From e26434979adc33bd949566ba7bf02dff64e80a3e Mon Sep 17 00:00:00 2001 From: Chris Dyer Date: Tue, 2 Oct 2012 00:19:43 -0400 Subject: cdec cleanup, remove bayesian stuff, parsing stuff --- phrasinator/Jamfile | 4 - phrasinator/Makefile.am | 14 -- phrasinator/README | 16 -- phrasinator/gibbs_train_plm.cc | 309 ----------------------------- phrasinator/gibbs_train_plm.notables.cc | 335 -------------------------------- phrasinator/train-phrasinator.pl | 89 --------- 6 files changed, 767 deletions(-) delete mode 100644 phrasinator/Jamfile delete mode 100644 phrasinator/Makefile.am delete mode 100644 phrasinator/README delete mode 100644 phrasinator/gibbs_train_plm.cc delete mode 100644 phrasinator/gibbs_train_plm.notables.cc delete mode 100755 phrasinator/train-phrasinator.pl (limited to 'phrasinator') diff --git a/phrasinator/Jamfile b/phrasinator/Jamfile deleted file mode 100644 index 1fc34f79..00000000 --- a/phrasinator/Jamfile +++ /dev/null @@ -1,4 +0,0 @@ -exe gibbs_train_plm : gibbs_train_plm.cc ..//utils ..//z ..//boost_program_options ; -exe gibbs_train_plm_notables : gibbs_train_plm.notables.cc ..//utils ..//z ..//boost_program_options ; - -alias programs : gibbs_train_plm gibbs_train_plm_notables ; diff --git a/phrasinator/Makefile.am b/phrasinator/Makefile.am deleted file mode 100644 index 3ddd1934..00000000 --- a/phrasinator/Makefile.am +++ /dev/null @@ -1,14 +0,0 @@ -bin_PROGRAMS = gibbs_train_plm gibbs_train_plm_notables - -#head_bigram_model - -gibbs_train_plm_notables_SOURCES = gibbs_train_plm.notables.cc -gibbs_train_plm_notables_LDADD = $(top_srcdir)/utils/libutils.a -lz - -gibbs_train_plm_SOURCES = gibbs_train_plm.cc -gibbs_train_plm_LDADD = $(top_srcdir)/utils/libutils.a -lz - -#head_bigram_model_SOURCES = head_bigram_model.cc -#head_bigram_model_LDADD = $(top_srcdir)/utils/libutils.a -lz - -AM_CPPFLAGS = -funroll-loops -ffast-math -W -Wall -I$(top_srcdir)/utils diff --git a/phrasinator/README b/phrasinator/README deleted file mode 100644 index fb5b93ef..00000000 --- a/phrasinator/README +++ /dev/null @@ -1,16 +0,0 @@ -The "phrasinator" uses a simple Bayesian nonparametric model to segment -text into chunks. The inferred model is then saved so that it can rapidly -predict segments in new (but related) texts. - - Input will be a corpus of sentences, e.g.: - - economists have argued that real interest rates have fallen . - - The output will be a model that, when run with cdec, will produce - a segmentation into phrasal units, e.g.: - - economists have argued that real_interest_rates have fallen . - - -To train a model, run ./train-phrasinator.pl and follow instructions. - diff --git a/phrasinator/gibbs_train_plm.cc b/phrasinator/gibbs_train_plm.cc deleted file mode 100644 index 7847a460..00000000 --- a/phrasinator/gibbs_train_plm.cc +++ /dev/null @@ -1,309 +0,0 @@ -#include -#include - -#include -#include - -#include "filelib.h" -#include "dict.h" -#include "sampler.h" -#include "ccrp.h" -#include "m.h" - -using namespace std; -using namespace std::tr1; -namespace po = boost::program_options; - -Dict d; // global dictionary - -string Join(char joiner, const vector& phrase) { - ostringstream os; - for (unsigned i = 0; i < phrase.size(); ++i) { - if (i > 0) os << joiner; - os << d.Convert(phrase[i]); - } - return os.str(); -} - -ostream& operator<<(ostream& os, const vector& phrase) { - for (unsigned i = 0; i < phrase.size(); ++i) - os << (i == 0 ? "" : " ") << d.Convert(phrase[i]); - return os; -} - -struct UnigramLM { - explicit UnigramLM(const string& fname) { - ifstream in(fname.c_str()); - assert(in); - } - - double logprob(unsigned word) const { - assert(word < freqs_.size()); - return freqs_[word]; - } - - vector freqs_; -}; - -void InitCommandLine(int argc, char** argv, po::variables_map* conf) { - po::options_description opts("Configuration options"); - opts.add_options() - ("samples,s",po::value()->default_value(1000),"Number of samples") - ("input,i",po::value(),"Read file from") - ("random_seed,S",po::value(), "Random seed") - ("write_cdec_grammar,g", po::value(), "Write cdec grammar to this file") - ("write_cdec_weights,w", po::value(), "Write cdec weights to this file") - ("poisson_length,p", "Use a Poisson distribution as the length of a phrase in the base distribuion") - ("no_hyperparameter_inference,N", "Disable hyperparameter inference"); - po::options_description clo("Command line options"); - clo.add_options() - ("config", po::value(), "Configuration file") - ("help,h", "Print this help message and exit"); - po::options_description dconfig_options, dcmdline_options; - dconfig_options.add(opts); - dcmdline_options.add(opts).add(clo); - - po::store(parse_command_line(argc, argv, dcmdline_options), *conf); - if (conf->count("config")) { - ifstream config((*conf)["config"].as().c_str()); - po::store(po::parse_config_file(config, dconfig_options), *conf); - } - po::notify(*conf); - - if (conf->count("help") || (conf->count("input") == 0)) { - cerr << dcmdline_options << endl; - exit(1); - } -} - -void ReadCorpus(const string& filename, vector >* c, set* vocab) { - c->clear(); - istream* in; - if (filename == "-") - in = &cin; - else - in = new ifstream(filename.c_str()); - assert(*in); - string line; - while(*in) { - getline(*in, line); - if (line.empty() && !*in) break; - c->push_back(vector()); - vector& v = c->back(); - d.ConvertWhitespaceDelimitedLine(line, &v); - for (unsigned i = 0; i < v.size(); ++i) vocab->insert(v[i]); - } - if (in != &cin) delete in; -} - -struct UniphraseLM { - UniphraseLM(const vector >& corpus, - const set& vocab, - const po::variables_map& conf) : - phrases_(1,1,1,1), - gen_(1,1,1,1), - corpus_(corpus), - uniform_word_(1.0 / vocab.size()), - gen_p0_(0.5), - p_end_(0.5), - use_poisson_(conf.count("poisson_length") > 0) {} - - double p0(const vector& phrase) const { - static vector p0s(10000, 0.0); - assert(phrase.size() < 10000); - double& p = p0s[phrase.size()]; - if (p) return p; - p = exp(log_p0(phrase)); - if (!p) { - cerr << "0 prob phrase: " << phrase << "\nAssigning std::numeric_limits::min()\n"; - p = std::numeric_limits::min(); - } - return p; - } - - double log_p0(const vector& phrase) const { - double len_logprob; - if (use_poisson_) - len_logprob = Md::log_poisson(phrase.size(), 1.0); - else - len_logprob = log(1 - p_end_) * (phrase.size() -1) + log(p_end_); - return log(uniform_word_) * phrase.size() + len_logprob; - } - - double llh() const { - double llh = gen_.log_crp_prob(); - llh += gen_.num_tables(false) * log(gen_p0_) + - gen_.num_tables(true) * log(1 - gen_p0_); - double llhr = phrases_.log_crp_prob(); - for (CCRP >::const_iterator it = phrases_.begin(); it != phrases_.end(); ++it) { - llhr += phrases_.num_tables(it->first) * log_p0(it->first); - //llhr += log_p0(it->first); - if (!isfinite(llh)) { - cerr << it->first << endl; - cerr << log_p0(it->first) << endl; - abort(); - } - } - return llh + llhr; - } - - void Sample(unsigned int samples, bool hyp_inf, MT19937* rng) { - cerr << "Initializing...\n"; - z_.resize(corpus_.size()); - int tc = 0; - for (unsigned i = 0; i < corpus_.size(); ++i) { - const vector& line = corpus_[i]; - const int ls = line.size(); - const int last_pos = ls - 1; - vector& z = z_[i]; - z.resize(ls); - int prev = 0; - for (int j = 0; j < ls; ++j) { - z[j] = rng->next() < 0.5; - if (j == last_pos) z[j] = true; // break phrase at the end of the sentence - if (z[j]) { - const vector p(line.begin() + prev, line.begin() + j + 1); - phrases_.increment(p, p0(p), rng); - //cerr << p << ": " << p0(p) << endl; - prev = j + 1; - gen_.increment(false, gen_p0_, rng); - ++tc; // remove - } - } - ++tc; - gen_.increment(true, 1.0 - gen_p0_, rng); // end of utterance - } - cerr << "TC: " << tc << endl; - cerr << "Initial LLH: " << llh() << endl; - cerr << "Sampling...\n"; - cerr << gen_ << endl; - for (unsigned s = 1; s < samples; ++s) { - cerr << '.'; - if (s % 10 == 0) { - cerr << " [" << s; - if (hyp_inf) ResampleHyperparameters(rng); - cerr << " LLH=" << llh() << "]\n"; - vector z(z_[0].size(), 0); - //for (int j = 0; j < z.size(); ++j) z[j] = z_[0][j]; - //SegCorpus::Write(corpus_[0], z, d); - } - for (unsigned i = 0; i < corpus_.size(); ++i) { - const vector& line = corpus_[i]; - const int ls = line.size(); - const int last_pos = ls - 1; - vector& z = z_[i]; - int prev = 0; - for (int j = 0; j < last_pos; ++j) { // don't resample last position - int next = j+1; while(!z[next]) { ++next; } - const vector p1p2(line.begin() + prev, line.begin() + next + 1); - const vector p1(line.begin() + prev, line.begin() + j + 1); - const vector p2(line.begin() + j + 1, line.begin() + next + 1); - - if (z[j]) { - phrases_.decrement(p1, rng); - phrases_.decrement(p2, rng); - gen_.decrement(false, rng); - gen_.decrement(false, rng); - } else { - phrases_.decrement(p1p2, rng); - gen_.decrement(false, rng); - } - - const double d1 = phrases_.prob(p1p2, p0(p1p2)) * gen_.prob(false, gen_p0_); - double d2 = phrases_.prob(p1, p0(p1)) * gen_.prob(false, gen_p0_); - phrases_.increment(p1, p0(p1), rng); - gen_.increment(false, gen_p0_, rng); - d2 *= phrases_.prob(p2, p0(p2)) * gen_.prob(false, gen_p0_); - phrases_.decrement(p1, rng); - gen_.decrement(false, rng); - z[j] = rng->SelectSample(d1, d2); - - if (z[j]) { - phrases_.increment(p1, p0(p1), rng); - phrases_.increment(p2, p0(p2), rng); - gen_.increment(false, gen_p0_, rng); - gen_.increment(false, gen_p0_, rng); - prev = j + 1; - } else { - phrases_.increment(p1p2, p0(p1p2), rng); - gen_.increment(false, gen_p0_, rng); - } - } - } - } -// cerr << endl << endl << gen_ << endl << phrases_ << endl; - cerr << gen_.prob(false, gen_p0_) << " " << gen_.prob(true, 1 - gen_p0_) << endl; - } - - void WriteCdecGrammarForCurrentSample(ostream* os) const { - CCRP >::const_iterator it = phrases_.begin(); - for (; it != phrases_.end(); ++it) { - (*os) << "[X] ||| " << Join(' ', it->first) << " ||| " - << Join('_', it->first) << " ||| C=1 P=" - << log(phrases_.prob(it->first, p0(it->first))) << endl; - } - } - - double OOVUnigramLogProb() const { - vector x(1,99999999); - return log(phrases_.prob(x, p0(x))); - } - - void ResampleHyperparameters(MT19937* rng) { - phrases_.resample_hyperparameters(rng); - gen_.resample_hyperparameters(rng); - cerr << " d=" << phrases_.discount() << ",s=" << phrases_.strength(); - } - - CCRP > phrases_; - CCRP gen_; - vector > z_; // z_[i] is there a phrase boundary after the ith word - const vector >& corpus_; - const double uniform_word_; - const double gen_p0_; - const double p_end_; // in base length distribution, p of the end of a phrase - const bool use_poisson_; -}; - - -int main(int argc, char** argv) { - po::variables_map conf; - InitCommandLine(argc, argv, &conf); - boost::shared_ptr prng; - if (conf.count("random_seed")) - prng.reset(new MT19937(conf["random_seed"].as())); - else - prng.reset(new MT19937); - MT19937& rng = *prng; - - vector > corpus; - set vocab; - ReadCorpus(conf["input"].as(), &corpus, &vocab); - cerr << "Corpus size: " << corpus.size() << " sentences\n"; - cerr << "Vocabulary size: " << vocab.size() << " types\n"; - - UniphraseLM ulm(corpus, vocab, conf); - ulm.Sample(conf["samples"].as(), conf.count("no_hyperparameter_inference") == 0, &rng); - cerr << "OOV unigram prob: " << ulm.OOVUnigramLogProb() << endl; - - for (unsigned i = 0; i < corpus.size(); ++i) -// SegCorpus::Write(corpus[i], shmmlm.z_[i], d); - ; - if (conf.count("write_cdec_grammar")) { - string fname = conf["write_cdec_grammar"].as(); - cerr << "Writing model to " << fname << " ...\n"; - WriteFile wf(fname); - ulm.WriteCdecGrammarForCurrentSample(wf.stream()); - } - - if (conf.count("write_cdec_weights")) { - string fname = conf["write_cdec_weights"].as(); - cerr << "Writing weights to " << fname << " .\n"; - WriteFile wf(fname); - ostream& os = *wf.stream(); - os << "# make C smaller to use more phrases\nP 1\nPassThrough " << ulm.OOVUnigramLogProb() << "\nC -3\n"; - } - - return 0; -} - diff --git a/phrasinator/gibbs_train_plm.notables.cc b/phrasinator/gibbs_train_plm.notables.cc deleted file mode 100644 index 4526eaa6..00000000 --- a/phrasinator/gibbs_train_plm.notables.cc +++ /dev/null @@ -1,335 +0,0 @@ -#include -#include - -#include -#include - -#include "filelib.h" -#include "dict.h" -#include "sampler.h" -#include "ccrp.h" -#include "ccrp_nt.h" - -using namespace std; -using namespace std::tr1; -namespace po = boost::program_options; - -Dict d; // global dictionary - -string Join(char joiner, const vector& phrase) { - ostringstream os; - for (unsigned i = 0; i < phrase.size(); ++i) { - if (i > 0) os << joiner; - os << d.Convert(phrase[i]); - } - return os.str(); -} - -template -void WriteSeg(const vector& line, const vector& label, const Dict& d) { - assert(line.size() == label.size()); - assert(label.back()); - unsigned prev = 0; - unsigned cur = 0; - while (cur < line.size()) { - if (label[cur]) { - if (prev) cout << ' '; - cout << "{{"; - for (unsigned i = prev; i <= cur; ++i) - cout << (i == prev ? "" : " ") << d.Convert(line[i]); - cout << "}}:" << label[cur]; - prev = cur + 1; - } - ++cur; - } - cout << endl; -} - -ostream& operator<<(ostream& os, const vector& phrase) { - for (unsigned i = 0; i < phrase.size(); ++i) - os << (i == 0 ? "" : " ") << d.Convert(phrase[i]); - return os; -} - -struct UnigramLM { - explicit UnigramLM(const string& fname) { - ifstream in(fname.c_str()); - assert(in); - } - - double logprob(unsigned word) const { - assert(word < freqs_.size()); - return freqs_[word]; - } - - vector freqs_; -}; - -void InitCommandLine(int argc, char** argv, po::variables_map* conf) { - po::options_description opts("Configuration options"); - opts.add_options() - ("samples,s",po::value()->default_value(1000),"Number of samples") - ("input,i",po::value(),"Read file from") - ("random_seed,S",po::value(), "Random seed") - ("write_cdec_grammar,g", po::value(), "Write cdec grammar to this file") - ("write_cdec_weights,w", po::value(), "Write cdec weights to this file") - ("poisson_length,p", "Use a Poisson distribution as the length of a phrase in the base distribuion") - ("no_hyperparameter_inference,N", "Disable hyperparameter inference"); - po::options_description clo("Command line options"); - clo.add_options() - ("config", po::value(), "Configuration file") - ("help,h", "Print this help message and exit"); - po::options_description dconfig_options, dcmdline_options; - dconfig_options.add(opts); - dcmdline_options.add(opts).add(clo); - - po::store(parse_command_line(argc, argv, dcmdline_options), *conf); - if (conf->count("config")) { - ifstream config((*conf)["config"].as().c_str()); - po::store(po::parse_config_file(config, dconfig_options), *conf); - } - po::notify(*conf); - - if (conf->count("help") || (conf->count("input") == 0)) { - cerr << dcmdline_options << endl; - exit(1); - } -} - -void ReadCorpus(const string& filename, vector >* c, set* vocab) { - c->clear(); - istream* in; - if (filename == "-") - in = &cin; - else - in = new ifstream(filename.c_str()); - assert(*in); - string line; - while(*in) { - getline(*in, line); - if (line.empty() && !*in) break; - c->push_back(vector()); - vector& v = c->back(); - d.ConvertWhitespaceDelimitedLine(line, &v); - for (unsigned i = 0; i < v.size(); ++i) vocab->insert(v[i]); - } - if (in != &cin) delete in; -} - -double log_poisson(unsigned x, const double& lambda) { - assert(lambda > 0.0); - return log(lambda) * x - lgamma(x + 1) - lambda; -} - -struct UniphraseLM { - UniphraseLM(const vector >& corpus, - const set& vocab, - const po::variables_map& conf) : - phrases_(1,1), - gen_(1,1), - corpus_(corpus), - uniform_word_(1.0 / vocab.size()), - gen_p0_(0.5), - p_end_(0.5), - use_poisson_(conf.count("poisson_length") > 0) {} - - double p0(const vector& phrase) const { - static vector p0s(10000, 0.0); - assert(phrase.size() < 10000); - double& p = p0s[phrase.size()]; - if (p) return p; - p = exp(log_p0(phrase)); - if (!p) { - cerr << "0 prob phrase: " << phrase << "\nAssigning std::numeric_limits::min()\n"; - p = std::numeric_limits::min(); - } - return p; - } - - double log_p0(const vector& phrase) const { - double len_logprob; - if (use_poisson_) - len_logprob = log_poisson(phrase.size(), 1.0); - else - len_logprob = log(1 - p_end_) * (phrase.size() -1) + log(p_end_); - return log(uniform_word_) * phrase.size() + len_logprob; - } - - double llh() const { - double llh = gen_.log_crp_prob(); - llh += log(gen_p0_) + log(1 - gen_p0_); - double llhr = phrases_.log_crp_prob(); - for (CCRP_NoTable >::const_iterator it = phrases_.begin(); it != phrases_.end(); ++it) { - llhr += log_p0(it->first); - //llhr += log_p0(it->first); - if (!isfinite(llh)) { - cerr << it->first << endl; - cerr << log_p0(it->first) << endl; - abort(); - } - } - return llh + llhr; - } - - void Sample(unsigned int samples, bool hyp_inf, MT19937* rng) { - cerr << "Initializing...\n"; - z_.resize(corpus_.size()); - int tc = 0; - for (unsigned i = 0; i < corpus_.size(); ++i) { - const vector& line = corpus_[i]; - const int ls = line.size(); - const int last_pos = ls - 1; - vector& z = z_[i]; - z.resize(ls); - int prev = 0; - for (int j = 0; j < ls; ++j) { - z[j] = rng->next() < 0.5; - if (j == last_pos) z[j] = true; // break phrase at the end of the sentence - if (z[j]) { - const vector p(line.begin() + prev, line.begin() + j + 1); - phrases_.increment(p); - //cerr << p << ": " << p0(p) << endl; - prev = j + 1; - gen_.increment(false); - ++tc; // remove - } - } - ++tc; - gen_.increment(true); // end of utterance - } - cerr << "TC: " << tc << endl; - cerr << "Initial LLH: " << llh() << endl; - cerr << "Sampling...\n"; - cerr << gen_ << endl; - for (unsigned s = 1; s < samples; ++s) { - cerr << '.'; - if (s % 10 == 0) { - cerr << " [" << s; - if (hyp_inf) ResampleHyperparameters(rng); - cerr << " LLH=" << llh() << "]\n"; - vector z(z_[0].size(), 0); - //for (int j = 0; j < z.size(); ++j) z[j] = z_[0][j]; - //SegCorpus::Write(corpus_[0], z, d); - } - for (unsigned i = 0; i < corpus_.size(); ++i) { - const vector& line = corpus_[i]; - const int ls = line.size(); - const int last_pos = ls - 1; - vector& z = z_[i]; - int prev = 0; - for (int j = 0; j < last_pos; ++j) { // don't resample last position - int next = j+1; while(!z[next]) { ++next; } - const vector p1p2(line.begin() + prev, line.begin() + next + 1); - const vector p1(line.begin() + prev, line.begin() + j + 1); - const vector p2(line.begin() + j + 1, line.begin() + next + 1); - - if (z[j]) { - phrases_.decrement(p1); - phrases_.decrement(p2); - gen_.decrement(false); - gen_.decrement(false); - } else { - phrases_.decrement(p1p2); - gen_.decrement(false); - } - - const double d1 = phrases_.prob(p1p2, p0(p1p2)) * gen_.prob(false, gen_p0_); - double d2 = phrases_.prob(p1, p0(p1)) * gen_.prob(false, gen_p0_); - phrases_.increment(p1); - gen_.increment(false); - d2 *= phrases_.prob(p2, p0(p2)) * gen_.prob(false, gen_p0_); - phrases_.decrement(p1); - gen_.decrement(false); - z[j] = rng->SelectSample(d1, d2); - - if (z[j]) { - phrases_.increment(p1); - phrases_.increment(p2); - gen_.increment(false); - gen_.increment(false); - prev = j + 1; - } else { - phrases_.increment(p1p2); - gen_.increment(false); - } - } - } - } -// cerr << endl << endl << gen_ << endl << phrases_ << endl; - cerr << gen_.prob(false, gen_p0_) << " " << gen_.prob(true, 1 - gen_p0_) << endl; - } - - void WriteCdecGrammarForCurrentSample(ostream* os) const { - CCRP_NoTable >::const_iterator it = phrases_.begin(); - for (; it != phrases_.end(); ++it) { - (*os) << "[X] ||| " << Join(' ', it->first) << " ||| " - << Join('_', it->first) << " ||| C=1 P=" - << log(phrases_.prob(it->first, p0(it->first))) << endl; - } - } - - double OOVUnigramLogProb() const { - vector x(1,99999999); - return log(phrases_.prob(x, p0(x))); - } - - void ResampleHyperparameters(MT19937* rng) { - phrases_.resample_hyperparameters(rng); - gen_.resample_hyperparameters(rng); - cerr << " " << phrases_.alpha(); - } - - CCRP_NoTable > phrases_; - CCRP_NoTable gen_; - vector > z_; // z_[i] is there a phrase boundary after the ith word - const vector >& corpus_; - const double uniform_word_; - const double gen_p0_; - const double p_end_; // in base length distribution, p of the end of a phrase - const bool use_poisson_; -}; - - -int main(int argc, char** argv) { - po::variables_map conf; - InitCommandLine(argc, argv, &conf); - boost::shared_ptr prng; - if (conf.count("random_seed")) - prng.reset(new MT19937(conf["random_seed"].as())); - else - prng.reset(new MT19937); - MT19937& rng = *prng; - - vector > corpus; - set vocab; - ReadCorpus(conf["input"].as(), &corpus, &vocab); - cerr << "Corpus size: " << corpus.size() << " sentences\n"; - cerr << "Vocabulary size: " << vocab.size() << " types\n"; - - UniphraseLM ulm(corpus, vocab, conf); - ulm.Sample(conf["samples"].as(), conf.count("no_hyperparameter_inference") == 0, &rng); - cerr << "OOV unigram prob: " << ulm.OOVUnigramLogProb() << endl; - - for (unsigned i = 0; i < corpus.size(); ++i) - WriteSeg(corpus[i], ulm.z_[i], d); - - if (conf.count("write_cdec_grammar")) { - string fname = conf["write_cdec_grammar"].as(); - cerr << "Writing model to " << fname << " ...\n"; - WriteFile wf(fname); - ulm.WriteCdecGrammarForCurrentSample(wf.stream()); - } - - if (conf.count("write_cdec_weights")) { - string fname = conf["write_cdec_weights"].as(); - cerr << "Writing weights to " << fname << " .\n"; - WriteFile wf(fname); - ostream& os = *wf.stream(); - os << "# make C smaller to use more phrases\nP 1\nPassThrough " << ulm.OOVUnigramLogProb() << "\nC -3\n"; - } - - - - return 0; -} - diff --git a/phrasinator/train-phrasinator.pl b/phrasinator/train-phrasinator.pl deleted file mode 100755 index c50b8e68..00000000 --- a/phrasinator/train-phrasinator.pl +++ /dev/null @@ -1,89 +0,0 @@ -#!/usr/bin/perl -w -use strict; -my $script_dir; BEGIN { use Cwd qw/ abs_path cwd /; use File::Basename; $script_dir = dirname(abs_path($0)); push @INC, $script_dir; } -use Getopt::Long; -use File::Spec qw (rel2abs); - -my $DECODER = "$script_dir/../decoder/cdec"; -my $TRAINER = "$script_dir/gibbs_train_plm_notables"; - -die "Can't find $TRAINER" unless -f $TRAINER; -die "Can't execute $TRAINER" unless -x $TRAINER; - -if (!GetOptions( - "decoder=s" => \$DECODER, -)) { usage(); } - -die "Can't find $DECODER" unless -f $DECODER; -die "Can't execute $DECODER" unless -x $DECODER; -if (scalar @ARGV != 2) { usage(); } -my $INFILE = shift @ARGV; -my $OUTDIR = shift @ARGV; -$OUTDIR = File::Spec->rel2abs($OUTDIR); -print STDERR " Input file: $INFILE\n"; -print STDERR "Output directory: $OUTDIR\n"; -open F, "<$INFILE" or die "Failed to open $INFILE for reading: $!"; -close F; -die "Please remove existing directory $OUTDIR\n" if (-f $OUTDIR || -d $OUTDIR); - -my $CMD = "mkdir $OUTDIR"; -safesystem($CMD) or die "Failed to create directory $OUTDIR\n$!"; - -my $grammar="$OUTDIR/grammar.gz"; -my $weights="$OUTDIR/weights"; -$CMD = "$TRAINER -w $weights -g $grammar -i $INFILE"; -safesystem($CMD) or die "Failed to train model!\n"; -my $cdecini = "$OUTDIR/cdec.ini"; -open C, ">$cdecini" or die "Failed to open $cdecini for writing: $!"; - -print C <> 8; - print STDERR "Exit code: $exitcode\n" if $exitcode; - return ! $exitcode; - } -} - -- cgit v1.2.3