.travis.yml

# As a general rule, the latest release of Keras is compatible with the latest
# release of TensorFlow as well as the previous release of TensorFlow (in most
# cases, it is actually compatible with several prior TF releases, but that is
# not guaranteed).
#
# -- https://github.com/keras-team/keras/issues/10440#issuecomment-397428200

language: python

env:
  - TENSORFLOW='tensorflow' KERAS='keras'

matrix:
  include:
    - os: linux
      python: 3.5
    - os: linux
      python: 3.6
      ###########
    - os: linux
      python: 3.6
      if: branch = master
      env: TENSORFLOW='tensorflow<1.12'   KERAS='keras'
    - os: linux
      python: 3.6
      if: branch = master
      env: TENSORFLOW='tensorflow<1.11'   KERAS='keras'
    - os: linux
      python: 3.6
      if: branch = master
      env: TENSORFLOW='tensorflow<1.10'   KERAS='keras==2.2.2'
    - os: linux
      python: 3.6
      if: branch = master
      # env: TENSORFLOW='tensorflow<1.9'    KERAS='keras==2.2.0' # causes segmentation fault, why?
      env: TENSORFLOW='tensorflow<1.8'    KERAS='keras==2.2.0'
    - os: linux
      python: 3.6
      if: branch = master
      env: TENSORFLOW='tensorflow<1.8'    KERAS='keras==2.1.6'
    - os: linux
      python: 3.6
      if: branch = master
      # lowest supported keras version, last tensorflow release that supports CUDA 8
      env: TENSORFLOW='tensorflow==1.4.1' KERAS='keras==2.1.6'
      ###########

install:
  - pip install $TENSORFLOW $KERAS
  - pip install .

script:
  - pytest -v --durations=50