anipose
anipose copied to clipboard
Consider using a conda config file for installation
Currently instructions indicate that deeplabcut should be installed as a pip dependency but don't go into how to get a CPU or GPU dlc version. DLC solved this by just making conda config files for both options. I've pasted a conda specification file that I used to install a anipose-modified GPU DLC installation. Might be useful to supply this (along with CPU version that I didn't make) for easy install.
# FIRST: INSTALL CORRECT DRIVER for GPU, see https://stackoverflow.com/questions/30820513/what-is-the-correct-version-of-cuda-for-my-nvidia-driver/30820690
#
# install: conda env create -f anipose.yaml
# update: conda env update -f anipose.yaml
name: anipose
dependencies:
- python=3.7
- pip
- tensorflow-gpu==1.13.1
- cudnn=7
- wxpython
- jupyter
- nb_conda
- Shapely
- ffmpeg
- mayavi
- pip:
- deeplabcut
- anipose
This is a good idea! Somehow I thought that people might install deeplabcut before anipose, and hence use their conda file. Still, a conda file just for Anipose would be useful, especially if we have different dependencies down the line.
I'll test out the conda file and make sure it installs correctly when I get a chance.