Bugfix

cube/api.py

@@ -126,6 +126,10 @@ def __call__(self, text: Union[str, Document], flavour: Optional[str] = None):
         self._lm_helper.apply(doc)
         self._parser.process(doc, self._parser_collate, num_workers=0)
         self._lemmatizer.process(doc, self._lemmatizer_collate, num_workers=0)
+        for seq in doc.sentences:
+            for w in seq.words:
+                if w.upos =='PUNCT':
+                    w.lemma = w.word
         return doc
 
 

cube/networks/parser.py

@@ -137,9 +137,9 @@ def forward(self, X):
         for ii in range(len(x_word_emb_packed)):
             we = unpack(x_word_emb_packed[ii], sl, x_sents.shape[1], self._get_device())
             if word_emb_ext is None:
-                word_emb_ext = self._ext_proj[ii](we)
+                word_emb_ext = self._ext_proj[ii](we.float())
             else:
-                word_emb_ext = word_emb_ext + self._ext_proj[ii](we)
+                word_emb_ext = word_emb_ext + self._ext_proj[ii](we.float())
 
         word_emb_ext = word_emb_ext / len(x_word_emb_packed)
         word_emb_ext = torch.tanh(word_emb_ext)

setup.py

@@ -9,7 +9,7 @@ def parse_requirements(filename, session=None):
 
 setuptools.setup(
     name="nlpcube",
-    version="0.3.1.0",
+    version="0.3.1.1",
     author="Multiple authors",
     author_email="[email protected]",
     description="Natural Language Processing Toolkit with support for tokenization, sentence splitting, lemmatization, tagging and parsing for more than 60 languages",