A few fixes to improve doc

doc/source/doc_util.py

@@ -59,7 +59,7 @@ def get_func_name(func_str):
     return name
 
 
-def create_doc_copy(in_file='../../python/_dynet.pyx', out_file='dynet.py'):
+def create_doc_copy(in_file, out_file):
 
     in_comment = False
     in_func = False
@@ -68,7 +68,7 @@ def create_doc_copy(in_file='../../python/_dynet.pyx', out_file='dynet.py'):
         with open(in_file, 'r') as pyx:
             for l in pyx:
                 # Check if this line is a function declaration (def or cpdef)
-                is_func = re.match(r'(\s*)(?:cp)?def (.*)\((.*)\):', l, re.I)
+                is_func = re.match(r'(\s*)(?:cp)?def ([^\(]*)\((.*)\):', l, re.I)
                 if is_func:
                     # If the previous line was a function, print pass
                     if in_func:

xnmt/batcher.py

@@ -21,9 +21,11 @@ class Batcher:
   PAIR_SRC = 0
   PAIR_TRG = 1
 
-  def __init__(self, batch_size, pad_token=Vocab.ES):
+  def __init__(self, batch_size, pad_token=None):
     self.batch_size = batch_size
-    self.pad_token = pad_token
+    # The only reason why we don't set Vocab.ES as the default is because it currently
+    # breaks our documentation pipeline
+    self.pad_token = pad_token if pad_token != None else Vocab.ES
 
   @staticmethod
   def is_batch_sent(sent):

xnmt/embedder.py

@@ -11,18 +11,14 @@ class Embedder:
   def embed(self, word):
     """
     Embed a single word.
-    :param word:
-      This will generally be an integer word ID, but could also be something
-      like a string.
+    :param word: This will generally be an integer word ID, but could also be something like a string.
     """
     raise NotImplementedError('embed must be implemented in Embedder subclasses')
 
   def embed_sent(self, sent):
     """
     Embed a full sent worth of words.
-    :param sent:
-      This will generally be a list of word IDs, but could also be a list
-      of strings or some other format.
+    :param sent: This will generally be a list of word IDs, but could also be a list of strings or some other format.
     """
     raise NotImplementedError('embed_sent must be implemented in Embedder subclasses')
 

xnmt/encoder.py

@@ -13,11 +13,8 @@ class Encoder:
   def transduce(self, sent):
     """
     Encode inputs into outputs.
-    :param sent: The input to be encoded. This is duck-typed, so it is the
-      appropriate input for this particular type of encoder. Frequently it will
-      be a list of word embeddings, but it can be anything else.
-    :returns: The encoded output. Frequently this will be a list of expressions
-      representing the encoded vectors for each word.
+    :param sent: The input to be encoded. This is duck-typed, so it is the appropriate input for this particular type of encoder. Frequently it will be a list of word embeddings, but it can be anything else.
+    :returns: The encoded output. Frequently this will be a list of expressions representing the encoded vectors for each word.
     """
     raise NotImplementedError('transduce must be implemented in Encoder subclasses')
 

xnmt/search_strategy.py

@@ -11,10 +11,12 @@ def generate_output(self):
 
 class BeamSearch(SearchStrategy):
 
-  def __init__(self, b, max_len=100, len_norm=NoNormalization()):
+  def __init__(self, b, max_len=100, len_norm=None):
     self.b = b
     self.max_len = max_len
-    self.len_norm = len_norm
+    # The only reason why we don't set NoNormalization as the default is because it currently
+    # breaks our documentation pipeline
+    self.len_norm = len_norm if len_norm != None else NoNormalization()
 
   class Hypothesis:
     def __init__(self, score, id, state):