fix HIT-SCIR#112 HIT-SCIR#113

2php · Jun 3, 2015 · 377e8aa · 377e8aa
1 parent 96ed927
commit 377e8aa
Show file tree

Hide file tree

Showing 5 changed files with 64 additions and 101 deletions.
diff --git a/src/ner/ner_frontend.cpp b/src/ner/ner_frontend.cpp
@@ -41,7 +41,7 @@ NamedEntityRecognizerFrontend::NamedEntityRecognizerFrontend(
   INFO_LOG("||| ltp ner, trainig ...");
   INFO_LOG("report: reference file = %s", train_opt.train_file.c_str());
   INFO_LOG("report: holdout file = %s", train_opt.holdout_file.c_str());
-  INFO_LOG("report: algorith = %s", train_opt.algorithm.c_str());
+  INFO_LOG("report: algorithm = %s", train_opt.algorithm.c_str());
   INFO_LOG("report: model name = %s", train_opt.model_name.c_str());
   INFO_LOG("report: maximum iteration = %d", train_opt.max_iter);
   INFO_LOG("report: rare threshold = %d", train_opt.rare_feature_threshold);
@@ -157,6 +157,7 @@ void NamedEntityRecognizerFrontend::build_feature_space(void) {
   model->space.set_num_labels(L);
 
   size_t interval = train_dat.size() / 10;
+  if (0 == interval) { interval = 1; }
   for (size_t i = 0; i < train_dat.size(); ++ i) {
     NamedEntityRecognizer::extract_features((*train_dat[i]), NULL, true);
     if ((i+ 1) % interval == 0) {
@@ -214,6 +215,7 @@ void NamedEntityRecognizerFrontend::train(void) {
     INFO_LOG("Training iteration #%d", (iter + 1));
 
     size_t interval = train_dat.size() / 10;
+    if (interval == 0) { interval = 1; }
     for (size_t i = 0; i < train_dat.size(); ++ i) {
       Instance* inst = train_dat[i];
       NamedEntityRecognizer::extract_features((*inst), &ctx, false);
@@ -252,18 +254,19 @@ void NamedEntityRecognizerFrontend::train(void) {
     double f_score;
     evaluate(f_score);
 
+    std::swap(model, new_model);
+
     if(f_score > best_f_score){
       best_f_score = f_score;
       best_iteration = iter;
-    }
 
-    std::string saved_model_file = (train_opt.model_name+ "."+ to_str(iter));
-    std::ofstream ofs(saved_model_file.c_str(), std::ofstream::binary);
-    std::swap(model, new_model);
-    new_model->save(model_header, Parameters::kDumpAveraged, ofs);
+      std::ofstream ofs(train_opt.model_name.c_str(), std::ofstream::binary);
+      new_model->save(model_header, Parameters::kDumpAveraged, ofs);
+      INFO_LOG("trace: model for iteration #%d is saved to %s",
+          iter+1, train_opt.model_name.c_str());
+    }
     delete new_model;
 
-    INFO_LOG("trace: model for iteration #%d is saved to %s", iter+1, saved_model_file.c_str());
   }
   INFO_LOG("Best result (iteration = %d) : F-score = %lf", best_iteration, best_f_score);
 }

diff --git a/src/parser.n/parser_frontend.cpp b/src/parser.n/parser_frontend.cpp
@@ -130,6 +130,7 @@ void NeuralNetworkParserFrontend::build_cluster(void) {
 
   std::string line;
   size_t interval = LineCountsReader(ifs).number_of_lines() / 10;
+  if (interval == 0) { interval = 1; }
   size_t nr_lines = 1;
 
   while (std::getline(ifs, line)) {
@@ -176,7 +177,7 @@ void NeuralNetworkParserFrontend::collect_precomputed_features() {
 
   size_t nr_processed = 0;
   nr_feature_types = 0;
-  size_t interval = train_dat.size() / 10; if (interval == 0) { interval = 10; }
+  size_t interval = train_dat.size() / 10; if (interval == 0) { interval = 1; }
   for (size_t d = 0; d < train_dat.size(); ++ d) {
     Instance* inst = train_dat[d];
     if (!inst->is_tree() || !inst->is_projective()) { continue; }

diff --git a/src/postagger/postagger_frontend.cpp b/src/postagger/postagger_frontend.cpp
@@ -39,7 +39,7 @@ PostaggerFrontend::PostaggerFrontend(const std::string& reference_file,
   INFO_LOG("||| ltp postagger, training ...");
   INFO_LOG("report: reference file = %s", train_opt.train_file.c_str());
   INFO_LOG("report: holdout file = %s", train_opt.holdout_file.c_str());
-  INFO_LOG("report: algorith = %s", train_opt.algorithm.c_str());
+  INFO_LOG("report: algorithm = %s", train_opt.algorithm.c_str());
   INFO_LOG("report: model name = %s", train_opt.model_name.c_str());
   INFO_LOG("report: maximum iteration = %d", train_opt.max_iter);
   INFO_LOG("report: rare threshold = %d", train_opt.rare_feature_threshold);
@@ -111,6 +111,7 @@ void PostaggerFrontend::build_feature_space(void) {
   model->space.set_num_labels(L);
 
   size_t interval = train_dat.size() / 10;
+  if (interval == 0) { interval = 1; }
   for (size_t i = 0; i < train_dat.size(); ++ i) {
     Postagger::extract_features((*train_dat[i]), NULL, true);
     if ((i+ 1) % interval == 0) {
@@ -163,6 +164,7 @@ void PostaggerFrontend::train(void) {
     INFO_LOG("Training iteraition #%d", (iter + 1));
 
     size_t interval= train_dat.size() / 10;
+    if (interval == 0) { interval = 1; }
     for (size_t i = 0; i < train_dat.size(); ++ i) {
       Instance* inst = train_dat[i];
       extract_features((*inst), &ctx, false);
@@ -199,19 +201,20 @@ void PostaggerFrontend::train(void) {
     std::swap(model, new_model);
     double p;
     evaluate(p);
+    std::swap(model, new_model);
 
     if(p > best_p){
       best_p = p;
       best_iteration = iter;
+
+      std::ofstream ofs(train_opt.model_name.c_str(), std::ofstream::binary);
+      new_model->save(model_header, Parameters::kDumpAveraged, ofs);
+
+      INFO_LOG("trace: model for iteration #%d is saved to %s",
+          iter+1, train_opt.model_name.c_str());
     }
 
-    std::string saved_model_file = (train_opt.model_name+ "."+ to_str(iter));
-    std::ofstream ofs(saved_model_file.c_str(), std::ofstream::binary);
-    std::swap(model, new_model);
-    new_model->save(model_header, Parameters::kDumpAveraged, ofs);
     delete new_model;
-
-    INFO_LOG("trace: model for iteration #%d is saved to %s", iter+1, saved_model_file.c_str());
   }
 
   INFO_LOG("Best result (iteration = %d) : P = %lf", best_iteration, best_p);

diff --git a/src/segmentor/segmentor_frontend.cpp b/src/segmentor/segmentor_frontend.cpp
@@ -45,7 +45,7 @@ SegmentorFrontend::SegmentorFrontend(const std::string& reference_file,
   INFO_LOG("||| ltp segmentor, training ...");
   INFO_LOG("report: reference file = %s", train_opt.train_file.c_str());
   INFO_LOG("report: holdout file = %s", train_opt.holdout_file.c_str());
-  INFO_LOG("report: algorith = %s", train_opt.algorithm.c_str());
+  INFO_LOG("report: algorithm = %s", train_opt.algorithm.c_str());
   INFO_LOG("report: model name = %s", train_opt.model_name.c_str());
   INFO_LOG("report: maximum iteration = %d", train_opt.max_iter);
   INFO_LOG("report: rare threshold = %d", train_opt.rare_feature_threshold);
@@ -176,6 +176,7 @@ void SegmentorFrontend::build_feature_space(void) {
   model->space.set_num_labels(L);
 
   size_t interval = train_dat.size() / 10;
+  if (0 == interval) { interval = 1; }
 
   for (size_t i = 0; i < train_dat.size(); ++ i) {
     build_lexicon_match_state(lexicons, train_dat[i]);
@@ -261,6 +262,7 @@ void SegmentorFrontend::train(void) {
     INFO_LOG("Training iteration #%d", (iter + 1));
 
     size_t interval = train_dat.size()/ 10;
+    if (interval == 0) { interval = 1; }
     for (size_t i = 0; i < train_dat.size(); ++ i) {
       increase_timestamp();
 
@@ -270,7 +272,6 @@ void SegmentorFrontend::train(void) {
 
       con.regist(&(inst->chartypes));
       decoder.decode(scm, con, inst->predict_tagsidx);
-      //decoder->decode((*scm), inst->predict_tagsidx);
 
       collect_features((*inst));
 
@@ -303,21 +304,20 @@ void SegmentorFrontend::train(void) {
     double p, r, f;
     evaluate(p,r,f);
 
+    std::swap(model, new_model);
+
     if (f > best_f) {
       best_p = p; best_r = r; best_f = f;
       best_iteration = iter;
-    }
 
-    std::string saved_model_file = (train_opt.model_name + "." + strutils::to_str(iter));
-    std::ofstream ofs(saved_model_file.c_str(), std::ofstream::binary);
-
-    std::swap(model, new_model);
-    new_model->save(model_header.c_str(),
-        (train_opt.dump_model_details? Parameters::kDumpDetails: Parameters::kDumpAveraged),
-        ofs);
+      std::ofstream ofs(train_opt.model_name.c_str(), std::ofstream::binary);
+      new_model->save(model_header.c_str(),
+          (train_opt.dump_model_details? Parameters::kDumpDetails: Parameters::kDumpAveraged),
+          ofs);
 
+      INFO_LOG("Model for iteration #%d is saved to [%s]", iter+1, train_opt.model_name.c_str());
+    }
     delete new_model;
-    INFO_LOG("Model for iteration #%d is saved to [%s]", iter+1, saved_model_file.c_str());
   }
 
   INFO_LOG("Best result (iteratin = %d) P = %lf | R = %lf | F = %lf",
@@ -351,7 +351,7 @@ void SegmentorFrontend::evaluate(double &p, double &r, double &f) {
     build_lexicon_match_state(lexicons, inst);
     extract_features((*inst), false);
     calculate_scores((*inst), true);
- 
+
     con.regist(&(inst->chartypes));
     decoder.decode(scm, con, inst->predict_tagsidx);
     ctx.clear();

diff --git a/tools/train/rock.sh b/tools/train/rock.sh
@@ -6,7 +6,6 @@
 
 ROOT=
 BUILD_DIR=build
-CONF_DIR=conf
 LOG_DIR=log
 
 # create model output folder
@@ -28,10 +27,7 @@ mkdir -p $LOG_DIR
 
 # cws dir config
 CWS_MODEL_DIR=$BUILD_DIR/cws
-CWS_MODEL_PATH=$CWS_MODEL_DIR/example-seg.0.model
-
-CWS_CONF_DIR=$CONF_DIR/cws
-CWS_CONF_TRAIN_PATH=$CWS_CONF_DIR/cws.cnf
+CWS_MODEL_PATH=$CWS_MODEL_DIR/example-seg.model
 
 CWS_LOG_DIR=$LOG_DIR/cws
 CWS_LOG_TRAIN_PATH=$CWS_LOG_DIR/example-seg.train.log
@@ -43,9 +39,12 @@ mkdir -p $CWS_MODEL_DIR
 mkdir -p $CWS_LOG_DIR
 
 # execute the example training process 
-$CWS_EXE $CWS_CONF_TRAIN_PATH >& $CWS_LOG_TRAIN_PATH
+$CWS_EXE learn --model $CWS_MODEL_PATH \
+    --reference sample/seg/example-train.seg \
+    --development sample/seg/example-holdout.seg \
+    --max-iter 1
 
-if [ ! -f $CWS_MODEL_PATH ]; then 
+if [ ! -f $CWS_MODEL_PATH ]; then
     echo "[1] ERROR: CWS model is not detected!"
 else
     echo "[1] TRACE: CWS train model test is passed."
@@ -56,10 +55,7 @@ fi
 #################################################
 
 POS_MODEL_DIR=$BUILD_DIR/pos
-POS_MODEL_PATH=$POS_MODEL_DIR/example-pos.0.model
-
-POS_CONF_DIR=$CONF_DIR/pos
-POS_CONF_TRAIN_PATH=$POS_CONF_DIR/pos.cnf
+POS_MODEL_PATH=$POS_MODEL_DIR/example-pos.model
 
 POS_LOG_DIR=$LOG_DIR/pos
 POS_LOG_TRAIN_PATH=$CWS_LOG_DIR/example-pos.train.log
@@ -70,7 +66,10 @@ POS_EXE=./otpos
 mkdir -p $POS_MODEL_DIR
 mkdir -p $POS_LOG_DIR
 
-$POS_EXE $POS_CONF_TRAIN_PATH >& $POS_LOG_TRAIN_PATH
+$POS_EXE learn --model $POS_MODEL_PATH \
+    --reference sample/pos/example-train.pos \
+    --development sample/pos/example-holdout.pos \
+    --max-iter 1
 
 if [ ! -f $CWS_MODEL_PATH ]; then 
     echo "[2] ERROR: POS model is not detected!"
@@ -84,10 +83,7 @@ fi
 
 # ner dir config
 NER_MODEL_DIR=$BUILD_DIR/ner
-NER_MODEL_PATH=$NER_MODEL_DIR/example-ner.0.model
-
-NER_CONF_DIR=$CONF_DIR/ner
-NER_CONF_TRAIN_PATH=$NER_CONF_DIR/ner.cnf
+NER_MODEL_PATH=$NER_MODEL_DIR/example-ner.model
 
 NER_LOG_DIR=$LOG_DIR/ner
 NER_LOG_TRAIN_PATH=$NER_LOG_DIR/example-ner.train.log
@@ -99,7 +95,10 @@ mkdir -p $NER_MODEL_DIR
 mkdir -p $NER_LOG_DIR
 
 # execute the example training process 
-$NER_EXE $NER_CONF_TRAIN_PATH >& $NER_LOG_TRAIN_PATH
+$NER_EXE learn --model $NER_MODEL_PATH \
+    --reference sample/ner/example-train.ner \
+    --development sample/ner/example-holdout.ner \
+    --max-iter 1
 
 if [ ! -f $NER_MODEL_PATH ]; then 
     echo "[3] ERROR: NER model is not detected!"
@@ -112,71 +111,28 @@ fi
 #################################################
 
 PARSER_MODEL_DIR=$BUILD_DIR/parser
-PARSER_MODEL_O1_PATH=$PARSER_MODEL_DIR/example-parser-o1.0.model
-
-PARSER_CONF_DIR=$CONF_DIR/parser
-PARSER_CONF_TRAIN_O1_PATH=$PARSER_CONF_DIR/parser-o1.cnf
+PARSER_MODEL_PATH=$PARSER_MODEL_DIR/example-parser.model
 
 PARSER_LOG_DIR=$LOG_DIR/parser
-PARSER_LOG_TRAIN_O1_PATH=$PARSER_LOG_DIR/example-train.conll
+PARSER_LOG_TRAIN_PATH=$PARSER_LOG_DIR/example-train.conll
 
-PARSER_EXE=./lgdpj
+PARSER_EXE=./nndepparser
 
 mkdir -p $PARSER_MODEL_DIR
 mkdir -p $PARSER_LOG_DIR
 
-$PARSER_EXE $PARSER_CONF_TRAIN_O1_PATH >& $PARSER_LOG_TRAIN_O1_PATH
-
-if [ ! -f $PARSER_MODEL_O1_PATH ]; then 
-    echo "[4.1] ERROR: Parser-o1 model is not detected!"
-else
-    echo "[4.1] TRACE: Parser-o1 train model test is passed."
-fi
-
-#################################################
-# THE PARSER O2 sibling SESSION                 #
-#################################################
-
-PARSER_MODEL_DIR=$BUILD_DIR/parser
-PARSER_MODEL_O2SIB_PATH=$PARSER_MODEL_DIR/example-parser-o2sib.0.model
-
-PARSER_CONF_DIR=$CONF_DIR/parser
-PARSER_CONF_TRAIN_O2SIB_PATH=$PARSER_CONF_DIR/parser-o2sib.cnf
-
-PARSER_LOG_DIR=$LOG_DIR/parser
-PARSER_LOG_TRAIN_O2SIB_PATH=$PARSER_LOG_DIR/example-train.conll
-
-PARSER_EXE=./lgdpj
-
-$PARSER_EXE $PARSER_CONF_TRAIN_O2SIB_PATH >& $PARSER_LOG_TRAIN_O2SIB_PATH
-
-if [ ! -f $PARSER_MODEL_O2SIB_PATH ]; then 
-    echo "[4.2] ERROR: Parser-o2sib model is not detected!"
-else
-    echo "[4.2] TRACE: Parser-o2sib train model test is passed."
-fi
-
-#################################################
-# THE PARSER O2 carreras SESSION                #
-#################################################
-
-PARSER_MODEL_DIR=$BUILD_DIR/parser
-PARSER_MODEL_O2CARRERAS_PATH=$PARSER_MODEL_DIR/example-parser-o2carreras.0.model
-
-PARSER_CONF_DIR=$CONF_DIR/parser
-PARSER_CONF_TRAIN_O2CARRERAS_PATH=$PARSER_CONF_DIR/parser-o2carreras.cnf
-
-PARSER_LOG_DIR=$LOG_DIR/parser
-PARSER_LOG_TRAIN_O2CARRERAS_PATH=$PARSER_LOG_DIR/example-train.conll
-
-PARSER_EXE=./lgdpj
-
-$PARSER_EXE $PARSER_CONF_TRAIN_O2CARRERAS_PATH >& $PARSER_LOG_TRAIN_O2CARRERAS_PATH
+./nndepparser learn \
+    --model $PARSER_MODEL_PATH \
+    --reference sample/parser/example-train.conll \
+    --development sample/parser/example-holdout.conll \
+    --embedding sample/parser/example.bin \
+    --root HED \
+    --max-iter 10
 
 if [ ! -f $PARSER_MODEL_O2SIB_PATH ]; then 
-    echo "[4.3] ERROR: Parser-o2carreras model is not detected!"
+    echo "[4] ERROR: neural network parser model is not detected!"
 else
-    echo "[4.3] TRACE: Parser-o2carreras train model test is passed."
+    echo "[4] TRACE: neural network parser model test is passed."
 fi
 
 #################################################
@@ -199,7 +155,7 @@ mkdir -p $SRL_PRG_INSTANCE_DIR
 
 SRL_PRG_EXE=./lgsrl
 
-$SRL_PRG_EXE $SRL_PRG_CONF_TRAIN_PATH >& $SRL_PRG_LOG_TRAIN_PATH
+$SRL_PRG_EXE $SRL_PRG_CONF_TRAIN_PATH
 
 if [ ! -f $SRL_PRG_MODEL_PATH ]; then
     echo "[5.1] ERROR: PRG model is not detected!"
@@ -229,7 +185,7 @@ mkdir -p $SRL_SRL_INSTANCE_DIR
 
 SRL_SRL_EXE=./lgsrl
 
-$SRL_SRL_EXE $SRL_SRL_CONF_TRAIN_PATH >& $SRL_SRL_LOG_TRAIN_PATH
+$SRL_SRL_EXE $SRL_SRL_CONF_TRAIN_PATH
 
 if [ ! -f $SRL_SRL_MODEL_PATH ]; then
     echo "[5.2] ERROR: SRL model is not detected!"