portable-ai-runner / modules /e2e-microtexts /src /e2e_dbap_train_and_predict.bash

CORTEX vault: module e2e-microtexts (active-mirror)

4f77642 verified about 1 month ago

2.89 kB

# Using: $ bash e2e_dbap_train_and_predict.bash --lang {en|ru} --train_data {en|ru|en_aug|ru_aug}

# Parsing parameters
lang=${lang:-"en"}
train_data=${train_data:-"en_aug"}
while [ $# -gt 0 ]; do
   if [[ $1 == *"--"* ]]; then
        param="${1/--/}"
        declare $param="$2"
        echo $1 $2
   fi
   shift
done

crossval ()
{
    LANG=$lang
    TRAIN_DATA=$train_data

    # Prepare directory
    DIRECTORY=e2e_dbap_${TRAIN_DATA}

    # Prepare configs for K-fold validation
    CONFIGFILE=configs/e2e_dbap_${LANG}_fold_0.jsonnet
    rm configs/e2e_dbap_${LANG}_fold*
    cp initial_configs/e2e_dbap_${LANG}_fold_0.jsonnet $CONFIGFILE
    echo CONFIGFILE $CONFIGFILE

    K=1
    python utils/make_k_copies.py --filename $CONFIGFILE --k $K

    # Default training
    rm -r $DIRECTORY
    mkdir $DIRECTORY
    for (( FOLD=0; FOLD<$K; FOLD++ ))
    do
        TRAIN_DATA=${TRAIN_DATA} allennlp train -s ${DIRECTORY}/fold_${FOLD} configs/e2e_dbap_${LANG}_fold_${FOLD}.jsonnet
        allennlp predict --use-dataset-reader --silent \
                          --output-file ${DIRECTORY}/fold_${FOLD}/predictions_test.json ${DIRECTORY}/fold_${FOLD}/model.tar.gz \
                          data/conll_cv_rst_end2end/${LANG}/test_fold_${FOLD}.conll \
                          --include-package dependency_parser
    done

    # Metrics report
    python utils/latex_report_line.py --model_name ${DIRECTORY} >> ${DIRECTORY}/latex_report.tex
    # Weights illustration (not for BAP)
    # python utils/crossval/rst_weights_plot.py --model_name ${DIRECTORY} --k ${K}
}

crossval_tune ()
{
    LANG=$lang
    TRAIN_DATA=$train_data

    # Prepare directory
    DIRECTORY=e2e_dbap_${TRAIN_DATA}

    # Prepare configs for K-fold validation
    CONFIGFILE=configs/e2e_dbap_${LANG}_fold_0.jsonnet
    rm configs/e2e_dbap_${LANG}_fold*
    cp initial_configs/e2e_dbap_${LANG}_fold_0.jsonnet $CONFIGFILE
    echo CONFIGFILE $CONFIGFILE

    K=10
    python utils/make_k_copies.py --filename $CONFIGFILE --k $K

    mkdir parameters_adjustments_results
    # Fill in the parameters for adjustment
    for DROPOUT in 0.0 0.1 0.3 0.4
    do
       rm -r $DIRECTORY
       mkdir $DIRECTORY
       for (( FOLD=0; FOLD<$K; FOLD++ ))
       do
           TRAIN_DATA=${TRAIN_DATA} DROPOUT=${DROPOUT} allennlp train -s ${DIRECTORY}/fold_${FOLD} configs/e2e_dbap_${LANG}_fold_${FOLD}.jsonnet
           allennlp predict --use-dataset-reader --silent \
                            --output-file ${DIRECTORY}/fold_${FOLD}/predictions_test.json ${DIRECTORY}/fold_${FOLD}/model.tar.gz \
                            data/conll_cv_rst_end2end/${LANG}/test_fold_${FOLD}.conll \
                            --include-package dependency_parser
       done
       python utils/latex_report_line.py --model_name ${DIRECTORY} >> parameters_adjustments_results/${DIRECTORY}_DROPOUT_${DROPOUT}.tex
    done
}

crossval

	# Using: $ bash e2e_dbap_train_and_predict.bash --lang {en\|ru} --train_data {en\|ru\|en_aug\|ru_aug}

	# Parsing parameters
	lang=${lang:-"en"}
	train_data=${train_data:-"en_aug"}
	while [ $# -gt 0 ]; do
	if [[ $1 == "--" ]]; then
	param="${1/--/}"
	declare $param="$2"
	echo $1 $2
	fi
	shift
	done

	crossval ()
	{
	LANG=$lang
	TRAIN_DATA=$train_data

	# Prepare directory
	DIRECTORY=e2e_dbap_${TRAIN_DATA}

	# Prepare configs for K-fold validation
	CONFIGFILE=configs/e2e_dbap_${LANG}_fold_0.jsonnet
	rm configs/e2e_dbap_${LANG}_fold*
	cp initial_configs/e2e_dbap_${LANG}_fold_0.jsonnet $CONFIGFILE
	echo CONFIGFILE $CONFIGFILE

	K=1
	python utils/make_k_copies.py --filename $CONFIGFILE --k $K

	# Default training
	rm -r $DIRECTORY
	mkdir $DIRECTORY
	for (( FOLD=0; FOLD<$K; FOLD++ ))
	do
	TRAIN_DATA=${TRAIN_DATA} allennlp train -s ${DIRECTORY}/fold_${FOLD} configs/e2e_dbap_${LANG}_fold_${FOLD}.jsonnet
	allennlp predict --use-dataset-reader --silent \
	--output-file ${DIRECTORY}/fold_${FOLD}/predictions_test.json ${DIRECTORY}/fold_${FOLD}/model.tar.gz \
	data/conll_cv_rst_end2end/${LANG}/test_fold_${FOLD}.conll \
	--include-package dependency_parser
	done

	# Metrics report
	python utils/latex_report_line.py --model_name ${DIRECTORY} >> ${DIRECTORY}/latex_report.tex
	# Weights illustration (not for BAP)
	# python utils/crossval/rst_weights_plot.py --model_name ${DIRECTORY} --k ${K}
	}

	crossval_tune ()
	{
	LANG=$lang
	TRAIN_DATA=$train_data

	# Prepare directory
	DIRECTORY=e2e_dbap_${TRAIN_DATA}

	# Prepare configs for K-fold validation
	CONFIGFILE=configs/e2e_dbap_${LANG}_fold_0.jsonnet
	rm configs/e2e_dbap_${LANG}_fold*
	cp initial_configs/e2e_dbap_${LANG}_fold_0.jsonnet $CONFIGFILE
	echo CONFIGFILE $CONFIGFILE

	K=10
	python utils/make_k_copies.py --filename $CONFIGFILE --k $K

	mkdir parameters_adjustments_results
	# Fill in the parameters for adjustment
	for DROPOUT in 0.0 0.1 0.3 0.4
	do
	rm -r $DIRECTORY
	mkdir $DIRECTORY
	for (( FOLD=0; FOLD<$K; FOLD++ ))
	do
	TRAIN_DATA=${TRAIN_DATA} DROPOUT=${DROPOUT} allennlp train -s ${DIRECTORY}/fold_${FOLD} configs/e2e_dbap_${LANG}_fold_${FOLD}.jsonnet
	allennlp predict --use-dataset-reader --silent \
	--output-file ${DIRECTORY}/fold_${FOLD}/predictions_test.json ${DIRECTORY}/fold_${FOLD}/model.tar.gz \
	data/conll_cv_rst_end2end/${LANG}/test_fold_${FOLD}.conll \
	--include-package dependency_parser
	done
	python utils/latex_report_line.py --model_name ${DIRECTORY} >> parameters_adjustments_results/${DIRECTORY}_DROPOUT_${DROPOUT}.tex
	done
	}

	crossval