Merge 5e0f761 into 1deb01a

saber/models/bert_for_ner.py

@@ -2,7 +2,7 @@
 from itertools import zip_longest
 
 import torch
-from pytorch_transformers import BertTokenizer
+from transformers import BertTokenizer
 from tqdm import tqdm
 
 from .. import constants

saber/models/bert_for_ner_and_re.py

@@ -2,8 +2,8 @@
 from itertools import zip_longest
 
 import torch
-from pytorch_transformers import BertConfig
-from pytorch_transformers import BertTokenizer
+from transformers import BertConfig
+from transformers import BertTokenizer
 from tqdm import tqdm
 
 from .. import constants

saber/models/modules/bert_for_entity_and_relation_extraction.py

@@ -2,9 +2,9 @@
 from itertools import permutations
 
 import torch
-from pytorch_transformers import BertModel
-# TODO (John): This can be shortned to from pytorch_transformers import x after next release
-from pytorch_transformers.modeling_bert import BertPreTrainedModel
+from transformers import BertModel
+# TODO (John): This can be shortned to from transformers import x after next release
+from transformers.modeling_bert import BertPreTrainedModel
 from torch import nn
 from torch.nn import CrossEntropyLoss
 
@@ -126,11 +126,10 @@ def __init__(self, config):
         # Biaffine transformation for relation classification
         self.rel_classifier = BiaffineAttention(head_tail_ffnns_size // 2, self.num_rel_labels)
 
-    def forward(self, input_ids, orig_to_tok_map, token_type_ids=None, attention_mask=None,
-                ent_labels=None, rel_labels=None, position_ids=None, head_mask=None):
+    def forward(self, input_ids, orig_to_tok_map, attention_mask=None, token_type_ids=None,
+                position_ids=None, head_mask=None, ent_labels=None, rel_labels=None):
         # Forward pass through BERT
-        outputs = self.bert(input_ids, position_ids=position_ids, token_type_ids=token_type_ids,
-                            attention_mask=attention_mask, head_mask=head_mask)
+        outputs = self.bert(input_ids, attention_mask, token_type_ids, position_ids, head_mask)
         sequence_output = outputs[0]
 
         # NER classification

saber/models/modules/bert_for_token_classification_multi_task.py

@@ -3,8 +3,8 @@
 from torch import nn
 from torch.nn import CrossEntropyLoss
 
-from pytorch_transformers.modeling_bert import BertPreTrainedModel
-from pytorch_transformers import BertModel
+from transformers.modeling_bert import BertPreTrainedModel
+from transformers import BertModel
 
 logger = logging.getLogger(__name__)
 
@@ -91,10 +91,9 @@ def __init__(self, config):
 
         self.init_weights()
 
-    def forward(self, input_ids, token_type_ids=None, attention_mask=None, labels=None,
-                position_ids=None, head_mask=None, model_idx=-1):
-        outputs = self.bert(input_ids, position_ids=position_ids, token_type_ids=token_type_ids,
-                            attention_mask=attention_mask, head_mask=head_mask)
+    def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None,
+                head_mask=None, labels=None, model_idx=-1):
+        outputs = self.bert(input_ids, attention_mask, token_type_ids, position_ids, head_mask)
         sequence_output = outputs[0]
 
         sequence_output = self.dropout(sequence_output)

saber/tests/conftest.py

@@ -2,7 +2,7 @@
 """
 import pytest
 import spacy
-from pytorch_transformers import BertTokenizer
+from transformers import BertTokenizer
 
 from ..config import Config
 from ..constants import SPACY_MODEL

saber/tests/test_bert_for_ner.py

@@ -2,11 +2,11 @@
 """
 import os
 
-from pytorch_transformers import CONFIG_NAME
-from pytorch_transformers import WEIGHTS_NAME
-from pytorch_transformers import BertForTokenClassification
-from pytorch_transformers import BertTokenizer
-from pytorch_transformers.optimization import AdamW
+from transformers import CONFIG_NAME
+from transformers import WEIGHTS_NAME
+from transformers import BertForTokenClassification
+from transformers import BertTokenizer
+from transformers.optimization import AdamW
 
 from ..constants import PARTITIONS
 from ..constants import WORDPIECE

saber/tests/test_bert_for_ner_and_re.py

@@ -3,11 +3,11 @@
 """
 import os
 
-from pytorch_transformers import CONFIG_NAME
-from pytorch_transformers import WEIGHTS_NAME
-from pytorch_transformers import BertForTokenClassification
-from pytorch_transformers import BertTokenizer
-from pytorch_transformers.optimization import AdamW
+from transformers import CONFIG_NAME
+from transformers import WEIGHTS_NAME
+from transformers import BertForTokenClassification
+from transformers import BertTokenizer
+from transformers.optimization import AdamW
 
 from ..constants import PARTITIONS
 from ..constants import WORDPIECE

saber/tests/test_bert_utils.py

@@ -1,7 +1,7 @@
 """Test suite for the `bert_utils` module (saber.utils.bert_utils).
 """
 import torch
-from pytorch_transformers.optimization import AdamW
+from transformers.optimization import AdamW
 from torch.utils.data import RandomSampler
 from torch.utils.data import SequentialSampler
 

saber/utils/bert_utils.py

@@ -1,6 +1,6 @@
 import torch
 from keras_preprocessing.sequence import pad_sequences
-from pytorch_transformers.optimization import AdamW
+from transformers.optimization import AdamW
 from torch.utils import data
 
 from saber import constants

setup.py

@@ -35,7 +35,7 @@
         'scikit-learn>=0.20.1',
         'scikit-multilearn>=0.2.0',
         'torch>=1.2.0',
-        'pytorch-transformers>=1.2.0',
+        'transformers>=2.0.0',
         'Flask>=1.0.2',
         'waitress>=1.1.0',
         'Keras-Preprocessing>=1.1.0',