Fix QA models binding for Flaubert, XLNet and XLM. (#3100)

Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>

Format & quality
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>

Again.
Signed-off-by: Morgan Funtowicz <morgan@huggingface.co>

src/transformers/modeling_auto.py

@@ -78,7 +78,7 @@ from .modeling_distilbert import (
 )
 from .modeling_flaubert import (
     FLAUBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
-    FlaubertForQuestionAnswering,
+    FlaubertForQuestionAnsweringSimple,
     FlaubertForSequenceClassification,
     FlaubertModel,
     FlaubertWithLMHeadModel,
@@ -97,7 +97,7 @@ from .modeling_t5 import T5_PRETRAINED_MODEL_ARCHIVE_MAP, T5Model, T5WithLMHeadM
 from .modeling_transfo_xl import TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_MAP, TransfoXLLMHeadModel, TransfoXLModel
 from .modeling_xlm import (
     XLM_PRETRAINED_MODEL_ARCHIVE_MAP,
-    XLMForQuestionAnswering,
+    XLMForQuestionAnsweringSimple,
     XLMForSequenceClassification,
     XLMModel,
     XLMWithLMHeadModel,
@@ -111,7 +111,7 @@ from .modeling_xlm_roberta import (
 )
 from .modeling_xlnet import (
     XLNET_PRETRAINED_MODEL_ARCHIVE_MAP,
-    XLNetForQuestionAnswering,
+    XLNetForQuestionAnsweringSimple,
     XLNetForSequenceClassification,
     XLNetForTokenClassification,
     XLNetLMHeadModel,
@@ -225,9 +225,9 @@ MODEL_FOR_QUESTION_ANSWERING_MAPPING = OrderedDict(
         (AlbertConfig, AlbertForQuestionAnswering),
         (RobertaConfig, RobertaForQuestionAnswering),
         (BertConfig, BertForQuestionAnswering),
-        (XLNetConfig, XLNetForQuestionAnswering),
-        (FlaubertConfig, FlaubertForQuestionAnswering),
-        (XLMConfig, XLMForQuestionAnswering),
+        (XLNetConfig, XLNetForQuestionAnsweringSimple),
+        (FlaubertConfig, FlaubertForQuestionAnsweringSimple),
+        (XLMConfig, XLMForQuestionAnsweringSimple),
     ]
 )