Skip to main content

Access to ECCO

The Economics Compute Cluster (ECCO) has migrated to the BioHPC environment, and accounts are not handled on this website anymore. See the BioHPC User Guide.

Latest Tweets

[ECCO] [SDS] Update to Anaconda Python (2.1.0 -> 4.0.0)

Print Friendly, PDF & Email

We have updated the Anaconda Python environment from 2.1.0 to Anaconda 4.0.0.

The key python changes are

  • python 2.7.9 -> 2.7.11
  • python 3.3.4 -> 3.3.5
  • ipython notebook -> jupyter notebook

The full Anaconda changelog (not fully synchronized with what ECCO has) can be found at https://docs.continuum.io/anaconda/changelog.  For what it's worth, environments were created prior to the upgrade:

$ conda info --envs
# conda environments:
#
mkl2.7                   /cac/contrib/anaconda/envs/mkl2.7
mkl3.3                   /cac/contrib/anaconda/envs/mkl3.3
py3_20161111             /cac/contrib/anaconda/envs/py3_20161111
python3.3                /cac/contrib/anaconda/envs/python3.3
root20161111             /cac/contrib/anaconda/envs/root20161111
root                  *  /cac/contrib/anaconda

where the py3_20161111 and root20161111 are the environments prior to the upgrade. As of 2016-11-11, the following python versions are installed in there:

  • root20161111: Python 2.7.9 :: Anaconda 2.1.0 (64-bit)
  • py3_20161111: Python 3.3.5 :: Anaconda 2.1.0 (64-bit)

The mkl2.7, mkl3.3, root, and python3.3 environments have been fully upgraded.

Note: In the core environments, the MKL high-performance libraries no longer require that you obtain a personal academic license. You may have to remove older licenses. See https://docs.continuum.io/mkl-optimizations/ for more information, and let us know if you run into any issues.

Note: For convenience, the core 3.3 environment can be called two ways:

  • by loading the appropriate module:
    module load python/anaconda/3.3
  • by loading the default anaconda module, and then activating the py3.3 environment:
    module load python/anaconda
    source activate py3.3

All other Python environments must be loaded using the second method.

Note: For replicability, you may want to always work with your own local environment. You can fix a particular environment by cloning it into your private Anaconda space (by default, this is your home directory, but it can be put into a group-readable directory as well): conda create -n MyClone --clone=root