From 671c21451542e2dd20e45b4033d44d8e8735f87b Mon Sep 17 00:00:00 2001 From: Chris Dyer Date: Thu, 3 Dec 2009 16:33:55 -0500 Subject: initial check in --- vest/dist-vest.pl | 642 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 642 insertions(+) create mode 100755 vest/dist-vest.pl (limited to 'vest/dist-vest.pl') diff --git a/vest/dist-vest.pl b/vest/dist-vest.pl new file mode 100755 index 00000000..5528838c --- /dev/null +++ b/vest/dist-vest.pl @@ -0,0 +1,642 @@ +#!/usr/bin/env perl + +use Getopt::Long; +use IPC::Open2; +use strict; +use POSIX ":sys_wait_h"; + +my $mydir = `dirname $0`; +chomp $mydir; +# Default settings +my $srcFile = "/fs/cliplab/mteval/Evaluation/Chinese-English/mt03.src.txt"; +my $refFiles = "/fs/cliplab/mteval/Evaluation/Chinese-English/mt03.ref.txt.*"; +my $bin_dir = "/fs/clip-software/cdec/bin"; +$bin_dir = "/Users/redpony/cdyer-svn-root/cdec/vest/bin_dir"; +die "Bin directory $bin_dir missing/inaccessible" unless -d $bin_dir; +my $FAST_SCORE="$bin_dir/fast_score"; +die "Can't find $FAST_SCORE" unless -x $FAST_SCORE; +my $MAPINPUT = "$bin_dir/mr_vest_generate_mapper_input"; +my $MAPPER = "$bin_dir/mr_vest_map"; +my $REDUCER = "$bin_dir/mr_vest_reduce"; +my $SCORER = $FAST_SCORE; +die "Can't find $MAPPER" unless -x $MAPPER; +my $forestUnion = "$bin_dir/union_forests"; +die "Can't find $forestUnion" unless -x $forestUnion; +my $cdec = "$bin_dir/cdec"; +die "Can't find decoder in $cdec" unless -x $cdec; +my $decoder = $cdec; +my $lines_per_mapper = 440; +my $rand_directions = 10; +my $iteration = 1; +my $run_local = 0; +my $best_weights; +my $max_iterations = 40; +my $optimization_iters = 6; +my $num_rand_points = 20; +my $mert_nodes = join(" ", grep(/^c\d\d$/, split(/\n/, `pbsnodes -a`))); # "1 1 1 1 1" fails due to file staging conflicts +my $decode_nodes = "1 1 1 1 1 1 1 1 1 1 1 1 1 1 1"; # start 15 jobs +my $pmem = "3g"; +my $disable_clean = 0; +my %seen_weights; +my $normalize; +my $help = 0; +my $epsilon = 0.0001; +my $interval = 5; +my $dryrun = 0; +my $ranges; +my $restart = 0; +my $metric = "ibm_bleu"; +my $dir; +my $iniFile; +my $weights; +my $initialWeights; +my $decoderOpt; + +# Process command-line options +Getopt::Long::Configure("no_auto_abbrev"); +if (GetOptions( + "decoder=s" => \$decoderOpt, + "decode-nodes=s" => \$decode_nodes, + "dont-clean" => \$disable_clean, + "dry-run" => \$dryrun, + "epsilon" => \$epsilon, + "help" => \$help, + "interval" => \$interval, + "iteration=i" => \$iteration, + "local" => \$run_local, + "max-iterations=i" => \$max_iterations, + "mert-nodes=s" => \$mert_nodes, + "normalize=s" => \$normalize, + "pmem=s" => \$pmem, + "ranges=s" => \$ranges, + "rand-directions=i" => \$rand_directions, + "ref-files=s" => \$refFiles, + "metric=s" => \$metric, + "restart" => \$restart, + "source-file=s" => \$srcFile, + "weights=s" => \$initialWeights, + "workdir=s" => \$dir +) == 0 || @ARGV!=1 || $help) { + print_help(); + exit; +} + +if ($metric =~ /^(combi|ter)$/i) { + $lines_per_mapper = 40; +} + +($iniFile) = @ARGV; + +sub write_config; +sub enseg; +sub print_help; + +my $nodelist; +my $host =`hostname`; chomp $host; +my $bleu; +my $best_bleu = 0.0; +my $interval_count = 0; +my $epsilon_bleu = $best_bleu + $epsilon; +my $logfile; +my $projected_score; + +my $refs_comma_sep = get_comma_sep_refs($refFiles); + +unless ($dir){ + $dir = "vest"; +} +unless ($dir =~ /^\//){ # convert relative path to absolute path + my $basedir = `pwd`; + chomp $basedir; + $dir = "$basedir/$dir"; +} +if ($restart){ + $iniFile = `ls $dir/*.ini`; chomp $iniFile; + unless (-e $iniFile){ + die "ERROR: Could not find ini file in $dir to restart\n"; + } + $logfile = "$dir/mert.log"; + open(LOGFILE, ">>$logfile"); + print LOGFILE "RESTARTING STOPPED OPTIMIZATION\n\n"; + + # figure out best weights so far and iteration number + open(LOG, "$dir/mert.log"); + my $wi = 0; + while (my $line = ){ + chomp $line; + if ($line =~ /ITERATION (\d+)/) { + $iteration = $1; + } + } + + $iteration = $wi + 1; +} + +if ($decoderOpt){ $decoder = $decoderOpt; } + + +# Initializations and helper functions +srand; + +my @childpids = (); +my @cleanupcmds = (); + +sub cleanup { + print STDERR "Cleanup...\n"; + for my $pid (@childpids){ `kill $pid`; } + for my $cmd (@cleanupcmds){`$cmd`; } + exit 1; +}; +$SIG{INT} = "cleanup"; +$SIG{TERM} = "cleanup"; +$SIG{HUP} = "cleanup"; + +my $decoderBase = `basename $decoder`; chomp $decoderBase; +my $newIniFile = "$dir/$decoderBase.ini"; +my $parallelize = "$mydir/parallelize.pl"; +my $inputFileName = "$dir/input"; +my $user = $ENV{"USER"}; + +# process ini file +-e $iniFile || die "Error: could not open $iniFile for reading\n"; +open(INI, $iniFile); + +if ($dryrun){ + write_config(*STDERR); + exit 0; +} else { + if (-e $dir){ + unless($restart){ + die "ERROR: working dir $dir already exists\n\n"; + } + } else { + mkdir $dir; + mkdir "$dir/hgs"; + mkdir "$dir/hgs-current"; + unless (-e $initialWeights) { + print STDERR "Please specify an initial weights file with --initial-weights\n"; + print_help(); + exit; + } + `cp $initialWeights $dir/weights.0`; + die "Can't find weights.0" unless (-e "$dir/weights.0"); + } + unless($restart){ + $logfile = "$dir/mert.log"; + open(LOGFILE, ">$logfile"); + } + write_config(*LOGFILE); +} + + +# Generate initial files and values +unless ($restart){ `cp $iniFile $newIniFile`; } +$iniFile = $newIniFile; + +my $newsrc = "$dir/dev.input"; +unless($restart){ enseg($srcFile, $newsrc); } +$srcFile = $newsrc; +my $devSize = 0; +open F, "<$srcFile" or die "Can't read $srcFile: $!"; +while() { $devSize++; } +close F; + +unless($best_weights){ $best_weights = $weights; } +unless($projected_score){ $projected_score = 0.0; } +$seen_weights{$weights} = 1; + +my $random_seed = int(time / 1000); +my $lastWeightsFile; +my $lastPScore = 0; +# main optimization loop +while (1){ + print LOGFILE "\n\nITERATION $iteration\n==========\n"; + + # iteration-specific files + my $runFile="$dir/run.raw.$iteration"; + my $onebestFile="$dir/1best.$iteration"; + my $logdir="$dir/logs.$iteration"; + my $decoderLog="$logdir/decoder.sentserver.log.$iteration"; + my $scorerLog="$logdir/scorer.log.$iteration"; + `mkdir -p $logdir`; + + #decode + print LOGFILE "DECODE\n"; + print LOGFILE `date`; + my $im1 = $iteration - 1; + my $weightsFile="$dir/weights.$im1"; + my $decoder_cmd = "$decoder -c $iniFile -w $weightsFile -O $dir/hgs-current"; + my $pcmd = "cat $srcFile | $parallelize -p $pmem -e $logdir -n \"$decode_nodes\" -- "; + if ($run_local) { $pcmd = "cat $srcFile |"; } + my $cmd = $pcmd . "$decoder_cmd 2> $decoderLog 1> $runFile"; + print LOGFILE "COMMAND:\n$cmd\n"; + my $result = 0; + $result = system($cmd); + unless ($result == 0){ + cleanup(); + print LOGFILE "ERROR: Parallel decoder returned non-zero exit code $result\n"; + die; + } + my $dec_score = `cat $runFile | $SCORER $refs_comma_sep -l $metric`; + chomp $dec_score; + print LOGFILE "DECODER SCORE: $dec_score\n"; + + # save space + `gzip $runFile`; + `gzip $decoderLog`; + + if ($iteration > $max_iterations){ + print LOGFILE "\nREACHED STOPPING CRITERION: Maximum iterations\n"; + last; + } + + # run optimizer + print LOGFILE "\nUNION FORESTS\n"; + print LOGFILE `date`; + my $mergeLog="$logdir/prune-merge.log.$iteration"; + $cmd = "$forestUnion -r $dir/hgs -n $dir/hgs-current -s $devSize"; + print LOGFILE "COMMAND:\n$cmd\n"; + $result = system($cmd); + unless ($result == 0){ + cleanup(); + print LOGFILE "ERROR: merge command returned non-zero exit code $result\n"; + die; + } + `rm -f $dir/hgs-current/*.json.gz`; # clean up old HGs, they've been moved to the repository + + my $score = 0; + my $icc = 0; + my $inweights="$dir/weights.$im1"; + for (my $opt_iter=1; $opt_iter<$optimization_iters; $opt_iter++) { + print LOGFILE "\nGENERATE OPTIMIZATION STRATEGY (OPT-ITERATION $opt_iter/$optimization_iters)\n"; + print LOGFILE `date`; + $icc++; + $cmd="$MAPINPUT -w $inweights -r $dir/hgs -s $devSize -d $rand_directions > $dir/agenda.$im1-$opt_iter"; + print LOGFILE "COMMAND:\n$cmd\n"; + $result = system($cmd); + unless ($result == 0){ + cleanup(); + print LOGFILE "ERROR: mapinput command returned non-zero exit code $result\n"; + die; + } + + `mkdir $dir/splag.$im1`; + $cmd="split -a 3 -l $lines_per_mapper $dir/agenda.$im1-$opt_iter $dir/splag.$im1/mapinput."; + print LOGFILE "COMMAND:\n$cmd\n"; + $result = system($cmd); + unless ($result == 0){ + cleanup(); + print LOGFILE "ERROR: split command returned non-zero exit code $result\n"; + die; + } + opendir(DIR, "$dir/splag.$im1") or die "Can't open directory: $!"; + my @shards = grep { /^mapinput\./ } readdir(DIR); + closedir DIR; + die "No shards!" unless scalar @shards > 0; + my $joblist = ""; + my $nmappers = 0; + my @mapoutputs = (); + @cleanupcmds = (); + my %o2i = (); + my $first_shard = 1; + for my $shard (@shards) { + my $mapoutput = $shard; + my $client_name = $shard; + $client_name =~ s/mapinput.//; + $client_name = "fmert.$client_name"; + $mapoutput =~ s/mapinput/mapoutput/; + push @mapoutputs, "$dir/splag.$im1/$mapoutput"; + $o2i{"$dir/splag.$im1/$mapoutput"} = "$dir/splag.$im1/$shard"; + my $script = "$MAPPER -l $metric $refs_comma_sep < $dir/splag.$im1/$shard | sort -k1 > $dir/splag.$im1/$mapoutput"; + if ($run_local) { + print LOGFILE "COMMAND:\n$script\n"; + $result = system($script); + unless ($result == 0){ + cleanup(); + print LOGFILE "ERROR: mapper returned non-zero exit code $result\n"; + die; + } + } else { + my $todo = "qsub -q batch -l pmem=3000mb,walltime=5:00:00 -N $client_name -o /dev/null -e $logdir/$client_name.ER"; + local(*QOUT, *QIN); + open2(\*QOUT, \*QIN, $todo); + print QIN $script; + if ($first_shard) { print LOGFILE "$script\n"; $first_shard=0; } + close QIN; + $nmappers++; + while (my $jobid=){ + chomp $jobid; + push(@cleanupcmds, "`qdel $jobid 2> /dev/null`"); + $jobid =~ s/^(\d+)(.*?)$/\1/g; + print STDERR "short job id $jobid\n"; + if ($joblist == "") { $joblist = $jobid; } + else {$joblist = $joblist . "\|" . $jobid; } + } + close QOUT; + } + } + if ($run_local) { + } else { + print LOGFILE "Launched $nmappers mappers.\n"; + print LOGFILE "Waiting for mappers to complete...\n"; + while ($nmappers > 0) { + sleep 5; + my @livejobs = grep(/$joblist/, split(/\n/, `qstat`)); + $nmappers = scalar @livejobs; + } + print LOGFILE "All mappers complete.\n"; + } + my $tol = 0; + my $til = 0; + for my $mo (@mapoutputs) { + my $olines = get_lines($mo); + my $ilines = get_lines($o2i{$mo}); + $tol += $olines; + $til += $ilines; + die "$mo: output lines ($olines) doesn't match input lines ($ilines)" unless $olines==$ilines; + } + print LOGFILE "Results for $tol/$til lines\n"; + print LOGFILE "\nSORTING AND RUNNING FMERT REDUCER\n"; + print LOGFILE `date`; + $cmd="sort -k1 @mapoutputs | $REDUCER > $dir/redoutput.$im1"; + print LOGFILE "COMMAND:\n$cmd\n"; + $result = system($cmd); + unless ($result == 0){ + cleanup(); + print LOGFILE "ERROR: reducer command returned non-zero exit code $result\n"; + die; + } + $cmd="sort -rnk3 '-t|' $dir/redoutput.$im1 | head -1"; + my $best=`$cmd`; chomp $best; + print LOGFILE "$best\n"; + my ($oa, $x, $xscore) = split /\|/, $best; + $score = $xscore; + print LOGFILE "PROJECTED SCORE: $score\n"; + if (abs($x) < $epsilon) { + print LOGFILE "\nOPTIMIZER: no score improvement: abs($x) < $epsilon\n"; + last; + } + my ($origin, $axis) = split /\s+/, $oa; + + my %ori = convert($origin); + my %axi = convert($axis); + + my $finalFile="$dir/weights.$im1-$opt_iter"; + open W, ">$finalFile" or die "Can't write: $finalFile: $!"; + for my $k (sort keys %ori) { + my $v = $ori{$k} + $axi{$k} * $x; + print W "$k $v\n"; + } + + `rm -rf $dir/splag.$im1`; + $inweights = $finalFile; + } + $lastWeightsFile = "$dir/weights.$iteration"; + `cp $inweights $lastWeightsFile`; + if ($icc < 2) { + print LOGFILE "\nREACHED STOPPING CRITERION: score change too little\n"; + last; + } + $lastPScore = $score; + $iteration++; + print LOGFILE "\n==========\n"; +} + +print LOGFILE "\nFINAL WEIGHTS: $dir/$lastWeightsFile\n(Use -w with hiero)\n\n"; + +sub normalize_weights { + my ($rfn, $rpts, $feat) = @_; + my @feat_names = @$rfn; + my @pts = @$rpts; + my $z = 1.0; + for (my $i=0; $i < scalar @feat_names; $i++) { + if ($feat_names[$i] eq $feat) { + $z = $pts[$i]; + last; + } + } + for (my $i=0; $i < scalar @feat_names; $i++) { + $pts[$i] /= $z; + } + print LOGFILE " NORM WEIGHTS: @pts\n"; + return @pts; +} + +sub get_lines { + my $fn = shift @_; + open FL, "<$fn" or die "Couldn't read $fn: $!"; + my $lc = 0; + while() { $lc++; } + return $lc; +} + +sub get_comma_sep_refs { + my ($p) = @_; + my $o = `echo $p`; + chomp $o; + my @files = split /\s+/, $o; + return "-r " . join(' -r ', @files); +} + +sub read_weights_file { + my ($file) = @_; + open F, "<$file" or die "Couldn't read $file: $!"; + my @r = (); + my $pm = -1; + while() { + next if /^#/; + next if /^\s*$/; + chomp; + if (/^(.+)\s+(.+)$/) { + my $m = $1; + my $w = $2; + die "Weights out of order: $m <= $pm" unless $m > $pm; + push @r, $w; + } else { + warn "Unexpected feature name in weight file: $_"; + } + } + close F; + return join ' ', @r; +} + +# subs +sub write_config { + my $fh = shift; + my $cleanup = "yes"; + if ($disable_clean) {$cleanup = "no";} + + print $fh "\n"; + print $fh "DECODER: $decoder\n"; + print $fh "INI FILE: $iniFile\n"; + print $fh "WORKING DIR: $dir\n"; + print $fh "SOURCE (DEV): $srcFile\n"; + print $fh "REFS (DEV): $refFiles\n"; + print $fh "EVAL METRIC: $metric\n"; + print $fh "START ITERATION: $iteration\n"; + print $fh "MAX ITERATIONS: $max_iterations\n"; + print $fh "MERT NODES: $mert_nodes\n"; + print $fh "DECODE NODES: $decode_nodes\n"; + print $fh "HEAD NODE: $host\n"; + print $fh "PMEM (DECODING): $pmem\n"; + print $fh "CLEANUP: $cleanup\n"; + print $fh "INITIAL WEIGHTS: $initialWeights\n"; + + if ($restart){ + print $fh "PROJECTED BLEU: $projected_score\n"; + print $fh "BEST WEIGHTS: $best_weights\n"; + print $fh "BEST BLEU: $best_bleu\n"; + } +} + +sub update_weights_file { + my ($neww, $rfn, $rpts) = @_; + my @feats = @$rfn; + my @pts = @$rpts; + my $num_feats = scalar @feats; + my $num_pts = scalar @pts; + die "$num_feats (num_feats) != $num_pts (num_pts)" unless $num_feats == $num_pts; + open G, ">$neww" or die; + for (my $i = 0; $i < $num_feats; $i++) { + my $f = $feats[$i]; + my $lambda = $pts[$i]; + print G "$f $lambda\n"; + } + close G; +} + +sub enseg { + my $src = shift; + my $newsrc = shift; + open(SRC, $src); + open(NEWSRC, ">$newsrc"); + my $i=0; + while (my $line=){ + chomp $line; + print NEWSRC "$line\n"; + $i++; + } + close SRC; + close NEWSRC; +} + +sub print_help { + + my $executable = `basename $0`; chomp $executable; + print << "Help"; + +Usage: $executable [options] + $executable --restart + + $executable [options] + Runs a complete MERT optimization and test set decoding, using + the decoder configuration in ini file. Note that many of the + options have default values that are inferred automatically + based on certain conventions. For details, refer to descriptions + of the options --decoder, --weights, and --workdir. + + $executable --restart + Continues an optimization run that was stopped for some reason, + using configuration information found in the working directory + left behind by the stopped run. + +Options: + + --local + Run the decoder and optimizer locally. + + --decoder + Decoder binary to use. + + --decode-nodes + A list of nodes used for parallel decoding. If specific nodes + are not desired, use "1" for each node requested. Defaults to + "1 1 1 1 1 1 1 1 1 1 1 1 1 1 1", which indicates a request for + 15 nodes. + + --dont-clean + If present, this flag prevents intermediate files, including + run files and cumulative files, from being automatically removed + after a successful optimization run (these files are left if the + run fails for any reason). If used, a makefile containing + cleanup commands is written to the directory. To clean up + the intermediate files, invoke make without any arguments. + + --dry-run + Prints out the settings and exits without doing anything. + + --epsilon + Require that the dev set BLEU score improve by at least + within iterations (controlled by parameter --interval). + If not specified, defaults to .002. + + --help + Print this message and exit. + + --interval + Require that the dev set BLEU score improve by at least + (controlled by parameter --epsilon) within iterations. + If not specified, defaults to 5. + + --iteration + Starting iteration number. If not specified, defaults to 1. + + --max-iterations + Maximum number of iterations to run. If not specified, defaults + to 10. + + --pmem + Amount of physical memory requested for parallel decoding jobs, + in the format expected by qsub. If not specified, defaults to + 2g. + + --ref-files + Dev set ref files. This option takes only a single string argument. + To use multiple files (including file globbing), this argument should + be quoted. If not specified, defaults to + /fs/cliplab/mteval/Evaluation/Chinese-English/mt03.ref.txt.* + + --metric + Metric to optimize. See fmert's --metric option for values. + Example values: IBM_BLEU, NIST_BLEU, Koehn_BLEU, TER, Combi + + --normalize + After each iteration, rescale all feature weights such that feature- + name has a weight of 1.0. + + --rand-directions + MERT will attempt to optimize along all of the principle directions, + set this parameter to explore other directions. Defaults to 5. + + --source-file + Dev set source file. If not specified, defaults to + /fs/cliplab/mteval/Evaluation/Chinese-English/mt03.src.txt + + --weights + A file specifying initial feature weights. The format is + FeatureName_1 value1 + FeatureName_2 value2 + + --workdir + Directory for intermediate and output files. If not specified, the + name is derived from the ini filename. Assuming that the ini + filename begins with the decoder name and ends with ini, the default + name of the working directory is inferred from the middle part of + the filename. E.g. an ini file named decoder.foo.ini would have + a default working directory name foo. + +Help +} + +sub convert { + my ($str) = @_; + my @ps = split /;/, $str; + my %dict = (); + for my $p (@ps) { + my ($k, $v) = split /=/, $p; + $dict{$k} = $v; + } + return %dict; +} + + -- cgit v1.2.3