Installation Troubleshooting#

Cannot install PyQt5 dependency#

Idtrackerai’s GUIs work (by default) with PyQt5. But if its installation fails, giving some of the next errors:

  • sipbuild.exceptions.UserException

  • error: metadata-generation-failed

  • sipbuild.pyproject.PyProjectOptionException

you can choose to install any of the remaining Qt bindings for Python (you only need one of them to succeed):

Note

If any Qt installation works on your computer, you still can use idtrackerai non-GUI parts by following the next commands without any Qt installation.

Qt binding options in Python#
python -m pip install PyQt5 # default
python -m pip install PyQt6
python -m pip install PySide6
python -m pip install PySide2

And, once any of them succeed, you will have to install idtrackerai without the PyQt5 dependency:

# install idtrackerai without any dependency
python -m pip install idtrackerai --no-deps

# install all remaining dependencies except PyQt5
python -m pip install numpy rich h5py scipy opencv-python-headless qtpy superqt toml matplotlib

Not recognized command#

If you just installed idtracker.ai and idtrackerai_test gets a very short error like No such file or directory or Not recognized command, try reactivating the Conda environment:

conda deactivate
conda activate idtrackerai

Could not load library libcudnn_cnn_infer.so.8#

If idtrackerai_test starts but after some seconds you get something like Could not load library libcudnn_cnn_infer.so.8, you are missing the Cuda Toolkit dependency, install it with:

conda install cudatoolkit=11.8 -c conda-forge

qt.qpa.plugin: Could not load the Qt platform plugin “xcb”#

Read this thread. Alternatively, in Ubuntu sudo apt install libxcb-cursor0 solves the problem.

module ‘torch’ has no attribute ‘_six’#

In MacOS, running

pip install --upgrade torch

seems to fix the issue.

No graphic device was found available#

If your computer has a NVIDIA or AMD GPU or uses MacOS >= 12.3 with M1, M2 or AMD GPU and idtrackerai cannot find the specific device available, your PyTorch installation is malfunctioning.

To fix that, you have to Uninstall the entire Conda environment and try again. Read carefully the PyTorch indications depending on your machine. Getting PyTorch to use GPU acceleration can be tricky sometimes.

Any other error#

Send us your error to idtrackerai@gmail.com and we will assist you.