blob: ba6c50c13cd9cf5dc20307fc3aba7e8ac5323721 [file] [log] [blame]
#!/usr/bin/env perl
use strict;
use warnings;
use Benchmark qw!:hireswallclock :all!;
use Data::Dumper;
use POSIX 'round';
our @ARGV;
# The first parameter is the number of iterations.
my $FILE = 'effi-1x-utf8.txt';
system 'gzip -dkf ./corpus/' . $FILE . '.gz';
my $iter = 1;
if ($ARGV[0]) {
$iter = $ARGV[0] + 0;
};
my $temp_dir = '/euralex/batches/';
mkdir($temp_dir);
# Read file
my $single = '';
if (open(my $f, '<', './corpus/' . $FILE)) {
while(!eof($f)) {
$single .= <$f>;
};
close($f);
};
warn "Concatenate file\n";
my $count = 300;
# Concat $count times
my $data = '';
foreach (1..$count) {
$data .= $single;
};
# save file in docker
open(X, '>', $temp_dir . 'effi-' . $count . 'x-utf8.txt') or die $!;
print X $data;
close(X);
warn "Calculate offsets\n";
# Calculate offsets with datok (just because) for segmenting
my $offsets = `cat ${temp_dir}effi-${count}x-utf8.txt | ./Datok/datok tokenize -t ./Datok/testdata/tokenizer.matok --no-tokens --no-sentences -p -`;
my @offsets = split / /, $offsets;
unlink './corpus/effi-' . $count . 'x-utf8.txt';
$FILE = undef;
my $models = {
'wc' => sub {
system 'wc -w ./corpus/'.$FILE.' > /dev/null';
},
'SoMaJo' => sub {
system 'somajo-tokenizer '.$temp_dir.$FILE.' --split_sentences > /dev/null';
},
'SoMaJo_p2' => sub {
system 'somajo-tokenizer '.$temp_dir.$FILE.' --parallel=2 --split_sentences > /dev/null';
},
'SoMaJo_p4' => sub {
system 'somajo-tokenizer '.$temp_dir.$FILE.' --parallel=4 --split_sentences > /dev/null';
},
'SoMaJo_p8' => sub {
system 'somajo-tokenizer '.$temp_dir.$FILE.' --parallel=8 --split_sentences > /dev/null';
},
'Datok_matok' => sub {
system 'cat '.$temp_dir.$FILE.' | ./Datok/datok tokenize -t ./Datok/testdata/tokenizer.matok - > /dev/null'
},
'Datok_datok' => sub {
system 'cat '.$temp_dir.$FILE.' | ./Datok/datok tokenize -t ./Datok/testdata/tokenizer.datok - > /dev/null'
},
'OpenNLP_Simple' => sub {
system 'cat '.$temp_dir.$FILE.' | ./opennlp/bin/opennlp SimpleTokenizer > /dev/null';
},
'OpenNLP_Tokenizer_de-ud-gsd' => sub {
system 'cat '.$temp_dir.$FILE.' | ./opennlp/bin/opennlp TokenizerME ./opennlp/models/opennlp-de-ud-gsd-tokens-1.0-1.9.3.bin > /dev/null';
},
'OpenNLP_Sentence_de-ud-gsd' => sub {
system 'cat '.$temp_dir.$FILE.' | ./opennlp/bin/opennlp SentenceDetector ./opennlp/models/opennlp-de-ud-gsd-sentence-1.0-1.9.3.bin > /dev/null';
},
'TreeTagger' => sub {
system 'cat '.$temp_dir.$FILE.' | perl ./treetagger/cmd/utf8-tokenize.perl -a ./treetagger/lib/german-abbreviations > /dev/null';
},
'deep-eos_bi-lstm-de' => sub {
system 'python3 ./deep-eos/main.py --input-file '.$temp_dir.$FILE.' --model-filename ./deep-eos/bi-lstm-de.model --vocab-filename ./deep-eos/bi-lstm-de.vocab --eos-marker "§" tag > /dev/null';
},
'deep-eos_cnn-de' => sub {
system 'python3 ./deep-eos/main.py --input-file '.$temp_dir.$FILE.' --model-filename ./deep-eos/cnn-de.model --vocab-filename ./deep-eos/cnn-de.vocab --eos-marker "§" tag > /dev/null';
},
'deep-eos_lstm-de' => sub {
system 'python3 ./deep-eos/main.py --input-file '.$temp_dir.$FILE.' --model-filename ./deep-eos/lstm-de.model --vocab-filename ./deep-eos/lstm-de.vocab --eos-marker "§" tag > /dev/null';
},
'JTok' => sub {
chdir '/euralex/JTok/bin';
system 'sh tokenize '.$temp_dir.$FILE.' de > /dev/null';
chdir '/euralex';
},
'KorAP-Tokenizer' => sub {
system 'cat '.$temp_dir.$FILE.' | java -jar ./KorAP-Tokenizer/KorAP-Tokenizer.jar -l de -s > /dev/null'
},
Syntok_tokenizer => sub {
system 'python3 -m syntok.tokenizer '.$temp_dir.$FILE.' > /dev/null';
},
Syntok_segmenter => sub {
system 'python3 -m syntok.segmenter '.$temp_dir.$FILE.' > /dev/null';
},
Waste => sub {
system 'cat '.$temp_dir.$FILE.' | waste -N -v0 --rcfile=./Waste/waste.rc > /dev/null';
},
nnsplit => sub {
system './nnsplit/nnsplit_bench '.$temp_dir.$FILE.' > /dev/null'
},
elephant => sub {
system './elephant-wrapper/bin/tokenize.sh -i '.$temp_dir.$FILE.' UD_German > /dev/null'
},
cutter => sub {
system 'python3 ./cutter/cutter.py nosent '.$temp_dir.$FILE.' > /dev/null'
},
blingfire_tok => sub {
system 'python3 ./blingfire/blingfire_tok.py '.$temp_dir.$FILE.' > /dev/null'
},
blingfire_sent => sub {
system 'python3 ./blingfire/blingfire_sent.py '.$temp_dir.$FILE.' > /dev/null'
},
spacy_tok => sub {
system 'python3 ./spacy/spacy_tok.py '.$temp_dir.$FILE.' > /dev/null'
},
spacy_dep => sub {
system 'python3 ./spacy/spacy_sent.py dep '.$temp_dir.$FILE.' > /dev/null'
},
spacy_stat => sub {
system 'python3 ./spacy/spacy_sent.py stat '.$temp_dir.$FILE.' > /dev/null'
},
spacy_sentencizer => sub {
system 'python3 ./spacy/spacy_sent.py sentencizer '.$temp_dir.$FILE.' > /dev/null'
},
Stanford => sub {
system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
'-props german -annotators tokenize,ssplit,mwt -tokenize.language=german -file '.$temp_dir . $FILE
},
Stanford_t2 => sub {
system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
'-props german -annotators tokenize,ssplit,mwt -tokenize.language=german -threads=2 -file '.$temp_dir . $FILE
},
Stanford_t4 => sub {
system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
'-props german -annotators tokenize,ssplit,mwt -tokenize.language=german -threads=4 -file '.$temp_dir . $FILE
},
Stanford_t8 => sub {
system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
'-props german -annotators tokenize,ssplit,mwt -tokenize.language=german -threads=8 -file '.$temp_dir . $FILE
},
Stanford_tokonly => sub {
system 'CLASSPATH=/euralex/stanford-corenlp-4.4.0/* java edu.stanford.nlp.pipeline.StanfordCoreNLP ' .
'-props german -annotators tokenize -tokenize.language=german -file '.$temp_dir . $FILE;
unlink $temp_dir . $FILE . '.out';
},
};
#delete $models->{'wc'};
#delete $models->{'SoMaJo'};
#delete $models->{'SoMaJo_p2'};
#delete $models->{'SoMaJo_p4'};
#delete $models->{'SoMaJo_p8'};
#delete $models->{'Datok_matok'};
#delete $models->{'Datok_datok'};
#delete $models->{'OpenNLP_Simple'};
#delete $models->{'OpenNLP_Tokenizer_de-ud-gsd'};
#delete $models->{'OpenNLP_Sentence_de-ud-gsd'};
#delete $models->{'TreeTagger'};
#delete $models->{'deep-eos_bi-lstm-de'};
#delete $models->{'deep-eos_cnn-de'};
#delete $models->{'deep-eos_lstm-de'};
#delete $models->{'JTok'};
#delete $models->{'KorAP-Tokenizer'};
#delete $models->{'Syntok_tokenizer'};
#delete $models->{'Syntok_segmenter'};
#delete $models->{'Waste'};
#delete $models->{'nnsplit'};
#delete $models->{'elephant'};
#delete $models->{'Stanford'};
#delete $models->{'Stanford_t2'};
#delete $models->{'Stanford_t4'};
#delete $models->{'Stanford_t8'};
#delete $models->{'Stanford_tokonly'};
#delete $models->{'cutter'};
#delete $models->{'spacy_tok'};
#delete $models->{'spacy_sentencizer'};
#delete $models->{'spacy_dep'};
#delete $models->{'spacy_stat'};
#delete $models->{'blingfire_tok'};
#delete $models->{'blingfire_sent'};
my $t0 = Benchmark->new;
# Get some batch files with
# 1000 tokens
# 2000 tokens
# 4000 tokens
# 8000 tokens
# ...
# 8192000 tokens
for (my $x = 0 ;$x < 100;$x++) {
my $i = (2**$x) * 1000;
warn "Create $i batch\n";
last if ($i*2)+1 > scalar(@offsets);
my $off = $offsets[($i*2)+1];
last unless $off;
$FILE = 'effi-batch-' . $i . '.txt';
open(Y, '>', $temp_dir . $FILE);
print Y substr($data, 0, $off);
close(Y);
my $cmp = timethese($iter => $models);
print "\n----------------------------------\n";
foreach my $tool (sort keys %$cmp) {
my $seconds_per_run = $cmp->{$tool}->[0] / $cmp->{$tool}->[5];
my $tokens_per_msecond = ($i / $seconds_per_run) / 1000;
print $tool,"\t",$i,"\t", $seconds_per_run, "\t", sprintf("%.2f", $seconds_per_run), "\t", $tokens_per_msecond, "\t", sprintf("%.2f", $tokens_per_msecond), "\n";
};
print "\n----------------------------------\n";
unlink $temp_dir . $FILE;
};
print "Benchmarking took: ", timestr(timediff(Benchmark->new, $t0)), "\n";