Split a dataset into train, eval and dev smaller subsets

select_datasets

#!/bin/sh

CATEGORIES=(desporto economia)

mkdir -p dev train 'eval'

for c in "${CATEGORIES[@]}"; do
    for i in $(find ./_/$c -type f | sort -R | head); do
        j=$(basename $i)
        mv $i dev/${c}_${j}
    done
done

for c in "${CATEGORIES[@]}"; do
    for i in $(find ./_/$c -type f | sort -R | head -n 2000); do
        j=$(basename $i)
        mv $i train/${c}_${j}
    done
done

for c in "${CATEGORIES[@]}"; do
    for i in $(find ./_/$c -type f | sort -R | head -n 700); do
        j=$(basename $i)
        mv $i 'eval'/${c}_${j}
    done
done