Merge branch 'develop'

CHANGELOG.md

@@ -6,6 +6,18 @@ Most recent releases are shown at the top. Each release shows:
 - **Changed**: Additional parameters, changes to inputs or outputs, etc
 - **Fixed**: Bug fixes that don't change documented behaviour
 
+## 0.20.2 (2020-08-27)
+
+### New:
+- N/A
+
+### Changed
+- N/A
+
+### Fixed:
+- Always use `*Auto*` classes to load `transformers` models to prevent loading errors
+
+
 ## 0.20.1 (2020-08-25)
 
 ### New:

FAQ.md

@@ -687,7 +687,7 @@ The set of integer labels in your training set need to be complete and consecuti
 
 ### Why am I seeing an ERROR when installing *ktrain* on Google Colab?
 
-These errors (e.g., `has requirement gast>=0.3.2, but you'll have gast 0.2.2 which is incompatible`) are related to TensorFlow and can be usually safely ignored and shouldn't affect operation of *ktrain*.
+These errors (e.g., `tensorboard 2.1.1 requires setuptools>=41.0.0, but you'll have setuptools 39.0.1 which is incompatible.`) are related to TensorFlow and can be usually safely ignored and shouldn't affect operation of *ktrain*.  The errors should go away if you perform the indicated upgrades (e.g., `pip install -U setuptools`).
 
 [[Back to Top](#frequently-asked-questions-about-ktrain)]
 

ktrain/text/preprocessor.py

@@ -4,30 +4,10 @@
 from ..data import SequenceDataset
 from . import textutils as TU
 
-DistilBertTokenizer = transformers.DistilBertTokenizer
-DISTILBERT= 'distilbert'
-
-from transformers import BertConfig, TFBertForSequenceClassification, BertTokenizer, TFBertModel
-from transformers import XLNetConfig, TFXLNetForSequenceClassification, XLNetTokenizer, TFXLNetModel 
-from transformers import XLMConfig, TFXLMForSequenceClassification, XLMTokenizer, TFXLMModel
-from transformers import RobertaConfig, TFRobertaForSequenceClassification, RobertaTokenizer, TFRobertaModel
-from transformers import DistilBertConfig, TFDistilBertForSequenceClassification, DistilBertTokenizer, TFDistilBertModel
-from transformers import AlbertConfig, TFAlbertForSequenceClassification, AlbertTokenizer, TFAlbertModel
-from transformers import CamembertConfig, TFCamembertForSequenceClassification, CamembertTokenizer, TFCamembertModel
-from transformers import XLMRobertaConfig, TFXLMRobertaForSequenceClassification, XLMRobertaTokenizer, TFXLMRobertaModel
 from transformers import AutoConfig, TFAutoModelForSequenceClassification, AutoTokenizer, TFAutoModel
 
-TRANSFORMER_MODELS = {
-    'bert':       (BertConfig, TFBertForSequenceClassification, BertTokenizer, TFBertModel),
-    'xlnet':      (XLNetConfig, TFXLNetForSequenceClassification, XLNetTokenizer, TFXLNetModel),
-    'xlm':        (XLMConfig, TFXLMForSequenceClassification, XLMTokenizer, TFXLMModel),
-    'roberta':    (RobertaConfig, TFRobertaForSequenceClassification, RobertaTokenizer, TFRobertaModel),
-    'distilbert': (DistilBertConfig, TFDistilBertForSequenceClassification, DistilBertTokenizer, TFDistilBertModel),
-    'albert':     (AlbertConfig, TFAlbertForSequenceClassification, AlbertTokenizer, TFAlbertModel),
-    'camembert':  (CamembertConfig, TFCamembertForSequenceClassification, CamembertTokenizer, TFCamembertModel),
-    'xlm_roberta':  (XLMRobertaConfig, TFXLMRobertaForSequenceClassification, XLMRobertaTokenizer, TFXLMRobertaModel)
-}
 
+DISTILBERT= 'distilbert'
 
 NOSPACE_LANGS = ['zh-cn', 'zh-tw', 'ja']
 
@@ -807,15 +787,9 @@ def __init__(self,  model_name,
             self.model_name = 'jplu/tf-' + self.model_name
         else:
             self.name = model_name.split('-')[0]
-        if self.name not in TRANSFORMER_MODELS:
-            #raise ValueError('unsupported model name %s' % (model_name))
-            self.config = AutoConfig.from_pretrained(model_name)
-            self.model_type = TFAutoModelForSequenceClassification
-            self.tokenizer_type = AutoTokenizer
-        else:
-            self.config = None # use default config
-            self.model_type = TRANSFORMER_MODELS[self.name][1]
-            self.tokenizer_type = TRANSFORMER_MODELS[self.name][2]
+        self.config = AutoConfig.from_pretrained(model_name)
+        self.model_type = TFAutoModelForSequenceClassification
+        self.tokenizer_type = AutoTokenizer
 
         if "bert-base-japanese" in model_name:
             self.tokenizer_type = transformers.BertJapaneseTokenizer
@@ -1196,14 +1170,10 @@ def __init__(self, model_name, layers=U.DEFAULT_TRANSFORMER_LAYERS):
         else:
             self.name = model_name.split('-')[0]
 
-        if self.name not in TRANSFORMER_MODELS:
-            self.config = AutoConfig.from_pretrained(model_name)
-            self.model_type = TFAutoModel
-            self.tokenizer_type = AutoTokenizer
-        else:
-            self.config = None # use default config
-            self.model_type = TRANSFORMER_MODELS[self.name][3]
-            self.tokenizer_type = TRANSFORMER_MODELS[self.name][2]
+        self.config = AutoConfig.from_pretrained(model_name)
+        self.model_type = TFAutoModel
+        self.tokenizer_type = AutoTokenizer
+
         if "bert-base-japanese" in model_name:
             self.tokenizer_type = transformers.BertJapaneseTokenizer
 

ktrain/version.py

@@ -1,2 +1,2 @@
 __all__ = ['__version__']
-__version__ = '0.20.1'
+__version__ = '0.20.2'