ImperialCollegeLondon / recode-perceptions

Convolutional Neural Networks exemplar for ReCoDE
https://imperialcollegelondon.github.io/recode-perceptions/
BSD 3-Clause "New" or "Revised" License
0 stars 2 forks source link

Slow to download and unzip the data #26

Closed AdrianDAlessandro closed 2 years ago

AdrianDAlessandro commented 2 years ago

Because of how large the data is and the time taken, I suggest writing a job script to download and unzip it. Something like this:

#PBS -l walltime=08:00:00
#PBS -l select=1:ncpus=1:mem=1gb

cd $PBS_O_WORKDIR

wget -O input/places365standard_easyformat.tar http://data.csail.mit.edu/places/places365/places365standard_easyformat.tar

cd input
tar -xvf places365standard_easyformat.tar

GLOBIGNORE=$(paste -s -d : keep.txt)
rm -rf places365_standard/train/*
rm -rf places365_standard/val/*
unset GLOBIGNORE
AdrianDAlessandro commented 2 years ago

Also to note: The GLOBIGNORE environment variable does not work for me in the HPC. So it unzips everything and then deletes it all