release code of ctunet

Author: qiaoliang6

davarocr/davar_ie/datasets/pipelines/tokenizer.py

@@ -6,9 +6,10 @@
 
 # Current Version:    1.0.0
 # Date           :    2020-05-31
+# Current Version:    1.0.1
+# Date           :    2022-12-12
 ##################################################################################################
 """
-import os
 import copy
 import numpy as np
 
@@ -18,6 +19,7 @@
 @PIPELINES.register_module()
 class CharPadTokenize():
     """Tokenize texts in characters and return their indexes ( padded if required)."""
+
     def __init__(self, vocab, targets, max_length=None, map_target_prefix=None):
         """
         Args:
@@ -32,15 +34,18 @@ def __init__(self, vocab, targets, max_length=None, map_target_prefix=None):
         self.max_length = max_length
         self.map_target_prefix = map_target_prefix
 
-        if os.path.exists(self.vocab):
+        if self.vocab is not None:
             with open(self.vocab, 'r', encoding='utf8') as read_f:
                 all_words = read_f.readline().strip()
-            default_token = ['[PAD]', '[UNK]']
             all_words = list(all_words)
-            self.character = default_token + all_words
+        else:
+            all_words = []
+
+        default_token = ['[PAD]', '[UNK]']
+        self.character = default_token + all_words
 
-            # default 0 to pad
-            self.word2idx = {char: idx for idx, char in enumerate(self.character)}
+        # default 0 to pad
+        self.word2idx = {char: idx for idx, char in enumerate(self.character)}
 
     def __call__(self, results):
         """Forward process, including tokenization and (optional) padding.
@@ -60,14 +65,14 @@ def __call__(self, results):
                 # pad to max length if required
                 if self.max_length is not None:
                     if len(tmp_per_line) > self.max_length:
-                        per_target_token = tmp_per_line[:self.max_length]
+                        tmp_per_line = tmp_per_line[:self.max_length]
                     else:
-                        tmp_per_line.extend([self.word2idx['[PAD]']]*(self.max_length - len(tmp_per_line)))
+                        tmp_per_line.extend([self.word2idx['[PAD]']] * (self.max_length - len(tmp_per_line)))
                 per_target_token.append(np.array(tmp_per_line))
 
             # add map_target to results if required.
             if self.map_target_prefix is not None:
-                results[self.map_target_prefix+key] = np.array(per_target_token)
+                results[self.map_target_prefix + key] = np.array(per_target_token)
             else:
                 results[key] = np.array(per_target_token)
 

davarocr/davar_ie/models/connects/__init__.py

@@ -5,12 +5,13 @@
 # Abstract       :
 
 # Current Version:    1.0.0
-# Date           :    2020-05-31
+# Date           :    2022-11-22
 ##################################################################################################
 """
 from .util import BertConfig
 from .multimodal_feature_merge import MultiModalFusion
 from .multimodal_context_module import MultiModalContextModule
+from .multimodal_context_module_plusplus import MultiModalContextModulePlusPlus
 from .relation_module import BertEncoder
 
-__all__ = ['MultiModalFusion', 'MultiModalContextModule', 'BertConfig', 'BertEncoder']
+__all__ = ['MultiModalFusion', 'MultiModalContextModule', 'BertConfig', 'BertEncoder', 'MultiModalContextModulePlusPlus']