lintenv.sh

#! /bin/bash

################################################################################
# lintenv.sh
#
# Set up a system-wide Anaconda virtualenv for running TensorFlow's
# tensorflow/tools/ci_build/ci_sanity.sh script outside of a Docker container. 
# Replicates the Python environment in the TensorFlow Dockerfiles under
# tensorflow/tools/ci_build.
#
# Requires that conda be installed and set up for calling from bash scripts.
#
# Also requires that you set the environment variable CONDA_HOME to the
# location of the root of your anaconda/miniconda distribution.
################################################################################

# The TensorFlow CPU dockerfile runs Python 3.5.2
PYTHON_VERSION=3.5.2

ENV_NAME="tflint"

############################
# HACK ALERT *** HACK ALERT 
# The friendly folks at Anaconda thought it would be a good idea to make the
# "conda" command a shell function. 
# See https://github.com/conda/conda/issues/7126
# The following workaround will probably be fragile.
if [ -z "$CONDA_HOME" ]
then 
    echo "Error: CONDA_HOME not set"
    exit
fi
. ${CONDA_HOME}/etc/profile.d/conda.sh
# END HACK
############################


# Remove the detrius of any previous runs of this script
conda env remove -n ${ENV_NAME}


# Create initial env, leaving it as empty as possible to prevent version
# conflicts with the specific libraries that the TF test env needs.
conda create -y --name ${ENV_NAME} \
    python=${PYTHON_VERSION} 

conda activate ${ENV_NAME}

# For now we pip install everything so as to tightly control version numbers
# See tensorflow/tools/ci_build/install/install_pip_packages.sh for more
# information.
# The original code set up both Python 2 and 3, but given the looming EOL date
# for Python 2, we only do Python 3 here.

# Start out by downgrading pip to the same version used in the TF script.
easy_install -U pip==18.1

# Install packages one at a time to avoid a long pause
pip install --upgrade wheel==0.31.1
pip install --upgrade setuptools==39.1.0
pip install --upgrade virtualenv
pip install --upgrade six==1.12.0
pip install --upgrade future>=0.17.1
pip install --upgrade absl-py
pip install --upgrade werkzeug==0.11.10
pip install --upgrade bleach==2.0.0
pip install --upgrade markdown==2.6.8
pip install --upgrade protobuf==3.6.1
pip install --upgrade numpy==1.14.5
pip install --upgrade scipy==1.1.0
pip install --upgrade scikit-learn==0.18.1
pip install --upgrade pandas==0.19.2
pip install --upgrade psutil
pip install --upgrade py-cpuinfo
pip install --upgrade pylint==1.6.4
pip install --upgrade pycodestyle
pip install --upgrade portpicker
pip install --upgrade grpcio
pip install --upgrade astor
pip install --upgrade gast
pip install --upgrade termcolor
pip install --upgrade h5py==2.8.0
pip install --upgrade argparse 

# These dependencies are installed separately because they will change
# frequently and need to be installed without dependent packages
# frequently and require --no-deps argument.
pip install keras_applications==1.0.6 --no-deps
pip install keras_preprocessing==1.0.5 --no-deps
pip install tf-estimator-nightly --no-deps


conda deactivate
    
        
echo << EOM
Anaconda virtualenv installed under name '${ENV_NAME}'
Run \"conda activate ${ENV_NAME}\" before running ./configure
EOM