Fix TF Flaubert and XLM (#9661)

* Fix Flaubert and XLM

* Fix Flaubert and XLM

* Apply style

src/transformers/models/flaubert/modeling_tf_flaubert.py

@@ -214,10 +214,13 @@ class TFFlaubertPreTrainedModel(TFPreTrainedModel):
         inputs_list = tf.constant([[7, 6, 0, 0, 1], [1, 2, 3, 0, 0], [0, 0, 0, 4, 5]])
         attns_list = tf.constant([[1, 1, 0, 0, 1], [1, 1, 1, 0, 0], [1, 0, 0, 1, 1]])
         if self.config.use_lang_emb and self.config.n_langs > 1:
-            langs_list = tf.constant([[1, 1, 0, 0, 1], [1, 1, 1, 0, 0], [1, 0, 0, 1, 1]])
+            return {
+                "input_ids": inputs_list,
+                "attention_mask": attns_list,
+                "langs": tf.constant([[1, 1, 0, 0, 1], [1, 1, 1, 0, 0], [1, 0, 0, 1, 1]]),
+            }
         else:
-            langs_list = None
-        return {"input_ids": inputs_list, "attention_mask": attns_list, "langs": langs_list}
+            return {"input_ids": inputs_list, "attention_mask": attns_list}
 
 
 @add_start_docstrings(

src/transformers/models/xlm/modeling_tf_xlm.py

@@ -536,10 +536,13 @@ class TFXLMPreTrainedModel(TFPreTrainedModel):
         inputs_list = tf.constant([[7, 6, 0, 0, 1], [1, 2, 3, 0, 0], [0, 0, 0, 4, 5]])
         attns_list = tf.constant([[1, 1, 0, 0, 1], [1, 1, 1, 0, 0], [1, 0, 0, 1, 1]])
         if self.config.use_lang_emb and self.config.n_langs > 1:
-            langs_list = tf.constant([[1, 1, 0, 0, 1], [1, 1, 1, 0, 0], [1, 0, 0, 1, 1]])
+            return {
+                "input_ids": inputs_list,
+                "attention_mask": attns_list,
+                "langs": tf.constant([[1, 1, 0, 0, 1], [1, 1, 1, 0, 0], [1, 0, 0, 1, 1]]),
+            }
         else:
-            langs_list = None
-        return {"input_ids": inputs_list, "attention_mask": attns_list, "langs": langs_list}
+            return {"input_ids": inputs_list, "attention_mask": attns_list}
 
 
 # Remove when XLMWithLMHead computes loss like other LM models
@@ -1045,10 +1048,16 @@ class TFXLMForMultipleChoice(TFXLMPreTrainedModel, TFMultipleChoiceLoss):
         Returns:
             tf.Tensor with dummy inputs
         """
+        # Sometimes XLM has language embeddings so don't forget to build them as well if needed
+        if self.config.use_lang_emb and self.config.n_langs > 1:
             return {
                 "input_ids": tf.constant(MULTIPLE_CHOICE_DUMMY_INPUTS),
                 "langs": tf.constant(MULTIPLE_CHOICE_DUMMY_INPUTS),
             }
+        else:
+            return {
+                "input_ids": tf.constant(MULTIPLE_CHOICE_DUMMY_INPUTS),
+            }
 
     @add_start_docstrings_to_model_forward(XLM_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(