clean up

4 files changed, 21 insertions, 193 deletions

diff --git a/phrasinator/Makefile.am b/phrasinator/Makefile.am
index aba98601..486cd21b 100644
--- a/phrasinator/Makefile.am
+++ b/phrasinator/Makefile.am
@@ -11,4 +11,4 @@ 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 -W -Wall -Wno-sign-compare $(GTEST_CPPFLAGS) -I$(top_srcdir)/utils -I$(top_srcdir)/decoder -I$(top_srcdir)/mteval
+AM_CPPFLAGS = -funroll-loops -ffast-math -W -Wall -Werror -I$(top_srcdir)/utils
diff --git a/phrasinator/ccrp_nt.h b/phrasinator/ccrp_nt.h
deleted file mode 100644
index 811bce73..00000000
--- a/phrasinator/ccrp_nt.h
+++ /dev/null
@@ -1,170 +0,0 @@
-#ifndef _CCRP_NT_H_
-#define _CCRP_NT_H_
-
-#include <numeric>
-#include <cassert>
-#include <cmath>
-#include <list>
-#include <iostream>
-#include <vector>
-#include <tr1/unordered_map>
-#include <boost/functional/hash.hpp>
-#include "sampler.h"
-#include "slice_sampler.h"
-
-// Chinese restaurant process (Pitman-Yor parameters) with table tracking.
-
-template <typename Dish, typename DishHash = boost::hash<Dish> >
-class CCRP_NoTable {
- public:
-  explicit CCRP_NoTable(double conc) :
-    num_customers_(),
-    concentration_(conc),
-    concentration_prior_shape_(std::numeric_limits<double>::quiet_NaN()),
-    concentration_prior_rate_(std::numeric_limits<double>::quiet_NaN()) {}
-
-  CCRP_NoTable(double c_shape, double c_rate, double c = 10.0) :
-    num_customers_(),
-    concentration_(c),
-    concentration_prior_shape_(c_shape),
-    concentration_prior_rate_(c_rate) {}
-
-  double concentration() const { return concentration_; }
-
-  bool has_concentration_prior() const {
-    return !std::isnan(concentration_prior_shape_);
-  }
-
-  void clear() {
-    num_customers_ = 0;
-    custs_.clear();
-  }
-
-  unsigned num_customers() const {
-    return num_customers_;
-  }
-
-  unsigned num_customers(const Dish& dish) const {
-    const typename std::tr1::unordered_map<Dish, unsigned, DishHash>::const_iterator it = custs_.find(dish);
-    if (it == custs_.end()) return 0;
-    return it->second;
-  }
-
-  int increment(const Dish& dish) {
-    int table_diff = 0;
-    if (++custs_[dish] == 1)
-      table_diff = 1;
-    ++num_customers_;
-    return table_diff;
-  }
-
-  int decrement(const Dish& dish) {
-    int table_diff = 0;
-    int nc = --custs_[dish];
-    if (nc == 0) {
-      custs_.erase(dish);
-      table_diff = -1;
-    } else if (nc < 0) {
-      std::cerr << "Dish counts dropped below zero for: " << dish << std::endl;
-      abort();
-    }
-    --num_customers_;
-    return table_diff;
-  }
-
-  double prob(const Dish& dish, const double& p0) const {
-    const unsigned at_table = num_customers(dish);
-    return (at_table + p0 * concentration_) / (num_customers_ + concentration_);
-  }
-
-  double logprob(const Dish& dish, const double& logp0) const {
-    const unsigned at_table = num_customers(dish);
-    return log(at_table + exp(logp0 + log(concentration_))) - log(num_customers_ + concentration_);
-  }
-
-  double log_crp_prob() const {
-    return log_crp_prob(concentration_);
-  }
-
-  static double log_gamma_density(const double& x, const double& shape, const double& rate) {
-    assert(x >= 0.0);
-    assert(shape > 0.0);
-    assert(rate > 0.0);
-    const double lp = (shape-1)*log(x) - shape*log(rate) - x/rate - lgamma(shape);
-    return lp;
-  }
-
-  // taken from http://en.wikipedia.org/wiki/Chinese_restaurant_process
-  // does not include P_0's
-  double log_crp_prob(const double& concentration) const {
-    double lp = 0.0;
-    if (has_concentration_prior())
-      lp += log_gamma_density(concentration, concentration_prior_shape_, concentration_prior_rate_);
-    assert(lp <= 0.0);
-    if (num_customers_) {
-      lp += lgamma(concentration) - lgamma(concentration + num_customers_) +
-        custs_.size() * log(concentration);
-      assert(std::isfinite(lp));
-      for (typename std::tr1::unordered_map<Dish, unsigned, DishHash>::const_iterator it = custs_.begin();
-             it != custs_.end(); ++it) {
-          lp += lgamma(it->second);
-      }
-    }
-    assert(std::isfinite(lp));
-    return lp;
-  }
-
-  void resample_hyperparameters(MT19937* rng, const unsigned nloop = 5, const unsigned niterations = 10) {
-    assert(has_concentration_prior());
-    ConcentrationResampler cr(*this);
-    for (int iter = 0; iter < nloop; ++iter) {
-        concentration_ = slice_sampler1d(cr, concentration_, *rng, 0.0,
-                               std::numeric_limits<double>::infinity(), 0.0, niterations, 100*niterations);
-    }
-  }
-
-  struct ConcentrationResampler {
-    ConcentrationResampler(const CCRP_NoTable& crp) : crp_(crp) {}
-    const CCRP_NoTable& crp_;
-    double operator()(const double& proposed_concentration) const {
-      return crp_.log_crp_prob(proposed_concentration);
-    }
-  };
-
-  void Print(std::ostream* out) const {
-    (*out) << "DP(alpha=" << concentration_ << ") customers=" << num_customers_ << std::endl;
-    int cc = 0;
-    for (typename std::tr1::unordered_map<Dish, unsigned, DishHash>::const_iterator it = custs_.begin();
-         it != custs_.end(); ++it) {
-      (*out) << " " << it->first << "(" << it->second << " eating)";
-      ++cc;
-      if (cc > 10) { (*out) << " ..."; break; }
-    }
-    (*out) << std::endl;
-  }
-
-  unsigned num_customers_;
-  std::tr1::unordered_map<Dish, unsigned, DishHash> custs_;
-
-  typedef typename std::tr1::unordered_map<Dish, unsigned, DishHash>::const_iterator const_iterator;
-  const_iterator begin() const {
-    return custs_.begin();
-  }
-  const_iterator end() const {
-    return custs_.end();
-  }
-
-  double concentration_;
-
-  // optional gamma prior on concentration_ (NaN if no prior)
-  double concentration_prior_shape_;
-  double concentration_prior_rate_;
-};
-
-template <typename T,typename H>
-std::ostream& operator<<(std::ostream& o, const CCRP_NoTable<T,H>& c) {
-  c.Print(&o);
-  return o;
-}
-
-#endif
diff --git a/phrasinator/gibbs_train_plm.cc b/phrasinator/gibbs_train_plm.cc
index 86fd7865..7847a460 100644
--- a/phrasinator/gibbs_train_plm.cc
+++ b/phrasinator/gibbs_train_plm.cc
@@ -18,7 +18,7 @@ Dict d; // global dictionary
 
 string Join(char joiner, const vector<int>& phrase) {
   ostringstream os;
-  for (int i = 0; i < phrase.size(); ++i) {
+  for (unsigned i = 0; i < phrase.size(); ++i) {
     if (i > 0) os << joiner;
     os << d.Convert(phrase[i]);
   }
@@ -26,7 +26,7 @@ string Join(char joiner, const vector<int>& phrase) {
 }
 
 ostream& operator<<(ostream& os, const vector<int>& phrase) {
-  for (int i = 0; i < phrase.size(); ++i)
+  for (unsigned i = 0; i < phrase.size(); ++i)
     os << (i == 0 ? "" : " ") << d.Convert(phrase[i]);
   return os;
 }
@@ -37,7 +37,7 @@ struct UnigramLM {
     assert(in);
   }
 
-  double logprob(int word) const {
+  double logprob(unsigned word) const {
     assert(word < freqs_.size());
     return freqs_[word];
   }
@@ -91,7 +91,7 @@ void ReadCorpus(const string& filename, vector<vector<int> >* c, set<int>* vocab
     c->push_back(vector<int>());
     vector<int>& v = c->back();
     d.ConvertWhitespaceDelimitedLine(line, &v);
-    for (int i = 0; i < v.size(); ++i) vocab->insert(v[i]);
+    for (unsigned i = 0; i < v.size(); ++i) vocab->insert(v[i]);
   }
   if (in != &cin) delete in;
 }
@@ -151,7 +151,7 @@ struct UniphraseLM {
     cerr << "Initializing...\n";
     z_.resize(corpus_.size());
     int tc = 0;
-    for (int i = 0; i < corpus_.size(); ++i) {
+    for (unsigned i = 0; i < corpus_.size(); ++i) {
       const vector<int>& line = corpus_[i];
       const int ls = line.size();
       const int last_pos = ls - 1;
@@ -177,7 +177,7 @@ struct UniphraseLM {
     cerr << "Initial LLH: " << llh() << endl;
     cerr << "Sampling...\n";
     cerr << gen_ << endl;
-    for (int s = 1; s < samples; ++s) {
+    for (unsigned s = 1; s < samples; ++s) {
       cerr << '.';
       if (s % 10 == 0) {
         cerr << " [" << s;
@@ -187,7 +187,7 @@ struct UniphraseLM {
         //for (int j = 0; j < z.size(); ++j) z[j] = z_[0][j];
         //SegCorpus::Write(corpus_[0], z, d);
       }
-      for (int i = 0; i < corpus_.size(); ++i) {
+      for (unsigned i = 0; i < corpus_.size(); ++i) {
         const vector<int>& line = corpus_[i];
         const int ls = line.size();
         const int last_pos = ls - 1;
@@ -286,7 +286,7 @@ int main(int argc, char** argv) {
   ulm.Sample(conf["samples"].as<unsigned>(), conf.count("no_hyperparameter_inference") == 0, &rng);
   cerr << "OOV unigram prob: " << ulm.OOVUnigramLogProb() << endl;
 
-  for (int i = 0; i < corpus.size(); ++i)
+  for (unsigned i = 0; i < corpus.size(); ++i)
 //    SegCorpus::Write(corpus[i], shmmlm.z_[i], d);
  ;
   if (conf.count("write_cdec_grammar")) {
@@ -304,8 +304,6 @@ int main(int argc, char** argv) {
     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
index 9dca9e8d..4526eaa6 100644
--- a/phrasinator/gibbs_train_plm.notables.cc
+++ b/phrasinator/gibbs_train_plm.notables.cc
@@ -18,7 +18,7 @@ Dict d; // global dictionary
 
 string Join(char joiner, const vector<int>& phrase) {
   ostringstream os;
-  for (int i = 0; i < phrase.size(); ++i) {
+  for (unsigned i = 0; i < phrase.size(); ++i) {
     if (i > 0) os << joiner;
     os << d.Convert(phrase[i]);
   }
@@ -29,13 +29,13 @@ template <typename BType>
 void WriteSeg(const vector<int>& line, const vector<BType>& label, const Dict& d) {
   assert(line.size() == label.size());
   assert(label.back());
-  int prev = 0;
-  int cur = 0;
+  unsigned prev = 0;
+  unsigned cur = 0;
   while (cur < line.size()) {
     if (label[cur]) {
       if (prev) cout << ' ';
       cout << "{{";
-      for (int i = prev; i <= cur; ++i)
+      for (unsigned i = prev; i <= cur; ++i)
         cout << (i == prev ? "" : " ") << d.Convert(line[i]);
       cout << "}}:" << label[cur];
       prev = cur + 1;
@@ -46,7 +46,7 @@ void WriteSeg(const vector<int>& line, const vector<BType>& label, const Dict& d
 }
 
 ostream& operator<<(ostream& os, const vector<int>& phrase) {
-  for (int i = 0; i < phrase.size(); ++i)
+  for (unsigned i = 0; i < phrase.size(); ++i)
     os << (i == 0 ? "" : " ") << d.Convert(phrase[i]);
   return os;
 }
@@ -57,7 +57,7 @@ struct UnigramLM {
     assert(in);
   }
 
-  double logprob(int word) const {
+  double logprob(unsigned word) const {
     assert(word < freqs_.size());
     return freqs_[word];
   }
@@ -111,7 +111,7 @@ void ReadCorpus(const string& filename, vector<vector<int> >* c, set<int>* vocab
     c->push_back(vector<int>());
     vector<int>& v = c->back();
     d.ConvertWhitespaceDelimitedLine(line, &v);
-    for (int i = 0; i < v.size(); ++i) vocab->insert(v[i]);
+    for (unsigned i = 0; i < v.size(); ++i) vocab->insert(v[i]);
   }
   if (in != &cin) delete in;
 }
@@ -175,7 +175,7 @@ struct UniphraseLM {
     cerr << "Initializing...\n";
     z_.resize(corpus_.size());
     int tc = 0;
-    for (int i = 0; i < corpus_.size(); ++i) {
+    for (unsigned i = 0; i < corpus_.size(); ++i) {
       const vector<int>& line = corpus_[i];
       const int ls = line.size();
       const int last_pos = ls - 1;
@@ -201,7 +201,7 @@ struct UniphraseLM {
     cerr << "Initial LLH: " << llh() << endl;
     cerr << "Sampling...\n";
     cerr << gen_ << endl;
-    for (int s = 1; s < samples; ++s) {
+    for (unsigned s = 1; s < samples; ++s) {
       cerr << '.';
       if (s % 10 == 0) {
         cerr << " [" << s;
@@ -211,7 +211,7 @@ struct UniphraseLM {
         //for (int j = 0; j < z.size(); ++j) z[j] = z_[0][j];
         //SegCorpus::Write(corpus_[0], z, d);
       }
-      for (int i = 0; i < corpus_.size(); ++i) {
+      for (unsigned i = 0; i < corpus_.size(); ++i) {
         const vector<int>& line = corpus_[i];
         const int ls = line.size();
         const int last_pos = ls - 1;
@@ -276,7 +276,7 @@ struct UniphraseLM {
   void ResampleHyperparameters(MT19937* rng) {
     phrases_.resample_hyperparameters(rng);
     gen_.resample_hyperparameters(rng);
-    cerr << " " << phrases_.concentration();
+    cerr << " " << phrases_.alpha();
   }
 
   CCRP_NoTable<vector<int> > phrases_;
@@ -310,7 +310,7 @@ int main(int argc, char** argv) {
   ulm.Sample(conf["samples"].as<unsigned>(), conf.count("no_hyperparameter_inference") == 0, &rng);
   cerr << "OOV unigram prob: " << ulm.OOVUnigramLogProb() << endl;
 
-  for (int i = 0; i < corpus.size(); ++i)
+  for (unsigned i = 0; i < corpus.size(); ++i)
     WriteSeg(corpus[i], ulm.z_[i], d);
 
   if (conf.count("write_cdec_grammar")) {