I faile to install SimBA on MacOS Ventura 13.4.1

[DorianBattivelli]

Hello,

I tried to follow these steps:

Create an environment for simba using anaconda terminal.

In the terminal type, pip install simba-uw-tf-dev

Then, conda install -c anaconda python.app

Then, conda install matplotlib

Then, conda uninstall shapely

Then, conda install -c conda-forge shapely

Then, pip install shap

Lastly, pip install h5py

In the terminal, type in simba to test if it works.

But here what I get:

I checked and both Pip and setuptools are up to date. Do you have an idea what is going wrong here?

Thank you, Best,

[sronilsson]

Hi @DorianBattivelli! It looks like your running python3.9. how does it look when you run it in 3.6?

PS. Latest SimBA should run in python >3 6, including 3.9, but requires slightly different installation, including downgrade setuptools and pip I think. But I haven't had time to write documentation. But if I really need 3.9 I can write it.

[DorianBattivelli]

Indeed, I tried to specify the python version, but somehow I cannot do it:

[DorianBattivelli]

Using "python3.6 pip install simba-uw-tf-dev" command, it looks that it moves forward, but I still face the issue:

[sronilsson]

I can't see it in the second screengrab, but in the first, you have created a conda environment for simba using python 3.9:

Instead, when creating the conda environment, do e.g.,:

1) conda create -n simba_env python=3.6 anaconda 2) conda activate simba_env 3) pip install simba-uw-tf-dev

[DorianBattivelli]

Everything run well at first, but an error popped out at the end, and it seems to prevent SimBA to work properly:

[DorianBattivelli]

Ohh wait maybe I should go through the following commands (matplotlib etc.), I'm giving a go..

[sronilsson]

maybe use pip install simba-uw-tf-dev --ignore-installed

[DorianBattivelli]

Still not working :/

[sronilsson]

I think this is a pandas and numpy version mismatch, which version of numpy and pandas do you have pip show numpy and pip show pandas ?

[DorianBattivelli]

Numpy: 1.18.1 Pandas: 1.0.5

[sronilsson]

If you try: pip install pandas==0.25.3, how does it look?

[DorianBattivelli]

Here again I have error:

[sronilsson]

Tro do do pip uninstall matplotlib then conda install matplotlib

[DorianBattivelli]

To resume, I did from scratch:

conda create -n simba_env python=3.6 anaconda
conda activate simba_env
pip install simba-uw-tf-dev --ignore-installed
pip install pandas==0.25.3
pip uninstall matplotlib
conda install matplotlib

And here what I get:

[sronilsson]

Did conda install matplotlib run properly? This suggests you don't have matplotlib in the environment. Did you type y after to confirm install?

[DorianBattivelli]

I just relaunched the command, and here another error:

[sronilsson]

This is a shapely error again, try:

pip uninstall shapely conda install shapely

[DorianBattivelli]

Then terminal returns:

[sronilsson]

:) This is an sklearn error, which version of scikit do you have? pip show scikit-learn

[DorianBattivelli]

0.23.1

[sronilsson]

If you do pip uninstall imblearn then pip install imblearn?

[DorianBattivelli]

Damn :)

[sronilsson]

can you try pip install scikit-learn==0.22.2 ?

[DorianBattivelli]

It's giving the same outcome

[sronilsson]

Does ant of these potential fixes do it? https://stackoverflow.com/questions/56549270/importerror-cannot-import-name-multioutputmixin-from-sklearn-base

[DorianBattivelli]

Ok, I managed to install it, what I did:

• Open anaconda navigator
• Choose create environement from there
• Specify Python 3.6.13

And then run all these commands from this env:

pip install simba-uw-tf-dev

Then, conda install -c anaconda python.app

Then, conda install matplotlib

Then, conda uninstall shapely

Then, conda install -c conda-forge shapely

Not sure what solved the problem here... :D

Thank you!

[sronilsson]

Great!

[DorianBattivelli]

Just to let you know, it's working, but when I'm launching the program I see some errors in the terminal:

Do you think it can be problematic for a stable use of SimBA?

[sronilsson]

You should be OK, just a warning that it might break if you update matplotlib to a newer version