Merge pull request #2565 from flairNLP/tars-fixes

Tars fixes

flair/data.py

@@ -1532,11 +1532,11 @@ def iob_iobes(tags):
         if tag.value == "O" or tag.value == "":
             tag.value = "O"
             continue
-        t, label = tag.value.split("-")
+        t, label = tag.value.split("-", 1)
         if len(tags) == i + 1 or tags[i + 1].value == "O":
             next_same = False
         else:
-            nt, next_label = tags[i + 1].value.split("-")
+            nt, next_label = tags[i + 1].value.split("-", 1)
             next_same = nt == "I" and next_label == label
         if t == "B":
             if not next_same:

flair/models/sequence_tagger_model.py

@@ -230,6 +230,10 @@ def RNN(
 
     def forward_loss(self, sentences: Union[List[Sentence], Sentence]) -> Tuple[torch.Tensor, int]:
 
+        # if there are no sentences, there is no loss
+        if len(sentences) == 0:
+            return torch.tensor(0., dtype=torch.float, device=flair.device, requires_grad=True), 0
+
         # forward pass to get scores
         scores, gold_labels = self.forward(sentences)  # type: ignore
 
@@ -241,7 +245,6 @@ def forward(self, sentences: Union[List[Sentence], Sentence]):
         Forward propagation through network. Returns gold labels of batch in addition.
         :param sentences: Batch of current sentences
         """
-
         self.embeddings.embed(sentences)
 
         # make a zero-padded tensor for the whole sentence

flair/models/tars_model.py

@@ -57,7 +57,6 @@ def _get_tars_formatted_sentence(self, label, sentence):
     def _get_tars_formatted_sentences(self, sentences: List[Sentence]):
         label_text_pairs = []
         all_labels = [label.decode("utf-8") for label in self.get_current_label_dictionary().idx2item]
-        # print(all_labels)
         for sentence in sentences:
             label_text_pairs_for_sentence = []
             if self.training and self.num_negative_labels_to_sample is not None:
@@ -88,7 +87,6 @@ def _get_nearest_labels_for(self, labels):
             import random
 
             sample = random.sample(tags, k=self.num_negative_labels_to_sample)
-            # print(sample)
             return sample
 
         already_sampled_negative_labels = set()