minor update

.gitignore

@@ -1,3 +1,4 @@
 *.py[cod]
 __pycache__
 *.dset
+*.model

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: 2019-01-02 00:35:39
+# @Last Modified time: 2019-01-10 14:53:57
 
 from __future__ import print_function
 from __future__ import absolute_import
@@ -42,7 +42,9 @@ def neg_log_likelihood_loss(self, word_inputs, feature_inputs, word_seq_lengths,
         _, tag_seq  = torch.max(score, 1)
         if self.average_batch:
             total_loss = total_loss / batch_size
+        print("aa")
         print(total_loss)
+        exit(0)
         return total_loss, tag_seq
 
 

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: 2019-01-02 00:37:16
+# @Last Modified time: 2019-01-10 15:01:16
 from __future__ import print_function
 from __future__ import absolute_import
 import torch
@@ -121,7 +121,7 @@ def sentence_representation(self, word_inputs, feature_inputs, word_seq_lengths,
                 Variable(batch_size, sent_len, hidden_dim)
         """
         word_represent = self.wordrep(word_inputs, feature_inputs, word_seq_lengths, char_inputs, char_seq_lengths, char_seq_recover)
-        print("b",word_represent)
+        # print("b",word_represent)
         ## word_embs (batch_size, seq_len, embed_size)
         batch_size = word_inputs.size(0)
         if self.word_feature_extractor == "CNN":
@@ -131,9 +131,10 @@ def sentence_representation(self, word_inputs, feature_inputs, word_seq_lengths,
                     cnn_feature = F.relu(self.cnn_list[idx](word_in))
                 else:
                     cnn_feature = F.relu(self.cnn_list[idx](cnn_feature))
+                # print("cnn: %s"%idx, cnn_feature)
                 cnn_feature = self.cnn_drop_list[idx](cnn_feature)
                 cnn_feature = self.cnn_batchnorm_list[idx](cnn_feature)
-                print("a", cnn_feature)
+                # print("a", cnn_feature)
             feature_out = F.max_pool1d(cnn_feature, cnn_feature.size(2)).view(batch_size, -1)
             print(feature_out)
             exit(0)

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: 2019-01-01 23:44:28
+# @Last Modified time: 2019-01-10 15:03:31
 from __future__ import print_function
 from __future__ import absolute_import
 import sys
@@ -81,6 +81,16 @@ def read_instance(input_file, word_alphabet, char_alphabet, feature_alphabets, l
                 word_Ids = []
                 feature_Ids = []
                 label_Ids = []
+        if (len(words) > 0) and ((max_sent_length < 0) or (len(words) < max_sent_length)) :
+            instence_texts.append([words, feat_list, chars, label])
+            instence_Ids.append([word_Ids, feat_Id, char_Ids,label_Id])
+            words = []
+            features = []
+            chars = []
+            char_Ids = []
+            word_Ids = []
+            feature_Ids = []
+            label_Ids = []
 
     else:
     ### for sequence labeling data format i.e. CoNLL 2003
@@ -136,6 +146,10 @@ def read_instance(input_file, word_alphabet, char_alphabet, feature_alphabets, l
                 feature_Ids = []
                 char_Ids = []
                 label_Ids = []
+        if (len(words) > 0) and ((max_sent_length < 0) or (len(words) < max_sent_length)) :
+            instence_texts.append([words, features, chars, labels])
+            instence_Ids.append([word_Ids, feature_Ids, char_Ids,label_Ids])
+
     return instence_texts, instence_Ids