There a few different images, see hcp4715/hddm/tags.

These three are currently used images:

hddm:0.9.8: Python 3.8.8, the latest version of HDDM 0.9.8. No example folder inside this image yet.

hddm:0.8: Python 3.8.8, HDDM 0.8, and kabuki 0.6.4. This version is more stable with classic HDDM functions.

hddm:0.8_tutorial: Python 3.8.8, HDDM 0.8, kabuki 0.6.4, and with functions bridging HDDM model results and arviz. This version is associated with our tutorial paper: A Hitchhiker’s Guide to Bayesian Hierarchical Drift-Diffusion Modeling with dockerHDDM .

These two are historical images, I keep them in case that some one has used it before.

hddm:arviz: Python 3.7, HDDM 0.8, kabuki 0.6.4, and testing scripts that aimed at working with arviz.

hddm:tutorial_alpha: Python 3.8.8, HDDM 0.8, and kabuki 0.6.4 installed directly from github, also with testing scripts that aimed at working with arviz

HDDM is a python package for hierarchical drift diffusion modelling, see here for more.

This docker image was based on a previous HDDM docker image by Mads (@madslupe), with a few improvements.

• Parallel processing with p_tqdm (pathos and tqdm).
• With the latest HDDM package.
• A few new python packages that are used for plotting (seaborn, plotly).
• Related packages for Bayesian inference (Arviz).

In the example folder of this docker image (see below on how to use this image), you can find jupyter notebooks. Most of them include a jupyter notebook that reproduce the official tutorial.

First, install docker and test it. There are many tutorials on this.

Please read this post on docker's website for installing docker for linux (ubuntu included).

Then, pull the current docker image from docker hub:

docker pull hcp4715/hddm:0.8

Note: you may need sudo permission to run the command docker.

Please read this post for installing docker on window 10. During the installation of docker, you might be instructed to install/update your window subsystem linux (WSL), please follow the instruction to finish the installation of docker.

Then, open window power shell, and pull the current docker image from docker hub:

docker pull hcp4715/hddm:0.8

After pulling it from docker hub, you can then run jupyter notebook in the container (e.g., in bash of linux or power shell of windows):

docker run -it --rm --cpus=4 \
-v /home/hcp4715/hddm_docker:/home/jovyan/work \
-p 8888:8888 hcp4715/hddm:0.8 jupyter notebook

docker run -it --rm --cpus=4 -v /d/hcp4715/hddm_docker:/home/jovyan/work -p 8888:8888 hcp4715/hddm:0.i jupyter notebook  

docker run ---- Run a docker image in a container

-it ---- Keep STDIN open even if not attached

--rm ---- Automatically remove the container when it exits

--cpus=5 ---- Number of cores will be used by docker

-v ---- Mount a folder to the container

/home/hcp4715/hddm_docker ---- The directory of a local folder where I stored my data. [For Linux]

/d/hcp4715/hddm_docker ---- The directory of a local folder under drive D. It appears as D:\hcp4715\hddm_docker in windows system.

/home/jovyan/work ---- The directory inside the docker image (the mounting point of the local folder in the docker image). Note that the docker container itself likes a mini virtual linux system, so the file system inside it is linux style.

-p ---- Publish a container’s port(s) to the host

hcp4715/hddm:0.8 ---- The docker image to run, 0.8 after : is the tag of the current docker image.

jupyter notebook ---- Open juypter notebook when start running the container.

After running the code above, bash will has output like this:

....
....
To access the notebook, open this file in a browser:
        file:https:///home/jovyan/.local/share/jupyter/runtime/nbserver-6-open.html
Or copy and paste one of these URLs:
    http:https://174196acc395:8888/?token=75f1a7a8ffcbb55f0c2802433a9a5d57ac00868e05089c09
 or http:https://127.0.0.1:8888/?token=75f1a7a8ffcbb55f0c2802433a9a5d57ac00868e05089c09

Copy the full url (http:https://127.0.0.1:8888/?token=.......) to a browser (firefox or chrome) and it will show a web page, this is the interface of jupyter notebook! Note, in Windows system, it might be localhost instead of 127.0.0.1 in the url.

Under the Files tab, there should be three folders: work and example. The work folder is the local folder mounted in docker container. The example folder was the one built in docker image, this folder includes one dataset and one jupyter notebook, you can test the parallel processing by running this jupyter notebook.

Enter work folder, you can start your analysis within jupyter notebook.

The example folder also includes two other jupyter notebooks, HDDM_official_tutorial_reproduced.ipynb reproduces the tutorial code, using the HDDM in this docker image.

docker run -it --rm --cpus=4 \
-p 8888:8888 hcp4715/hddm:0.8 jupyter notebook

• Permission denied. If you still encounter this error, please see this post about the permission problem.

An alternative way to get the docker image is to build it from Dockerfile.

I built this docker image under Ubuntu 20.04.

This Dockerfile is modified by Dr. Rui Yuan @ Stanford, based on the Dockerfile of jupyter/scipy-notebook. We installed additional packages for HDDM and ipyparallel, and configured the ipyparallel so that we can run it in jupyter noebook (doesn't work for jupyterlab yet). See Dockerfile for the details

Code for building the docker image (don't forget the . in the end):

docker build -t hcp4715/hddm:0.8 -f Dockerfile .

Thank @madslupe for his previous HDDM image, which laid the base for the current version.

Thank Dr Rui Yuan for his help in creating the Dockerfile.

If you have any problem in using this docker image, please report an issue at the github repo