Skip to content

Commit

Permalink
Merge pull request HIT-SCIR#127 from icycandy/master
Browse files Browse the repository at this point in the history 
improve program options processing and logging for parser.n
  • Loading branch information
@Oneplus
Oneplus committed Aug 10, 2015
2 parents c05d7e4 + ccd36b3 commit 7eae6a7
Show file tree
Hide file tree
Showing 4 changed files with 71 additions and 123 deletions.
180 changes: 63 additions & 117 deletions src/parser.n/main.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -23,43 +23,31 @@ int test(int argc, char** argv) {
usage += "Testing suite for " DESCRIPTION "\n\n";
usage += "usage: ./" EXECUTABLE " test <options>\n\n";
usage += "options:";
options_description optparser = options_description(usage);
TestOption opt;
options_description optparser(usage);
optparser.add_options()
("model", value<std::string>(), "The path to the model.")
("input", value<std::string>(), "The path to the reference.")
("evaluate", value<bool>()->default_value(false),
("model", value<std::string>(&opt.model_file)->required(), "The path to the model.")
("input", value<std::string>(&opt.input_file)->required(), "The path to the reference.")
("evaluate", value<bool>(&opt.evaluate)->default_value(false),
"if configured, perform evaluation, heads and deprels columns should be filled.")
("help,h", "Show help information");

if (argc == 1) { std::cerr << optparser << std::endl; return 1; }

variables_map vm;
store(parse_command_line(argc, argv, optparser), vm);

if (vm.count("help")) {
std::cerr << optparser << std::endl;
return 1;
}

TestOption opt;
opt.model_file = "";
if (!vm.count("model")) {
ERROR_LOG("model path should be specified [--model].");
return 1;
} else {
opt.model_file = vm["model"].as<std::string>();
try {
store(parse_command_line(argc, argv, optparser), vm);
if (vm.count("help")) {
std::cerr << optparser << std::endl;
return 1;
}

boost::program_options::notify(vm);
} catch(const boost::program_options::error &e) {
std::cerr << e.what() << std::endl;
std::cout << optparser << std::endl;
}

opt.input_file = "";
if (!vm.count("input")) {
ERROR_LOG("input file should be specified [--input].");
return 1;
} else {
opt.input_file = vm["input"].as<std::string>();
}

opt.evaluate = vm["evaluate"].as<bool>();

NeuralNetworkParserFrontend frontend(opt);
frontend.test();
return 0;
Expand All @@ -70,118 +58,76 @@ int learn(int argc, char** argv) {
usage += "Training suite for " DESCRIPTION "\n";
usage += "usage: ./" EXECUTABLE " learn <options>\n\n";
usage += "options:";
options_description optparser = options_description(usage);
LearnOption opt;
options_description optparser(usage);
optparser.add_options()
("model", value<std::string>(), "The path to the model.")
("embedding", value<std::string>(), "The path to the embedding file.")
("reference", value<std::string>(), "The path to the reference file.")
("development", value<std::string>(), "The path to the development file.\n")
("init-range", value<double>()->default_value(0.01), "The initialization range. [default=0.01]")
("word-cutoff", value<int>()->default_value(1), "The frequency of rare word. [default=1]")
("max-iter", value<int>()->default_value(20000), "The number of max iteration. [default=20000]")
("batch-size", value<int>()->default_value(10000), "The size of batch. [default=10000]")
("hidden-size", value<int>()->default_value(200), "The size of hidden layer. [default=200]")
("embedding-size", value<int>()->default_value(50), "The size of embedding. [default=50]")
("features-number", value<int>()->default_value(48), "The number of features. [default=48]")
("precomputed-number", value<int>()->default_value(100000), "The number of precomputed. [default=100000]")
("evaluation-stops", value<int>()->default_value(100), "Evaluation on per-iteration. [default=100]")
("ada-eps", value<double>()->default_value(1e-6), "The EPS in AdaGrad. [defautl=1e-6]")
("ada-alpha", value<double>()->default_value(0.01), "The Alpha in AdaGrad. [default=0.01]")
("lambda", value<double>()->default_value(1e-8), "The regularizer parameter. [default=1e-8]")
("dropout-probability", value<double>()->default_value(0.5), "The probability for dropout. [default=0.5]")
("oracle", value<std::string>()->default_value("static"),
("model", value<std::string>(&opt.model_file)->required(), "The path to the model.")
("embedding", value<std::string>(&opt.embedding_file)->required(), "The path to the embedding file.")
("reference", value<std::string>(&opt.reference_file)->required(), "The path to the reference file.")
("development", value<std::string>(&opt.devel_file), "The path to the development file.\n")
("init-range", value<double>(&opt.init_range)->default_value(0.01), "The initialization range. [default=0.01]")
("word-cutoff", value<int>(&opt.word_cutoff)->default_value(1), "The frequency of rare word. Word with frequency less than cutoff will be considered unknown")
("max-iter", value<int>(&opt.max_iter)->default_value(20000), "The number of max iteration. [default=20000]")
("batch-size", value<int>(&opt.batch_size)->default_value(10000), "The size of batch. [default=10000]")
("hidden-size", value<int>(&opt.hidden_layer_size)->default_value(200), "The size of hidden layer. [default=200]")
("embedding-size", value<int>(&opt.embedding_size)->default_value(50), "The size of embedding. [default=50]")
//("features-number", value<int>(), "The number of features. [default=48]")
("precomputed-number", value<int>(&opt.nr_precomputed)->default_value(100000), "The number of precomputed. [default=100000]")
("evaluation-stops", value<int>(&opt.evaluation_stops)->default_value(100), "Evaluation on per-iteration. [default=100]")
("ada-eps", value<double>(&opt.ada_eps)->default_value(1e-6), "The EPS in AdaGrad. [defautl=1e-6]")
("ada-alpha", value<double>(&opt.ada_alpha)->default_value(0.01), "The Alpha in AdaGrad. [default=0.01]")
("lambda", value<double>(&opt.lambda)->default_value(1e-8), "The regularizer parameter. [default=1e-8]")
("dropout-probability", value<double>(&opt.dropout_probability)->default_value(0.5), "The probability for dropout. [default=0.5]")
("oracle", value<std::string>(&opt.oracle)->default_value("static"),
"The oracle type\n"
" - static: The static oracle [default]\n"
" - nondet: The non-deterministic oracle\n"
" - explore: The explore oracle.")
("save-intermediate", value<bool>()->default_value(true), "Save the intermediate. [default=true]")
("fix-embeddings", value<bool>()->default_value(false), "Fix the embeddings. [default=false]")
("use-distance", value<bool>()->default_value(false), "Specify to use distance feature. [default=false]")
("use-valency", value<bool>()->default_value(false), "Specify to use valency feature. [default=false]")
("use-cluster", value<bool>()->default_value(false), "Specify to use cluster feature. [default=false]")
("cluster", value<std::string>(), "Specify the path to the cluster file.")
("root", value<std::string>()->default_value("ROOT"), "The root tag. [default=ROOT]")
("save-intermediate", value<bool>(&opt.save_intermediate)->default_value(true), "Save the intermediate. [default=true]")
("fix-embeddings", value<bool>(&opt.fix_embeddings)->default_value(false), "Fix the embeddings. [default=false]")
("use-distance", value<bool>(&opt.use_distance)->default_value(false), "Specify to use distance feature. [default=false]")
("use-valency", value<bool>(&opt.use_valency)->default_value(false), "Specify to use valency feature. [default=false]")
("use-cluster", value<bool>(&opt.use_cluster)->default_value(false), "Specify to use cluster feature. [default=false]")
("cluster", value<std::string>(&opt.cluster_file), "Specify the path to the cluster file.")
("root", value<std::string>(&opt.root)->default_value("ROOT"), "The root tag. [default=ROOT]")
("verbose", "Logging more details.")
("help,h", "Show help information.");

if (argc == 1) { std::cerr << optparser << std::endl; return 1; }

variables_map vm;
store(parse_command_line(argc, argv, optparser), vm);

if (vm.count("help")) {
std::cerr << optparser << std::endl;
return 1;
try {
store(parse_command_line(argc, argv, optparser), vm);
if (vm.count("help")) {
std::cerr << optparser << std::endl;
return 1;
}

boost::program_options::notify(vm);
} catch (const boost::program_options::error &e) {
std::cerr << e.what() << std::endl;
std::cout << optparser << std::endl;
}

LearnOption opt;

opt.model_file = "";
if (!vm.count("model")) {
ERROR_LOG("parse opt: model file must be specified [--model].");
return 1;
} else {
opt.model_file= vm["model"].as<std::string>();
}
// opt.clear_gradient_per_iter = 0;
// if (vm.count("clear-gradient-per-iter")) {
// opt.clear_gradient_per_iter = vm["clear-gradient-per-iter"].as<int>();
// }

opt.embedding_file = "";
if (!vm.count("embedding")) {
ERROR_LOG("parse opt: embedding file must be specified [--embedding].");
return 1;
} else {
opt.embedding_file = vm["embedding"].as<std::string>();
}

opt.reference_file = "";
if (!vm.count("reference")) {
ERROR_LOG("parse opt: reference file must be specified [--reference].");
return false;
} else {
opt.reference_file = vm["reference"].as<std::string>();
}

opt.devel_file = "";
if (vm.count("development")) {
opt.devel_file = vm["development"].as<std::string>();
}

opt.ada_eps = vm["ada-eps"].as<double>();
opt.ada_alpha = vm["ada-alpha"].as<double>();
opt.lambda = vm["lambda"].as<double>();
opt.dropout_probability = vm["dropout-probability"].as<double>();
opt.hidden_layer_size = vm["hidden-size"].as<int>();
opt.embedding_size = vm["embedding-size"].as<int>();
opt.max_iter = vm["max-iter"].as<int>();
opt.init_range = vm["init-range"].as<double>();
opt.word_cutoff = vm["word-cutoff"].as<int>();
opt.batch_size = vm["batch-size"].as<int>();
opt.nr_precomputed = vm["precomputed-number"].as<int>();
opt.evaluation_stops = vm["evaluation-stops"].as<int>();
opt.clear_gradient_per_iter = 0;
opt.oracle = vm["oracle"].as<std::string>();
//some extra validate of opt
if (opt.oracle != "static" && opt.oracle != "nondet" && opt.oracle != "explore") {
ERROR_LOG("parse opt [--oracle]: oracle value error, allowed values are static|nondet|explore.");
opt.oracle = "static";
}

opt.save_intermediate = vm["save-intermediate"].as<bool>();
opt.fix_embeddings = vm["fix-embeddings"].as<bool>();
opt.root = vm["root"].as<std::string>();
opt.use_distance = vm["use-distance"].as<bool>();
opt.use_valency = vm["use-valency"].as<bool>();
opt.use_cluster = vm["use-cluster"].as<bool>();

opt.cluster_file = "";
if (opt.use_cluster) {
if (vm.count("cluster")) {
opt.cluster_file = vm["cluster"].as<std::string>();
} else {
ERROR_LOG("cluster file should be specified when using cluster feature.");
if (opt.use_cluster && vm.count("cluster") == 0) {
ERROR_LOG("parse opt [--cluster]: cluster file should be specified when using cluster feature.");
return 1;
}
}

NeuralNetworkParserFrontend frontend(opt);
frontend.train();

return 0;
}

Expand Down
9 changes: 5 additions & 4 deletions src/parser.n/parser.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -261,17 +261,18 @@ void NeuralNetworkParser::report() {

void NeuralNetworkParser::transduce_instance_to_dependency(const Instance& data,
Dependency* dependency, bool with_dependencies) {
size_t L = data.forms.size();
size_t L = data.size();
for (size_t i = 0; i < L; ++ i) {
int form = forms_alphabet.index(data.forms[i]);
if (form == -1) { form = forms_alphabet.index(SpecialOption::UNKNOWN); }
int postag = postags_alphabet.index(data.postags[i]);
int head = with_dependencies? data.heads[i]: -1;
int deprel = (with_dependencies ? deprels_alphabet.index(data.deprels[i]): -1);

dependency->forms.push_back(form);
dependency->postags.push_back(postag);
dependency->heads.push_back(with_dependencies? data.heads[i]: -1);
dependency->deprels.push_back(with_dependencies? deprel: -1);
dependency->heads.push_back(head);
dependency->deprels.push_back(deprel);
}
}

Expand All @@ -280,7 +281,7 @@ void NeuralNetworkParser::get_cluster_from_dependency(const Dependency& data,
std::vector<int>& cluster6,
std::vector<int>& cluster) {
if (use_cluster) {
size_t L = data.forms.size();
size_t L = data.size();
for (size_t i = 0; i < L; ++ i) {
int form = data.forms[i];
cluster4.push_back(i == 0?
Expand Down
3 changes: 2 additions & 1 deletion src/parser.n/parser_frontend.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -129,6 +129,7 @@ void NeuralNetworkParserFrontend::build_cluster(void) {
}

std::string line;
//TODO 构造函数 LineCountsReader 会计算 interval，此处的计算重复了
size_t interval = LineCountsReader(ifs).number_of_lines() / 10;
if (interval == 0) { interval = 1; }
size_t nr_lines = 1;
Expand Down Expand Up @@ -274,7 +275,7 @@ void NeuralNetworkParserFrontend::initialize_classifier() {
if (form == -1) { continue; }

if (items.size() != learn_opt->embedding_size + 1) {
WARNING_LOG("report: embedding dimension not match to configuration.");
WARNING_LOG("report: line %d, embedding dimension(%d) not match to configuration.", nr_lines, items.size()-1);
continue;
}

Expand Down
2 changes: 1 addition & 1 deletion src/parser.n/parser_frontend.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@

#include <iostream>

#include "framework/frontend.h"
//#include "framework/frontend.h"
#include "parser.n/options.h"
#include "parser.n/instance.h"
#include "parser.n/parser.h"
Expand Down

0 comments on commit 7eae6a7

Please sign in to comment.