summaryrefslogtreecommitdiff
path: root/extractor/run_extractor.cc
diff options
context:
space:
mode:
authorPaul Baltescu <pauldb89@gmail.com>2013-11-26 15:01:14 +0000
committerPaul Baltescu <pauldb89@gmail.com>2013-11-26 15:01:14 +0000
commitbed3e4b867e4132917fa0640956e8ce713f0e451 (patch)
treea442aa2233c85313aebcf364ec0a0804922d2db7 /extractor/run_extractor.cc
parente633526bc2ba1f73e88989f495d70c0d2ec84a97 (diff)
Script for grammar extraction only.
Diffstat (limited to 'extractor/run_extractor.cc')
-rw-r--r--extractor/run_extractor.cc20
1 files changed, 8 insertions, 12 deletions
diff --git a/extractor/run_extractor.cc b/extractor/run_extractor.cc
index 6b22a302..f1aa5e35 100644
--- a/extractor/run_extractor.cc
+++ b/extractor/run_extractor.cc
@@ -5,10 +5,10 @@
#include <string>
#include <vector>
-#include <omp.h>
#include <boost/filesystem.hpp>
#include <boost/program_options.hpp>
#include <boost/program_options/variables_map.hpp>
+#include <omp.h>
#include "alignment.h"
#include "data_array.h"
@@ -78,7 +78,8 @@ int main(int argc, char** argv) {
("tight_phrases", po::value<bool>()->default_value(true),
"False if phrases may be loose (better, but slower)")
("leave_one_out", po::value<bool>()->zero_tokens(),
- "do leave-one-out estimation of grammars (e.g. for extracting grammars for the training set");
+ "do leave-one-out estimation of grammars "
+ "(e.g. for extracting grammars for the training set");
po::variables_map vm;
po::store(po::parse_command_line(argc, argv, desc), vm);
@@ -99,11 +100,6 @@ int main(int argc, char** argv) {
return 1;
}
- bool leave_one_out = false;
- if (vm.count("leave_one_out")) {
- leave_one_out = true;
- }
-
int num_threads = vm["threads"].as<int>();
cerr << "Grammar extraction will use " << num_threads << " threads." << endl;
@@ -178,8 +174,8 @@ int main(int argc, char** argv) {
<< GetDuration(preprocess_start_time, preprocess_stop_time)
<< " seconds" << endl;
- // Features used to score each grammar rule.
Clock::time_point extraction_start_time = Clock::now();
+ // Features used to score each grammar rule.
vector<shared_ptr<Feature>> features = {
make_shared<TargetGivenSourceCoherent>(),
make_shared<SampleSourceCount>(),
@@ -206,9 +202,6 @@ int main(int argc, char** argv) {
vm["max_samples"].as<int>(),
vm["tight_phrases"].as<bool>());
- // Releases extra memory used by the initial precomputation.
- precomputation.reset();
-
// Creates the grammars directory if it doesn't exist.
fs::path grammar_path = vm["grammars"].as<string>();
if (!fs::is_directory(grammar_path)) {
@@ -224,6 +217,7 @@ int main(int argc, char** argv) {
}
// Extracts the grammar for each sentence and saves it to a file.
+ bool leave_one_out = vm.count("leave_one_out");
vector<string> suffixes(sentences.size());
#pragma omp parallel for schedule(dynamic) num_threads(num_threads)
for (size_t i = 0; i < sentences.size(); ++i) {
@@ -236,7 +230,9 @@ int main(int argc, char** argv) {
suffixes[i] = suffix;
unordered_set<int> blacklisted_sentence_ids;
- if (leave_one_out) blacklisted_sentence_ids.insert(i);
+ if (leave_one_out) {
+ blacklisted_sentence_ids.insert(i);
+ }
Grammar grammar = extractor.GetGrammar(
sentences[i], blacklisted_sentence_ids);
ofstream output(GetGrammarFilePath(grammar_path, i).c_str());