Refactoring checkpoint names for multiple models (#10527)

* Refactor checkpoint name in ALBERT and ALBERT_tf

* Refactor checkpoint name in BART and BART_tf

* Refactor checkpoint name in BERT generation

* Refactor checkpoint name in Blenderbot_tf

* Refactor checkpoint name in Blenderbot_small_tf

* Refactor checkpoint name in ConvBERT AND CONVBERT_TF

* Refactor checkpoint name in CTRL AND CTRL_TF

* Refactor checkpoint name in DistilBERT AND DistilBERT_TF

* Refactor checkpoint name in DistilBERT redo

* Refactor checkpoint name in Electra and Electra_tf

* Refactor checkpoint name in FlauBERT and FlauBERT_tf

* Refactor checkpoint name in FSMT

* Refactor checkpoint name in GPT2 and GPT2_tf

* Refactor checkpoint name in IBERT

* Refactor checkpoint name in LED and LED_tf

* Refactor checkpoint name in Longformer and Longformer_tf

* Refactor checkpoint name in Lxmert and Lxmert_tf

* Refactor checkpoint name in Marian_tf

* Refactor checkpoint name in MBART and MBART_tf

* Refactor checkpoint name in MobileBERT and MobileBERT_tf

* Refactor checkpoint name in mpnet and mpnet_tf

* Refactor checkpoint name in openai and openai_tf

* Refactor checkpoint name in pegasus_tf

* Refactor checkpoint name in reformer

* Refactor checkpoint name in Roberta and Roberta_tf

* Refactor checkpoint name in SqueezeBert

* Refactor checkpoint name in Transformer_xl and Transformer_xl_tf

* Refactor checkpoint name in XLM and XLM_tf

* Refactor checkpoint name in XLNET and XLNET_tf

* Refactor checkpoint name in BERT_tf

* run make tests, style, quality, fixup

src/transformers/models/albert/modeling_albert.py

@@ -52,6 +52,7 @@ from .configuration_albert import AlbertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "albert-base-v2"
 _CONFIG_FOR_DOC = "AlbertConfig"
 _TOKENIZER_FOR_DOC = "AlbertTokenizer"
 
@@ -658,7 +659,7 @@ class AlbertModel(AlbertPreTrainedModel):
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithPooling,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -901,7 +902,7 @@ class AlbertForMaskedLM(AlbertPreTrainedModel):
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -979,7 +980,7 @@ class AlbertForSequenceClassification(AlbertPreTrainedModel):
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1067,7 +1068,7 @@ class AlbertForTokenClassification(AlbertPreTrainedModel):
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1155,7 +1156,7 @@ class AlbertForQuestionAnswering(AlbertPreTrainedModel):
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=QuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1254,7 +1255,7 @@ class AlbertForMultipleChoice(AlbertPreTrainedModel):
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/albert/modeling_tf_albert.py

@@ -59,6 +59,7 @@ from .configuration_albert import AlbertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "albert-base-v2"
 _CONFIG_FOR_DOC = "AlbertConfig"
 _TOKENIZER_FOR_DOC = "AlbertTokenizer"
 
@@ -780,7 +781,7 @@ class TFAlbertModel(TFAlbertPreTrainedModel):
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutputWithPooling,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -998,7 +999,7 @@ class TFAlbertForMaskedLM(TFAlbertPreTrainedModel, TFMaskedLanguageModelingLoss)
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1103,7 +1104,7 @@ class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel, TFSequenceClass
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1207,7 +1208,7 @@ class TFAlbertForTokenClassification(TFAlbertPreTrainedModel, TFTokenClassificat
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1308,7 +1309,7 @@ class TFAlbertForQuestionAnswering(TFAlbertPreTrainedModel, TFQuestionAnsweringL
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1436,7 +1437,7 @@ class TFAlbertForMultipleChoice(TFAlbertPreTrainedModel, TFMultipleChoiceLoss):
     @add_start_docstrings_to_model_forward(ALBERT_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="albert-base-v2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/bart/modeling_bart.py

@@ -49,6 +49,7 @@ from .configuration_bart import BartConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "facebook/bart-large"
 _CONFIG_FOR_DOC = "BartConfig"
 _TOKENIZER_FOR_DOC = "BartTokenizer"
 
@@ -1109,7 +1110,7 @@ class BartModel(BartPretrainedModel):
     @add_start_docstrings_to_model_forward(BART_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/bart-large",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1377,7 +1378,7 @@ class BartForSequenceClassification(BartPretrainedModel):
     @add_start_docstrings_to_model_forward(BART_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/bart-large",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1482,7 +1483,7 @@ class BartForQuestionAnswering(BartPretrainedModel):
     @add_start_docstrings_to_model_forward(BART_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/bart-large",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/bart/modeling_tf_bart.py

@@ -52,6 +52,7 @@ from .configuration_bart import BartConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "facebook/bart-large"
 _CONFIG_FOR_DOC = "BartConfig"
 _TOKENIZER_FOR_DOC = "BartTokenizer"
 
@@ -1170,7 +1171,7 @@ class TFBartModel(TFBartPretrainedModel):
     @add_start_docstrings_to_model_forward(BART_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/bart-large",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSeq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/bert/modeling_tf_bert.py

@@ -64,6 +64,7 @@ from .configuration_bert import BertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "bert-base-cased"
 _CONFIG_FOR_DOC = "BertConfig"
 _TOKENIZER_FOR_DOC = "BertTokenizer"
 
@@ -850,7 +851,7 @@ class TFBertModel(TFBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="bert-base-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutputWithPooling,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1067,7 +1068,7 @@ class TFBertForMaskedLM(TFBertPreTrainedModel, TFMaskedLanguageModelingLoss):
     @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="bert-base-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1171,7 +1172,7 @@ class TFBertLMHeadModel(TFBertPreTrainedModel, TFCausalLanguageModelingLoss):
 
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="bert-base-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFCausalLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1382,7 +1383,7 @@ class TFBertForSequenceClassification(TFBertPreTrainedModel, TFSequenceClassific
     @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="bert-base-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1492,7 +1493,7 @@ class TFBertForMultipleChoice(TFBertPreTrainedModel, TFMultipleChoiceLoss):
     @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="bert-base-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1648,7 +1649,7 @@ class TFBertForTokenClassification(TFBertPreTrainedModel, TFTokenClassificationL
     @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="bert-base-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1755,7 +1756,7 @@ class TFBertForQuestionAnswering(TFBertPreTrainedModel, TFQuestionAnsweringLoss)
     @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="bert-base-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/bert_generation/modeling_bert_generation.py

@@ -35,6 +35,7 @@ from .configuration_bert_generation import BertGenerationConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "google/bert_for_seq_generation_L-24_bbc_encoder"
 _CONFIG_FOR_DOC = "BertGenerationConfig"
 _TOKENIZER_FOR_DOC = "BertGenerationTokenizer"
 
@@ -300,7 +301,7 @@ class BertGenerationEncoder(BertGenerationPreTrainedModel):
     @add_start_docstrings_to_model_forward(BERT_GENERATION_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/bert_for_seq_generation_L-24_bbc_encoder",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithPastAndCrossAttentions,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/blenderbot/modeling_tf_blenderbot.py

@@ -54,6 +54,7 @@ from .configuration_blenderbot import BlenderbotConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "facebook/blenderbot-400M-distill"
 _CONFIG_FOR_DOC = "BlenderbotConfig"
 _TOKENIZER_FOR_DOC = "BlenderbotTokenizer"
 
@@ -1186,7 +1187,7 @@ class TFBlenderbotModel(TFBlenderbotPreTrainedModel):
     @add_start_docstrings_to_model_forward(BLENDERBOT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/blenderbot-400M-distill",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSeq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/blenderbot_small/modeling_tf_blenderbot_small.py

@@ -52,6 +52,7 @@ from .configuration_blenderbot_small import BlenderbotSmallConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "facebook/blenderbot_small-90M"
 _CONFIG_FOR_DOC = "BlenderbotSmallConfig"
 _TOKENIZER_FOR_DOC = "BlenderbotSmallTokenizer"
 
@@ -1174,7 +1175,7 @@ class TFBlenderbotSmallModel(TFBlenderbotSmallPreTrainedModel):
     @add_start_docstrings_to_model_forward(BLENDERBOT_SMALL_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/blenderbot_small-90M",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSeq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/convbert/modeling_convbert.py

@@ -47,6 +47,7 @@ from .configuration_convbert import ConvBertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "YituTech/conv-bert-base"
 _CONFIG_FOR_DOC = "ConvBertConfig"
 _TOKENIZER_FOR_DOC = "ConvBertTokenizer"
 
@@ -773,7 +774,7 @@ class ConvBertModel(ConvBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithCrossAttentions,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -870,7 +871,7 @@ class ConvBertForMaskedLM(ConvBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -969,7 +970,7 @@ class ConvBertForSequenceClassification(ConvBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1053,7 +1054,7 @@ class ConvBertForMultipleChoice(ConvBertPreTrainedModel):
     )
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1145,7 +1146,7 @@ class ConvBertForTokenClassification(ConvBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("(batch_size, sequence_length)"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1232,7 +1233,7 @@ class ConvBertForQuestionAnswering(ConvBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("(batch_size, sequence_length)"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=QuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/convbert/modeling_tf_convbert.py

@@ -51,6 +51,7 @@ from .configuration_convbert import ConvBertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "YituTech/conv-bert-base"
 _CONFIG_FOR_DOC = "ConvBertConfig"
 _TOKENIZER_FOR_DOC = "ConvBertTokenizer"
 
@@ -750,7 +751,7 @@ class TFConvBertModel(TFConvBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -882,7 +883,7 @@ class TFConvBertForMaskedLM(TFConvBertPreTrainedModel, TFMaskedLanguageModelingL
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1003,7 +1004,7 @@ class TFConvBertForSequenceClassification(TFConvBertPreTrainedModel, TFSequenceC
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1112,7 +1113,7 @@ class TFConvBertForMultipleChoice(TFConvBertPreTrainedModel, TFMultipleChoiceLos
     )
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1247,7 +1248,7 @@ class TFConvBertForTokenClassification(TFConvBertPreTrainedModel, TFTokenClassif
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1342,7 +1343,7 @@ class TFConvBertForQuestionAnswering(TFConvBertPreTrainedModel, TFQuestionAnswer
     @add_start_docstrings_to_model_forward(CONVBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="YituTech/conv-bert-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/ctrl/modeling_ctrl.py

@@ -31,6 +31,7 @@ from .configuration_ctrl import CTRLConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "ctrl"
 _CONFIG_FOR_DOC = "CTRLConfig"
 _TOKENIZER_FOR_DOC = "CTRLTokenizer"
 
@@ -355,7 +356,7 @@ class CTRLModel(CTRLPreTrainedModel):
     @add_start_docstrings_to_model_forward(CTRL_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ctrl",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithPast,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -515,7 +516,7 @@ class CTRLLMHeadModel(CTRLPreTrainedModel):
     @add_start_docstrings_to_model_forward(CTRL_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ctrl",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=CausalLMOutputWithPast,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -618,7 +619,7 @@ class CTRLForSequenceClassification(CTRLPreTrainedModel):
     @add_start_docstrings_to_model_forward(CTRL_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ctrl",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/ctrl/modeling_tf_ctrl.py

@@ -38,6 +38,7 @@ from .configuration_ctrl import CTRLConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "ctrl"
 _CONFIG_FOR_DOC = "CTRLConfig"
 _TOKENIZER_FOR_DOC = "CTRLTokenizer"
 
@@ -543,7 +544,7 @@ class TFCTRLModel(TFCTRLPreTrainedModel):
     @add_start_docstrings_to_model_forward(CTRL_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ctrl",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutputWithPast,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -671,7 +672,7 @@ class TFCTRLLMHeadModel(TFCTRLPreTrainedModel, TFCausalLanguageModelingLoss):
     @add_start_docstrings_to_model_forward(CTRL_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ctrl",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFCausalLMOutputWithPast,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -795,7 +796,7 @@ class TFCTRLForSequenceClassification(TFCTRLPreTrainedModel, TFSequenceClassific
     @add_start_docstrings_to_model_forward(CTRL_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ctrl",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/distilbert/modeling_distilbert.py

@@ -52,7 +52,7 @@ from .configuration_distilbert import DistilBertConfig
 
 
 logger = logging.get_logger(__name__)
-
+_CHECKPOINT_FOR_DOC = "distilbert-base-uncased"
 _CONFIG_FOR_DOC = "DistilBertConfig"
 _TOKENIZER_FOR_DOC = "DistilBertTokenizer"
 
@@ -441,11 +441,10 @@ class DistilBertModel(DistilBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, num_choices"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
-    @add_code_sample_docstrings(tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint="distilbert-base-uncased")
     def forward(
         self,
         input_ids=None,
@@ -517,7 +516,7 @@ class DistilBertForMaskedLM(DistilBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, num_choices"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -593,7 +592,7 @@ class DistilBertForSequenceClassification(DistilBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, num_choices"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -674,7 +673,7 @@ class DistilBertForQuestionAnswering(DistilBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, num_choices"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=QuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -770,7 +769,7 @@ class DistilBertForTokenClassification(DistilBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/distilbert/modeling_tf_distilbert.py

@@ -53,6 +53,7 @@ from .configuration_distilbert import DistilBertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "distilbert-base-uncased"
 _CONFIG_FOR_DOC = "DistilBertConfig"
 _TOKENIZER_FOR_DOC = "DistilBertTokenizer"
 
@@ -545,7 +546,7 @@ class TFDistilBertModel(TFDistilBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -660,7 +661,7 @@ class TFDistilBertForMaskedLM(TFDistilBertPreTrainedModel, TFMaskedLanguageModel
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -761,7 +762,7 @@ class TFDistilBertForSequenceClassification(TFDistilBertPreTrainedModel, TFSeque
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -856,7 +857,7 @@ class TFDistilBertForTokenClassification(TFDistilBertPreTrainedModel, TFTokenCla
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -964,7 +965,7 @@ class TFDistilBertForMultipleChoice(TFDistilBertPreTrainedModel, TFMultipleChoic
     )
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1090,7 +1091,7 @@ class TFDistilBertForQuestionAnswering(TFDistilBertPreTrainedModel, TFQuestionAn
     @add_start_docstrings_to_model_forward(DISTILBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="distilbert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/electra/modeling_electra.py

@@ -54,6 +54,7 @@ from .configuration_electra import ElectraConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "google/electra-small-discriminator"
 _CONFIG_FOR_DOC = "ElectraConfig"
 _TOKENIZER_FOR_DOC = "ElectraTokenizer"
 
@@ -812,7 +813,7 @@ class ElectraModel(ElectraPreTrainedModel):
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithCrossAttentions,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -910,7 +911,7 @@ class ElectraForSequenceClassification(ElectraPreTrainedModel):
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1092,7 +1093,7 @@ class ElectraForMaskedLM(ElectraPreTrainedModel):
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1171,7 +1172,7 @@ class ElectraForTokenClassification(ElectraPreTrainedModel):
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1258,7 +1259,7 @@ class ElectraForQuestionAnswering(ElectraPreTrainedModel):
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=QuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1359,7 +1360,7 @@ class ElectraForMultipleChoice(ElectraPreTrainedModel):
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/electra/modeling_tf_electra.py

@@ -57,6 +57,7 @@ from .configuration_electra import ElectraConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "google/electra-small-discriminator"
 _CONFIG_FOR_DOC = "ElectraConfig"
 _TOKENIZER_FOR_DOC = "ElectraTokenizer"
 
@@ -732,7 +733,7 @@ class TFElectraModel(TFElectraPreTrainedModel):
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -951,7 +952,7 @@ class TFElectraForMaskedLM(TFElectraPreTrainedModel, TFMaskedLanguageModelingLos
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-generator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1071,7 +1072,7 @@ class TFElectraForSequenceClassification(TFElectraPreTrainedModel, TFSequenceCla
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1179,7 +1180,7 @@ class TFElectraForMultipleChoice(TFElectraPreTrainedModel, TFMultipleChoiceLoss)
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1316,7 +1317,7 @@ class TFElectraForTokenClassification(TFElectraPreTrainedModel, TFTokenClassific
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1413,7 +1414,7 @@ class TFElectraForQuestionAnswering(TFElectraPreTrainedModel, TFQuestionAnswerin
     @add_start_docstrings_to_model_forward(ELECTRA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/electra-small-discriminator",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/flaubert/modeling_flaubert.py

@@ -38,6 +38,7 @@ from .configuration_flaubert import FlaubertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "flaubert/flaubert_base_cased"
 _CONFIG_FOR_DOC = "FlaubertConfig"
 _TOKENIZER_FOR_DOC = "FlaubertTokenizer"
 
@@ -143,7 +144,7 @@ class FlaubertModel(XLMModel):
     @add_start_docstrings_to_model_forward(FLAUBERT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="flaubert/flaubert_base_cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/flaubert/modeling_tf_flaubert.py

@@ -52,6 +52,7 @@ from .configuration_flaubert import FlaubertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "flaubert/flaubert_base_cased"
 _CONFIG_FOR_DOC = "FlaubertConfig"
 _TOKENIZER_FOR_DOC = "FlaubertTokenizer"
 
@@ -236,7 +237,7 @@ class TFFlaubertModel(TFFlaubertPreTrainedModel):
     @add_start_docstrings_to_model_forward(FLAUBERT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="jplu/tf-flaubert-small-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -822,7 +823,7 @@ class TFFlaubertWithLMHeadModel(TFFlaubertPreTrainedModel):
     @add_start_docstrings_to_model_forward(FLAUBERT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="jplu/tf-flaubert-small-cased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFFlaubertWithLMHeadModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/fsmt/modeling_fsmt.py

@@ -57,6 +57,7 @@ from .configuration_fsmt import FSMTConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "facebook/wmt19-ru-en"
 _CONFIG_FOR_DOC = "FSMTConfig"
 _TOKENIZER_FOR_DOC = "FSMTTokenizer"
 
@@ -989,7 +990,7 @@ class FSMTModel(PretrainedFSMTModel):
     @add_start_docstrings_to_model_forward(FSMT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/wmt19-ru-en",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/gpt2/modeling_gpt2.py

@@ -51,6 +51,7 @@ from .configuration_gpt2 import GPT2Config
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "gpt2"
 _CONFIG_FOR_DOC = "GPT2Config"
 _TOKENIZER_FOR_DOC = "GPT2Tokenizer"
 
@@ -599,7 +600,7 @@ class GPT2Model(GPT2PreTrainedModel):
     @add_start_docstrings_to_model_forward(GPT2_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="gpt2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithPastAndCrossAttentions,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -871,7 +872,7 @@ class GPT2LMHeadModel(GPT2PreTrainedModel):
     @add_start_docstrings_to_model_forward(GPT2_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="gpt2",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=CausalLMOutputWithCrossAttentions,
         config_class=_CONFIG_FOR_DOC,
     )