environment.yaml

name: space-type-recognition
channels:
  - anaconda
  - conda-forge
dependencies:
  - python=3.6
#  - python=3.7
#  - numpy=1.15.4
  - numpy=1.16
  - pandas=0.23.4
# It is strange but MKL package does not support AVX
#  - tensorflow-mkl=1.12.0
#  - anaconda::tensorflow=1.13.1
#  - anaconda::tensorboard=1.13.1
# required to support PyCharm
  - jupyter
  - jupyterlab
  - pydot=1.3.0
  - graphviz=2.40.1
  - matplotlib=3.0.1
  - keras=2.2.4
  - scipy=1.1.0
  - scikit-learn=0.20.1
  - tqdm=4.28.1
  - pip
  - IPython
  - conda-forge::hdbscan=0.8.18
  # Use Conda Forge package because it supports CPU and AVX. Tensorflow-eigen does not support it for some reason and slow
  - conda-forge::tensorflow==1.13.1
  - conda-forge::tensorboard==1.13.1
  - pip:
      - git+https://www.github.com/keras-team/keras-contrib.git@46fcdb9384b3bc9399c651b2b43640aa54098e64
#      - keras
#      - tensorflow>=1.12
#      - tensorboard>=1.12
#      - tensorflow==1.13.1
#      - tensorboard==1.13.1