gistfile1.sh

#!/bin/bash

# copy and archive latest reports/scores from the cluster
# filenames are {corpus}-{date}__{machine}__{logword}, eg
#
# scores-TRAINING-2015-02-05__eric-laptop__scikit.txt
# scores-TRAINING-2015-02-07__eric-laptop__ptb-sent.txt
# scores-TRAINING-2015-02-09__cluster__hello.txt
# scores-TRAINING-2015-02-12__cluster__parallel.txt
# scores-TRAINING-2015-02-16__cluster__intra-mst.txt
# scores-TRAINING-2015-02-16__eric-laptop__parallel.txt
# scores-TRAINING-2015-02-17__cluster__skperceptron.txt

# cwd-proofing
ZERO_DIR=$(dirname "$0")
pushd "$ZERO_DIR" > /dev/null
SCRIPT_DIR=$PWD
popd > /dev/null
cd "$SCRIPT_DIR"

# force user to give us a log word
if [ $# -ne 2 ]; then
    echo >&2 "Usage: $0 double|training log-word"
    exit 1
fi

if [ "$1" == "double" ]; then
    CORPUS=RSTtrees-WSJ-double-1.0
else
    CORPUS=TRAINING
fi
LOGWORD=$2

# copy from the cluster (preserve hardlinks too, eh?)
TODAY=$(date +%Y-%m-%d)
SUFFIX=${CORPUS}-${TODAY}__cluster__${LOGWORD}
rsync -avH cluster:irit-rst-dt/TMP/latest/eval-current/reports-"$CORPUS" reports/"reports-${SUFFIX}"
cp "reports/reports-${SUFFIX}"/scores.txt "scores/scores-${SUFFIX}".txt

cd reports
tar cjvf "reports-${SUFFIX}.tar.bz2" "reports-${SUFFIX}"
cd ..

tar cjvf "all-scores-${TODAY}.tar.bz2" scores