Position pipeline to incorporate DeepLabCut into Spyglass
- Install miniconda (or anaconda) if it isn't already installed.
- INSTALL CORRECT DRIVER for GPU, see https://stackoverflow.com/questions/30820513/what-is-the-correct-version-of-cuda-for-my-nvidia-driver/30820690
git clone https://github.com/LorenFrankLab/spyglass-dlc.git- Setup editiable package with dependencies
cd spyglass-dlc
conda env create -f environment.yml
conda env update -f environment.yamlafter installing do the following:
conda activate spyglass-dlc
conda env config vars set LD_LIBRARY_PATH=~/path/to/anaconda3/envs/spyglass-dlc/lib/
mamba install -c conda-forge wxpython
mamba install pytorch torchvision torchaudio cudatoolkit=11.6 -c pytorch -c conda-forge
cd spyglass
pip install -e .
cd spyglass-dlc
pip install -e .Add the following environment variables (e.g. in ~/.bashrc). If from the Frank Lab, we assume that you are interacting with the database on a computer that has mounted nimbus at /nimbus (if the mount location is different, change accordingly). For this to take effect, log out and log back in, or run source ~/.bashrc in the terminal.
nano ~/.bashrcand add:
export DLC_PROJECT_PATH="/nimbus/deeplabcut/projects"
export DLC_VIDEO_PATH="/nimbus/deeplabcut/video"
export DLC_OUTPUT_PATH="/nimbus/deeplabcut/output"