Jupyter quant research environment.
Project description
Jupyter Quant
A dockerized Jupyter quant research environment.
Highlights
- It can be used as a docker image or pypi package.
- Includes tools for quant analysis, statsmodels, pymc, arch, py_vollib, zipline-reloaded, PyPortfolioOpt, etc.
- The usual suspects are included, numpy, pandas, sci-py, scikit-learn, yellowbricks, shap, optuna.
- ib_async for Interactive Broker connectivity. Works well with IB Gateway docker image. ib_insync has been invaluable all this time, RIP Ewald.
- Includes all major Python packages for statistical and time series analysis, see requirements. For an extensive list check list installed packages section.
- Zipline-reloaded, pyfolio-reloaded and alphalens-reloaded.
- ib_fundamental for IBKR fundamental data.
- You can install it as a python package, just
pip install -U jupyter-quant
- Designed for ephemeral containers. Relevant data for your environment will survive your container.
- Optimized for size, it's a 2GB image vs 4GB for jupyter/scipy-notebook
- Includes jedi language server, jupyterlab-lsp, black and isort.
- It does NOT include conda/mamba. All packages are installed with pip under
~/.local/lib/python
. Which should be mounted in a dedicated volume to preserve your environment. - Includes Cython, Numba, bottleneck and numexpr to speed up things
- sudo, so you can install new packages if needed.
- bash and stow, so you can BYODF (bring your dotfiles). Plus common command line utilities like git, less, nano (tiny), jq, ssh, curl, bash completion and others.
- Support for apt cache. If you have other Linux boxes using it can leverage your package cache.
- It does not include a built environment. If you need to install a package that does not provide wheels you can build your wheels, as explained in common tasks
Quick Start
To use jupyter-quant
as a pypi package
see install quant package.
Create a docker-compose.yml
file with this content
services:
jupyter-quant:
image: gnzsnz/jupyter-quant:${IMAGE_VERSION}
environment:
APT_PROXY: ${APT_PROXY:-}
BYODF: ${BYODF:-}
SSH_KEYDIR: ${SSH_KEYDIR:-}
START_SCRIPTS: ${START_SCRIPTS:-}
TZ: ${QUANT_TZ:-}
restart: unless-stopped
ports:
- ${LISTEN_PORT}:8888
volumes:
- quant_conf:/home/gordon/.config
- quant_data:/home/gordon/.local
- ${PWD}/Notebooks:/home/gordon/Notebooks
volumes:
quant_conf:
quant_data:
You can use .env-dist
as your starting point.
cp .env-dist .env
# verify everything looks good
docker compose config
docker compose up
Volumes
The image is designed to work with 3 volumes:
quant_data
- volume for ~/.local folder. It contains caches and all Python packages. This enables to install additional packages through pip.quant_conf
- volume for ~/.config, all config goes here. This includes jupyter, ipython, matplotlib, etc- Bind mount (but you could use a named volume) - volume for all notebooks,
under
~/Notebooks
.
This allows to have ephemeral containers and to keep your notebooks (3), your
config (2) and your additional packages (1). Eventually, you would need to
update the image, in this case, your notebooks (3) can move without issues,
your config (2) should still work but no warranty and your packages in
quant_data
could still be used but you should refresh it with a new image.
Eventually, you would need to refresh (1) and less frequently (2)
Common tasks
Get running server URL
docker exec -it jupyterquant jupyter-server list
Currently running servers:
http://40798f7a604a:8888/?token=
ebf9e870d2aa0ed877590eb83b4d3bbbdfbd55467422a167 :: /home/gordon/Notebooks
or
docker logs -t jupyter-quant 2>&1 | grep '127.0.0.1:8888/lab?token='
You will need to change hostname (40798f7a604a in this case) or 127.0.0.1 by your docker host ip.
Show jupyter config
docker exec -it jupyter-quant jupyter-server --show-config
Set password
docker exec -it jupyter-quant jupyter-server password
Get command line help
docker exec -it jupyter-quant jupyter-server --help
docker exec -it jupyter-quant jupyter-lab --help
List installed packages
docker exec -it jupyter-quant pip list
# outdated packages
docker exec -it jupyter-quant pip list -o
Pass parameters to jupyter-lab
docker run -it --rm gnzsnz/jupyter-quant --core-mode
docker run -it --rm gnzsnz/jupyter-quant --show-config-json
Run a command in the container
docker run -it --rm gnzsnz/jupyter-quant bash
Build wheels outside the container
Build wheels outside the container and import wheels into the container
# make sure python version match .env-dist
docker run -it --rm -v $PWD/wheels:/wheels python:3.11 bash
pip wheel --no-cache-dir --wheel-dir /wheels numpy
This will build wheels for numpy (or any other package that you need) and save
the file in $PWD/wheels
. Then you can copy the wheels in your notebook mount
(3 above) and install it within the container. You can even drag and drop into
Jupyter.
Install your dotfiles
git clone
your dotfiles to Notebook/etc/dotfiles
, set environment variable
BYODF=/home/gordon/Notebook/etc/dotfiles
in your docker-compose.yml
When
the container starts up stow will create links like /home/gordon/.bashrc
Install your SSH keys
You need to define environment variable SSH_KEY_DIR
which should point to a
location with your keys. The suggested place is
SSH_KEYDIR=/home/gordon/Notebooks/etc/ssh
, make sure the director has the
right permissions. Something like chmod 700 Notebooks/etc/ssh
should work.
The entrypoint.sh
script will create a symbolic link pointing to
$SSH_KEYDIR
on /home/gordon/.ssh
.
Within Jupyter's terminal, you can then:
# start agent
eval $(ssh-agent)
# add keys to agent
ssh-add
# open a tunnel
ssh -fNL 4001:localhost:4001 gordon@bastion-ssh
Run scripts at start up
If you define START_SCRIPTS
env variable with a path, all scripts on that
directory will be executed at start up. The sample .env-dist
file contains
a commented line with START_SCRIPTS=/home/gordon/Notebooks/etc/start_scripts
as an example and recommended location.
Files should have a .sh
suffix and should run under bash
. in directory
start_scripts
you will find example scripts to load ssh keys and install python packages.
Install jupyter-quant package
Jupyter-quant is available as a package in pypi. It's a meta-package that pulls all dependencies in it's highest possible version.
Dependencies:
- hdf5 (see below)
- TA-lib see instructions
# ubuntu/debian, see install instructions above for TA-lib
sudo apt-get install libhdf5-dev
# osx
brew install hdf5 ta-lib
Install pypi package.
pip install -U jupyter-quant
Additional options supported are
pip install -U jupyter-quant[bayes] # to install pymc & arviz/graphviz
pip install -U jupyter-quant[sk-util] # to install skfolio & sktime
jupyter-quant
it's a meta-package that pins all it's dependencies versions.
If you need/want to upgrade a dependency you can uninstall jupyter-quant
,
although this can break interdependencies. Or install from git, where it's
updated regularly.
# git install
pip install -U git+https://github.com/quantbelt/jupyter-quant.git
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for jupyter_quant-2409.3-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | f01ca475a89bdfd38d0fceb6b18e241476313bcd004d37c1539121a6db8c022d |
|
MD5 | baef825c080b7bce22d5282f803543ee |
|
BLAKE2b-256 | 00fe05e25fbf44a8ccab74db96cd14c17de883f7fa108907c7bbdd25e93cd869 |