fixing errors

neuralmonkey/decoders/autoregressive.py

@@ -176,7 +176,7 @@ def embedding_size(self) -> int:
                      "size of the reused embeddings from the "
                      "`embeddings_source`.")
 
-        return self.embeddings_source.dimension
+        return self.embeddings_source.embedding_matrix.get_shape()[1].value
 
     @tensor
     def go_symbols(self) -> tf.Tensor:

neuralmonkey/decoders/sequence_labeler.py

@@ -23,7 +23,7 @@ def __init__(self,
                  name: str,
                  encoder: TemporalStateful,
                  data_id: str,
-                 vocabulary: Vocabulary,
+                 vocabulary: Vocabulary = None,
                  embeddings_source: EmbeddedSequence = None,
                  dropout_keep_prob: float = 1.0,
                  reuse: ModelPart = None,
@@ -34,19 +34,20 @@ def __init__(self,
         ModelPart.__init__(self, name, reuse, save_checkpoint, load_checkpoint,
                            initializers)
 
-        self.vocabulary = vocabulary
         self.embeddings_source = embeddings_source
         self.encoder = encoder
         self.data_id = data_id
         self.dropout_keep_prob = dropout_keep_prob
 
-        # We provide only embedding_source when we want to input and output
+        # We provide only embedding_source when we want to tie input and output
         # projections
-        if self.embeddings_source is not None and self.vocabulary is not None:
-            warn("Both `vocabulary` and `embedding_source` was provided. "
-                 "using `embedding_source.vocabulary` instead of provided "
-                 "`vocabulary`")
+        if (self.embeddings_source is None) == (vocabulary is None):
+            raise ValueError("You must specify either `vocabulary or` or "
+                             "`embeddings_source`, not both")
+        elif self.embeddings_source is not None:
             self.vocabulary = self.embeddings_source.vocabulary
+        elif vocabulary is not None:
+            self.vocabulary = vocabulary
     # pylint: enable=too-many-arguments
 
     @property

neuralmonkey/readers/string_vector_reader.py

@@ -13,7 +13,7 @@ def process_line(line: str, lineno: int, path: str) -> np.ndarray:
 
         return np.array(numbers, dtype=dtype)
 
-    def reader(files: List[str])-> Iterable[List[np.ndarray]]:
+    def reader(files: List[str]) -> Iterable[List[np.ndarray]]:
         for path in files:
             current_line = 0
 

tests/bert.ini

@@ -66,7 +66,6 @@ dropout_keep_prob=0.9
 class=decoders.sequence_labeler.SequenceLabeler
 name="labeler_bert"
 encoder=<encoder>
-vocabulary=<vocabulary>
 data_id="source_masked"
 dropout_keep_prob=0.5
 embeddings_source=<sequence>