Upgrading

While we try to make upgrades from one version of MDBenchmark to another as easy as possible, we are sometimes forced to break things along the way. This page gives some guidelines on how to migrate to specific versions of MDBenchmark, if you have been using a different version before.

To update from a previous MDBenchmark version to a new release, run the following, if you are using conda and your conda environment is called benchmark:

conda update -n benchmark mdbenchmark

and if you use pip, run:

pip install mdbenchmark --upgrade

Refer to our installation guide, if you need help with conda/virtual environments.

Migrating to a newer version

Version 2.0.0

Note

TL;DR: You need to uninstall the unneeded datreant.core, datreant.data and mdsynthesis Python packages. Or just get rid of the current environment, e.g., conda env remove -n benchmark if you are using conda, and create it from scratch.

Starting from version 2.0.0, MDBenchmark migrated to a different dependency, than it used before. While the specialized mdsynthesis package was used in previous version, we migrated to the more general datreant package. This leads to two things:

  1. Previously, every folder representing a specific number of nodes contained a Sim.<uuid>.json file, where <uuid> is a UUID. This file contained all information representing a specific benchmark with its parameters. With the newer version of datreant, the format was changed and now each of these folders contains a .datreant folder instead.
  2. datreant version 1.0 changed its package layout and led to the necessity of uninstalling previous packages. Because we were using a previous datreant version, you as a user will also need to uninstall those packages.

Uninstalling old packages

For proper migration, you need to uninstall the following three packages:

  1. datreant.core
  2. datreant.data
  3. mdsynthesis

If you are using conda and your environment is called benchmark, run the following command:

conda remove -n benchmark datreant.core datreant.data mdsynthesis

You can also decide to fully remove the environment (conda env remove -n benchamrk) and create it from scratch.

If you are using pip simply run:

pip uninstall datreant.core datreant.data mdsynthesis