Setup HDDM on Windows

setup_hddm_win.md

Install HDDM

HDDM is a python project aiming at fitting hierarchical drift diffusion model. But it is complex to install it on Windows. Here I record what I tried.

Prerequisite

• Install Microsoft C++ Build Tools, but see this discussion. Remember to add C++ tools when installing.
• Install Miniconda 3.
• Git for Windows is also required for the GFW blocks Github. In this way, we need ssh protocol (see this documentation) access to Github.

Steps

The steps is generally the very similar to those on README of HDDM.

pip install cython
pip install pymc==2.3.8
pip install git+https://github.com/hddm-devs/kabuki
pip install git+https://github.com/hddm-devs/hddm
# Optional
pip install torch torchvision torchaudio

What I really do is:

1. Create a new environment by conda create -n ddm python=3.8. Note we have to use python 3.8 because pymc 2.3.8 does not support python 3.9 or higher, but HDDM uses pymc 2.3.8.
2. Activate ddm environment and execute these lines:

conda install cython
conda install pymc==2.3.8 # pymc install from conda will work
pip install git+ssh://git@github.com/hddm-devs/kabuki # use ssh protocol to bypass GFW
pip install git+ssh://git@github.com/hddm-devs/hddm

It is done then.