#!/usr/bin/env sh
set -e
# The following example allows for the MNIST example (using LeNet) to be
# trained using the caffe docker image instead of building from source.
#
# The GPU-enabled version of Caffe can be used, assuming that nvidia-docker
# is installed, and the GPU-enabled Caffe image has been built.
# Setting the GPU environment variable to 1 will enable the use of nvidia-docker.
# e.g.
#   GPU=1 ./examples/mnist/train_lenet_docker.sh [ADDITIONAL_CAFFE_ARGS]
#
# With any arguments following the script being passed directly to caffe
# when training the network.
#
# The steps that are performed by the script are as follows:
# 1. The MNIST data set is downloaded
#    (see data/mnist/get_mnist.sh)
# 2. An LMDB database is created from the downloaded data
#    (see examples/mnist/create_mnist.sh.
# 3. A caffe network based on the LeNet solver is trained.
#    (see examples/mnist/lenet_solver.prototxt)
#
# For each of these, a step is executed to ensure that certain prerequisites
# are available, after which a command that actually performs the work is
# executed.
#
# In order to provide additional flexibility, the following shell (environment)
# variables can be used to control the execution of each of the phases:
#
# DOWNLOAD_DATA: Enable (1) or disable (0) the downloading of the MNIST dataset
# CREATE_LMDB: Enable (1) or disable (0) the creation of the LMDB database
# TRAIN: Enable (1) or disable (0) the training of the LeNet networkd.
#
# As an example, assuming that the data set has been downloaded, and an LMDB
# database created, the following command can be used to train the LeNet
# network with GPU computing enabled.
#
# DOWNLOAD_DATA=0 CREATE_LMDB=0 GPU=1 ./examples/mnist/train_lenet_docker.sh
#


if [ x"$(uname -s)" != x"Linux" ]
then
echo ""
echo "This script is designed to run on Linux."
echo "There may be problems with the way Docker mounts host volumes on other"
echo "systems which will cause the docker commands to fail."
echo ""
read -p "Press [ENTER] to continue..." key
echo ""
fi


# Check if GPU mode has been enabled and set the docker executable accordingly
if [ ${GPU:-0} -eq 1 ]
then
DOCKER_CMD=nvidia-docker
IMAGE=caffe:gpu
else
DOCKER_CMD=docker
IMAGE=caffe:cpu
fi
echo "Using $DOCKER_CMD to launch $IMAGE"

# On non-Linux systems, the Docker host is typically a virtual machine.
# This means that the user and group id's may be different.
# On OS X, for example, the user and group are 1000 and 50, respectively.
if [ x"$(uname -s)" != x"Linux" ]
then
CUID=1000
CGID=50
else
CUID=$(id -u)
CGID=$(id -g)
fi

# Define some helper variables to make the running of the actual docker
# commands less verbose.
# Note:
#   -u $CUID:$CGID             runs the docker image as the current user to ensure
#                              that the file permissions are compatible with the
#                              host system. The variables CUID and CGID have been
#                              set above depending on the host operating system.
#   --volume $(pwd):/workspace mounts the current directory as the docker volume
#                              /workspace
#   --workdir /workspace       Ensures that the docker container starts in the right
#                              working directory
DOCKER_OPTIONS="--rm -ti -u $CUID:$CGID --volume=$(pwd):/workspace --workdir=/workspace"
DOCKER_RUN="$DOCKER_CMD run $DOCKER_OPTIONS $IMAGE"

# Download the data
if [ ${DOWNLOAD_DATA:-1} -eq 1 ]
then
$DOCKER_RUN bash -c "mkdir -p ./data/mnist;
                     cp -ru \$CAFFE_ROOT/data/mnist/get_mnist.sh ./data/mnist/"
$DOCKER_RUN ./data/mnist/get_mnist.sh
fi

# Create the LMDB database
if [ ${CREATE_LMDB:-1} -eq 1 ]
then
$DOCKER_RUN bash -c "mkdir -p ./examples/mnist;
                     cp -ru \$CAFFE_ROOT/examples/mnist/create_mnist.sh ./examples/mnist/;
                     sed -i s#BUILD=build#BUILD=\$CAFFE_ROOT/build## ./examples/mnist/create_mnist.sh"
$DOCKER_RUN ./examples/mnist/create_mnist.sh
fi

# Train the network
if [ ${TRAIN:-1} -eq 1 ]
then
$DOCKER_RUN bash -c "cp \$CAFFE_ROOT/examples/mnist/lenet_solver.prototxt ./examples/mnist/;
                     cp \$CAFFE_ROOT/examples/mnist/lenet_train_test.prototxt ./examples/mnist/"
    # Ensure that the solver_mode is compatible with the desired GPU mode.
    if [ ${GPU:-0} -eq 0 ]
    then
    $DOCKER_RUN sed -i 's#solver_mode: GPU#solver_mode: CPU##' ./examples/mnist/lenet_solver.prototxt
    fi
$DOCKER_RUN caffe train --solver=examples/mnist/lenet_solver.prototxt $*
fi