environment.yml

# Notes:
# - This file only lists packages that elektronn3 *directly* depends on.
#   Transitive dependencies will be resolved by conda.
# - Dependencies are not pinned to specific versions. The minimum versions
#   listed below are the ones that are known to work. If some package causes
#   problems, just downgrade to the minimum version that is documented here.
# - If you encounter problems with CUDA versions (unwanted up-/downgrades,
#   incompatible versions), make sure to pin the cudatoolkit package to
#   a version that is known to work with your GPU driver, for example if
#   you want to stay on CUDA 10.1, run the following commands:
#       conda config --add pinned_packages 'cudatoolkit=10.1'
#       conda update --all

name: elektronn3
channels:
  - pytorch
  - conda-forge
  - defaults
dependencies:
  - python >=3.6
  - colorlog >=4.0
  - h5py >=2.9
  - ipython >=7
  - matplotlib >=3.1
  - numba >=0.45
  - tbb
  - tqdm >=4.23
  - numpy >=1.17
  - scikit-learn >=0.21
  - scikit-image >=0.15
  - scipy >=1.3
  - tensorboardx >=2.0
  - pytorch >=1.6  # pytorch-cpu also works, but it's not recommended.

  # Only required for running a tensorboard server:
  - tensorboard >=2.1.1

  # For docs:
  - sphinx >=2.2
  - sphinx_rtd_theme >=0.4
  - sphinx-autodoc-typehints >=1.6