mkconda is a scientific computing stack of useful Python and R scientific computing conda packages for GPU servers with Intel CPUS and NVIDIA GPUS, among them:

• mkpy - Kutas Lab data interchange utilities
• spudtr - some useful Pandas data transforms
• fitgrid - multichannel regression modeling
• MNE Python - EEG/MEG data analysis
• jupyterlab - Jupyter notebook desktop + Python and R kernels
• spyder - Python IDE
• tidyverse - data wrangling utilities in R
• pytorch - tensor library for deep learning
• numba - CPU and GPU Python acclerators
• Intel Math Kernel (MKL) library - math libraries optimized for Intel CPUs

DROPPED PENDING FIX OF cudatoolkit-dev 11.4 MAMBA INSTALL

• cudatoolkit-dev
• rapids - NVIDIA Python GPU accelerators

Installation

Users

In a bash terminal window, create a new named mkconda working environment like so. The order of conda channels is important and if you copy the command make sure the backslash is the last character on the line, no trailing whitespace.

mamba create --name mkconda_072221 mkconda \
    --strict-channel-priority -c conda-forge -c defaults -c ejolly -c kutaslab

Then you can activate the environment as usual.

conda activate mkconda_072221

The mkconda environment comes with Python, R, and over 500 packages including matplotlib, seaborn. To see which, install a mkconda working environment and then run conda list.

If you need additional packages, mamba (or conda) install them into the working environment like so:

(mkconda_072221) $ conda list
(mkconda_072221) $ mamba install --strict-channel-priority -c conda-forge -c defaults package_a package_b ...

Development mode

To run a mkconda member package in development mode, create a working mkconda environment, navigate to where you want the package source code to reside, and install the development branch from source into the working environment in editable ("development") mode like so:

(base) $ mamba create --name mkconda_072221 --strict-channel-priority -c conda-forge -c defaults -c ejolly -c kutaslab
(base) $ conda activate mkconda_072221
(mkconda_072221) $ cd path/to/local_source_dirs
(mkconda_072221) $ git clone https://github.com/the_package --single-branch -b the_branch
(mkconda_072221) $ cd the_package
(mkconda_072221) $ git checkout the_branch
(mkconda_072221) $ pip install --no-deps -e .
(mkconda_072221) $