaxy do not work

[pedro-acunha]

Description:

I encountered issues while following the current installation instructions for PyAutoGalaxy. The steps provided in the documentation did not work as expected. After troubleshooting, an alternative method that successfully installed PyAutoGalaxy was found.

Steps to Reproduce:

1. Follow the current installation instructions as provided in the documentation.
2. Attempt to install PyAutoGalaxy using the specified commands.
3. Running example notebook.

Expected Result: PyAutoGalaxy should be run without any errors.

Actual Result: The installation is successful, running the examples failed with various errors.

Solution: I was able to successfully install and run PyAutoGalaxy after using the following steps:

conda create -n pyautogalaxy python=3.11
conda activate pyautogalaxy
pip install --upgrade pip
pip install autogalaxy --no-cache-dir
pip install numba --no-cache-dir

Steps Taken:

1. Created a new conda environment with Python 3.11.
2. Activated the newly created environment.
3. Upgraded pip to the latest version.
4. Installed autogalaxy using pip with the --no-cache-dir option to avoid caching issues.
5. Installed numba using pip with the --no-cache-dir option.

Request: Please update the installation instructions in the documentation to reflect the above steps or provide alternative troubleshooting steps for users who encounter similar issues.

Environment:

• Operating System: Ubuntu 24.04 LTS
• Python Version: 3.11
• Conda Version: 24.5.0

Thank you for your attention.

Additional Context: Examples of error messages and conflicts:

AttributeError: 'Parameter' object has no attribute 'clone'
AttributeError: module 'numpy' has no attribute 'long'

[Jammy2211]

Thank you for the info.

Do you know what step in your instructions is different to this example?

https://pyautogalaxy.readthedocs.io/en/latest/installation/conda.html

They look identical to me, except that you are using --no-cache-dir which doesn't feel like it should matter.

[pedro-acunha]

Hi @Jammy2211,

The main difference lies in the --no-cache-dir, setting the python version to 3.11, and not doing conda create -n autogalaxy astropy scikit-image scikit-learn scipy. By creating the env and then running pip install autogalaxy allows for conflicts to be solved. In my previous message I posted some of the AtributeErrors that was shown when trying to run the example notebooks. I believe that by using --no-cache-dir it helps to avoid cache-related issues, specially if different package versions are requested, probably created when the requested packages are installed, alongside with the new env.

This came around after multiple people tried to install PyAutoGalaxy and we were stuck with AttributeErrors. It is probably not the most elegant one...

[Jammy2211]

Ok thanks, I'll try doing some installs via conda and update the installation instructions!

[Jammy2211]

I have updated the docs based on your suggestions :).

The docs themselves are not currently building due to some other error, but I'll close this.