optimize code

.gitignore

@@ -6,3 +6,4 @@ __pycache__
 demo.clf.*
 sent.*
 *.out
+*.log

demo.clf.config

@@ -4,10 +4,11 @@ sentence_classification=True
 
 ### I/O ###
 train_dir=../data/Sentclf/SST1/stsa.fine.train.clf
-dev_dir=../data/Sentclf/SST1/stsa.fine.train.clf
+dev_dir=../data/Sentclf/SST1/stsa.fine.dev.clf
 test_dir=../data/Sentclf/SST1/stsa.fine.test.clf
 model_dir=sample_data/clf
-word_emb_dir=../data/glove.6B.100d.txt
+word_emb_dir=../data/glove.840B.300d.txt
+
 
 #raw_dir=
 #decode_dir=
@@ -33,19 +34,19 @@ char_seq_feature=CNN
 
 ###TrainingSetting###
 status=train
-optimizer=Adadelta
-iteration=10
+optimizer=SGD
+iteration=50
 batch_size=10
 ave_batch_loss=False
 
 ###Hyperparameters###
 cnn_layer=4
 char_hidden_dim=50
 hidden_dim=400
-dropout=0.5
+dropout=0
 lstm_layer=1
 bilstm=True
-learning_rate=1
+learning_rate=0.2
 lr_decay=0.05
 momentum=0
 l2=1e-8

main.py

@@ -2,7 +2,7 @@
 # @Author: Jie
 # @Date:   2017-06-15 14:11:08
 # @Last Modified by:   Jie Yang,     Contact: [email protected]
-# @Last Modified time: 2019-01-14 16:09:16
+# @Last Modified time: 2019-02-13 12:41:44
 
 from __future__ import print_function
 import time
@@ -51,6 +51,8 @@ def predict_check(pred_variable, gold_variable, mask_variable, sentence_classifi
     mask = mask_variable.cpu().data.numpy()
     overlaped = (pred == gold)
     if sentence_classification:
+        # print(overlaped)
+        # print(overlaped*pred)
         right_token = np.sum(overlaped)
         total_token = overlaped.shape[0] ## =batch_size
     else:
@@ -359,7 +361,7 @@ def train(data):
         model = SentClassifier(data)
     else:
         model = SeqLabel(data)
-    # loss_function = nn.NLLLoss()
+
     if data.optimizer.lower() == "sgd":
         optimizer = optim.SGD(model.parameters(), lr=data.HP_lr, momentum=data.HP_momentum,weight_decay=data.HP_l2)
     elif data.optimizer.lower() == "adagrad":
@@ -407,7 +409,7 @@ def train(data):
                 continue
             batch_word, batch_features, batch_wordlen, batch_wordrecover, batch_char, batch_charlen, batch_charrecover, batch_label, mask  = batchify_with_label(instance, data.HP_gpu, True, data.sentence_classification)
             instance_count += 1
-            loss, tag_seq = model.neg_log_likelihood_loss(batch_word, batch_features, batch_wordlen, batch_char, batch_charlen, batch_charrecover, batch_label, mask)
+            loss, tag_seq = model.calculate_loss(batch_word, batch_features, batch_wordlen, batch_char, batch_charlen, batch_charrecover, batch_label, mask)
             right, whole = predict_check(tag_seq, batch_label, mask, data.sentence_classification)
             right_token += right
             whole_token += whole
@@ -504,13 +506,41 @@ def load_model_decode(data, name):
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(description='Tuning with NCRF++')
     # parser.add_argument('--status', choices=['train', 'decode'], help='update algorithm', default='train')
-    parser.add_argument('--config',  help='Configuration File' )
+    parser.add_argument('--config',  help='Configuration File', default='None')
+    parser.add_argument('--wordemb',  help='Embedding for words', default='None')
+    parser.add_argument('--charemb',  help='Embedding for chars', default='None')
+    parser.add_argument('--status', choices=['train', 'decode'], help='update algorithm', default='train')
+    parser.add_argument('--savemodel', default="data/model/saved_model.lstmcrf.")
+    parser.add_argument('--savedset', help='Dir of saved data setting')
+    parser.add_argument('--train', default="data/conll03/train.bmes") 
+    parser.add_argument('--dev', default="data/conll03/dev.bmes" )  
+    parser.add_argument('--test', default="data/conll03/test.bmes") 
+    parser.add_argument('--seg', default="True") 
+    parser.add_argument('--raw') 
+    parser.add_argument('--loadmodel')
+    parser.add_argument('--output') 
 
     args = parser.parse_args()
     data = Data()
     data.HP_gpu = torch.cuda.is_available()
-    data.read_config(args.config)
-    data.show_data_summary()
+    if args.config == 'None':
+        data.train_dir = args.train 
+        data.dev_dir = args.dev 
+        data.test_dir = args.test
+        data.model_dir = args.savemodel
+        data.dset_dir = args.savedset
+        print("Save dset directory:",data.dset_dir)
+        save_model_dir = args.savemodel
+        data.word_emb_dir = args.wordemb
+        data.char_emb_dir = args.charemb
+        if args.seg.lower() == 'true':
+            data.seg = True
+        else:
+            data.seg = False
+        print("Seed num:",seed_num)
+    else:
+        data.read_config(args.config)
+    # data.show_data_summary()
     status = data.status.lower()
     print("Seed num:",seed_num)
 

main_parse.py

@@ -2,7 +2,7 @@
 # @Author: Jie
 # @Date:   2017-06-15 14:11:08
 # @Last Modified by:   Jie Yang,     Contact: [email protected]
-# @Last Modified time: 2019-01-18 21:02:01
+# @Last Modified time: 2019-02-13 10:58:43
 
 from __future__ import print_function
 import time
@@ -434,7 +434,7 @@ def load_model_decode(data, name):
     data.test_dir = args.test
     data.model_dir = args.savemodel
     data.dset_dir = args.savedset
-    print("aaa",data.dset_dir)
+    print("dset directory:",data.dset_dir)
     status = args.status.lower()
     save_model_dir = args.savemodel
     data.HP_gpu = torch.cuda.is_available()

model/sentclassifier.py

@@ -2,7 +2,7 @@
 # @Author: Jie Yang
 # @Date:   2019-01-01 21:11:50
 # @Last Modified by:   Jie Yang,     Contact: [email protected]
-# @Last Modified time: 2019-01-14 14:56:28
+# @Last Modified time: 2019-02-13 12:30:56
 
 from __future__ import print_function
 from __future__ import absolute_import
@@ -27,7 +27,7 @@ def __init__(self, data):
 
 
 
-    def neg_log_likelihood_loss(self, word_inputs, feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover, batch_label, mask):
+    def calculate_loss(self, word_inputs, feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover, batch_label, mask):
         outs = self.word_hidden.sentence_representation(word_inputs,feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover)
         batch_size = word_inputs.size(0)
         # loss_function = nn.CrossEntropyLoss(ignore_index=0, reduction='sum')
@@ -40,6 +40,7 @@ def neg_log_likelihood_loss(self, word_inputs, feature_inputs, word_seq_lengths,
         # exit(0)
         total_loss = F.cross_entropy(outs, batch_label.view(batch_size))
         # total_loss = loss_function(score, batch_label.view(batch_size))
+
         _, tag_seq  = torch.max(outs, 1)
         if self.average_batch:
             total_loss = total_loss / batch_size
@@ -51,6 +52,8 @@ def forward(self, word_inputs, feature_inputs, word_seq_lengths, char_inputs, ch
         batch_size = word_inputs.size(0)
         outs = outs.view(batch_size, -1)
         _, tag_seq  = torch.max(outs, 1)
+        # if a == 0:
+        #     print(tag_seq)
         return tag_seq
 
 

model/seqlabel.py

@@ -2,7 +2,7 @@
 # @Author: Jie Yang
 # @Date:   2017-10-17 16:47:32
 # @Last Modified by:   Jie Yang,     Contact: [email protected]
-# @Last Modified time: 2019-01-01 21:10:00
+# @Last Modified time: 2019-02-13 11:49:38
 
 from __future__ import print_function
 from __future__ import absolute_import
@@ -33,7 +33,7 @@ def __init__(self, data):
             self.crf = CRF(label_size, self.gpu)
 
 
-    def neg_log_likelihood_loss(self, word_inputs, feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover, batch_label, mask):
+    def calculate_loss(self, word_inputs, feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover, batch_label, mask):
         outs = self.word_hidden(word_inputs,feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover)
         batch_size = word_inputs.size(0)
         seq_len = word_inputs.size(1)

model/wordrep.py

@@ -2,7 +2,7 @@
 # @Author: Jie Yang
 # @Date:   2017-10-17 16:47:32
 # @Last Modified by:   Jie Yang,     Contact: [email protected]
-# @Last Modified time: 2019-01-10 16:41:16
+# @Last Modified time: 2019-02-01 15:52:01
 from __future__ import print_function
 from __future__ import absolute_import
 import torch
@@ -87,7 +87,9 @@ def forward(self, word_inputs,feature_inputs, word_seq_lengths, char_inputs, cha
         """
         batch_size = word_inputs.size(0)
         sent_len = word_inputs.size(1)
+
         word_embs =  self.word_embedding(word_inputs)
+
         word_list = [word_embs]
         if not self.sentence_classification:
             for idx in range(self.feature_num):
@@ -109,5 +111,8 @@ def forward(self, word_inputs,feature_inputs, word_seq_lengths, char_inputs, cha
                 ## concat word and char together
                 word_list.append(char_features_extra)    
         word_embs = torch.cat(word_list, 2)
+        # if a == 0:
+        #     print("inputs", word_inputs)
+        #     print("embeddings:", word_embs)
         word_represent = self.drop(word_embs)
         return word_represent

model/wordsequence.py

@@ -2,7 +2,7 @@
 # @Author: Jie Yang
 # @Date:   2017-10-17 16:47:32
 # @Last Modified by:   Jie Yang,     Contact: [email protected]
-# @Last Modified time: 2019-01-11 13:55:41
+# @Last Modified time: 2019-02-01 15:59:26
 from __future__ import print_function
 from __future__ import absolute_import
 import torch

utils/functions.py

@@ -2,7 +2,7 @@
 # @Author: Jie
 # @Date:   2017-06-15 14:23:06
 # @Last Modified by:   Jie Yang,     Contact: [email protected]
-# @Last Modified time: 2019-01-14 11:08:45
+# @Last Modified time: 2019-02-14 12:23:52
 from __future__ import print_function
 from __future__ import absolute_import
 import sys
@@ -204,8 +204,10 @@ def load_pretrain_emb(embedding_path):
             tokens = line.split()
             if embedd_dim < 0:
                 embedd_dim = len(tokens) - 1
-            else:
-                assert (embedd_dim + 1 == len(tokens))
+            elif embedd_dim + 1 != len(tokens):
+                ## ignore illegal embedding line
+                continue
+                # assert (embedd_dim + 1 == len(tokens))
             embedd = np.empty([1, embedd_dim])
             embedd[:] = tokens[1:]
             if sys.version_info[0] < 3:

utils/metric.py

@@ -2,7 +2,7 @@
 # @Author: Jie
 # @Date:   2017-02-16 09:53:19
 # @Last Modified by:   Jie Yang,     Contact: [email protected]
-# @Last Modified time: 2019-01-18 21:04:10
+# @Last Modified time: 2019-02-17 22:46:59
 
 # from operator import add
 #
@@ -27,7 +27,7 @@ def get_ner_fmeasure(golden_lists, predict_lists, label_type="BMES"):
             if golden_list[idy] == predict_list[idy]:
                 right_tag += 1
         all_tag += len(golden_list)
-        if label_type == "BMES" or "BIOES":
+        if label_type == "BMES" or label_type == "BIOES":
             gold_matrix = get_ner_BMES(golden_list)
             pred_matrix = get_ner_BMES(predict_list)
         else: