relax transformers packag requirement, remove BERTForSequenceClassifcation

docs/apidoc/textattack.models.helpers.rst

@@ -7,12 +7,6 @@ textattack.models.helpers package
    :show-inheritance:
 
 
-.. automodule:: textattack.models.helpers.bert_for_classification
-   :members:
-   :undoc-members:
-   :show-inheritance:
-
-
 .. automodule:: textattack.models.helpers.glove_embedding_layer
    :members:
    :undoc-members:

requirements.txt

@@ -11,9 +11,8 @@ numpy<1.19.0 #TF 2.0 requires this
 pandas>=1.0.1
 scipy==1.4.1
 torch
-transformers==3.3.0
+transformers>=3.3.0
 terminaltables
-tokenizers==0.8.1-rc2
 tqdm>=4.27,<4.50.0
 word2number
 num2words

textattack/constraints/grammaticality/cola.py

@@ -5,9 +5,10 @@ CoLA for Grammaticality
 """
 import lru
 import nltk
-from transformers import AutoModelForSequenceClassification, AutoTokenizer
+from transformers import AutoModelForSequenceClassification
 
 from textattack.constraints import Constraint
+from textattack.models.tokenizers import AutoTokenizer
 from textattack.models.wrappers import HuggingFaceModelWrapper
 
 
@@ -45,7 +46,7 @@ class COLA(Constraint):
         self.model_name = model_name
         self._reference_score_cache = lru.LRU(2 ** 10)
         model = AutoModelForSequenceClassification.from_pretrained(model_name)
-        tokenizer = AutoTokenizer.from_pretrained(model_name)
+        tokenizer = AutoTokenizer(model_name)
         self.model = HuggingFaceModelWrapper(model, tokenizer)
 
     def clear_cache(self):

textattack/models/helpers/bert_for_classification.py

@@ -1,39 +0,0 @@
-"""
-BERT Classification
-^^^^^^^^^^^^^^^^^^^^^
-
-"""
-
-
-import torch
-from transformers.modeling_bert import BertForSequenceClassification
-
-from textattack.models.tokenizers import AutoTokenizer
-from textattack.shared import utils
-
-
-class BERTForClassification:
-    """BERT fine-tuned for textual classification.
-
-    Args:
-        model_path(:obj:`string`): Path to the pre-trained model.
-        num_labels(:obj:`int`, optional):  Number of class labels for
-            prediction, if different than 2.
-    """
-
-    def __init__(self, model_path, num_labels=2):
-        model_file_path = utils.download_if_needed(model_path)
-        self.model = BertForSequenceClassification.from_pretrained(
-            model_file_path, num_labels=num_labels
-        )
-
-        self.model.to(utils.device)
-        self.model.eval()
-        self.tokenizer = AutoTokenizer(model_file_path)
-
-    def __call__(self, input_ids=None, **kwargs):
-        # The tokenizer will return ``input_ids`` along with ``token_type_ids``
-        # and an ``attention_mask``. Our pre-trained models only need the input
-        # IDs.
-        pred = self.model(input_ids=input_ids)[0]
-        return torch.nn.functional.softmax(pred, dim=-1)