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/gpt2/modeling_tf_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"
 
@@ -587,7 +588,7 @@ class TFGPT2Model(TFGPT2PreTrainedModel):
     @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=TFBaseModelOutputWithPast,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -679,7 +680,7 @@ class TFGPT2LMHeadModel(TFGPT2PreTrainedModel, TFCausalLanguageModelingLoss):
     @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=TFCausalLMOutputWithPast,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/ibert/modeling_ibert.py

@@ -43,6 +43,7 @@ from .quant_modules import IntGELU, IntLayerNorm, IntSoftmax, QuantAct, QuantEmb
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "ibert-roberta-base"
 _CONFIG_FOR_DOC = "IBertConfig"
 _TOKENIZER_FOR_DOC = "RobertaTokenizer"
 
@@ -776,7 +777,7 @@ class IBertModel(IBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(IBERT_INPUTS_DOCSTRING.format("(batch_size, sequence_length)"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ibert-roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithPoolingAndCrossAttentions,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -880,7 +881,7 @@ class IBertForMaskedLM(IBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(IBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ibert-roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
         mask="<mask>",
@@ -986,7 +987,7 @@ class IBertForSequenceClassification(IBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(IBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ibert-roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1069,7 +1070,7 @@ class IBertForMultipleChoice(IBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(IBERT_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ibert-roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1163,7 +1164,7 @@ class IBertForTokenClassification(IBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(IBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ibert-roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1272,7 +1273,7 @@ class IBertForQuestionAnswering(IBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(IBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="ibert-roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=QuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/led/modeling_led.py

@@ -49,6 +49,7 @@ from .configuration_led import LEDConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "allenai/led-base-16384"
 _CONFIG_FOR_DOC = "LEDConfig"
 _TOKENIZER_FOR_DOC = "LEDTokenizer"
 
@@ -2168,7 +2169,7 @@ class LEDModel(LEDPreTrainedModel):
     @add_start_docstrings_to_model_forward(LED_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/led-base-16384",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2451,7 +2452,7 @@ class LEDForSequenceClassification(LEDPreTrainedModel):
     @add_start_docstrings_to_model_forward(LED_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/led-base-16384",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2559,7 +2560,7 @@ class LEDForQuestionAnswering(LEDPreTrainedModel):
     @add_start_docstrings_to_model_forward(LED_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/led-base-16384",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/led/modeling_tf_led.py

@@ -47,6 +47,7 @@ from .configuration_led import LEDConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "allenai/led-base-16384"
 _CONFIG_FOR_DOC = "LEDConfig"
 _TOKENIZER_FOR_DOC = "LEDTokenizer"
 
@@ -2228,7 +2229,7 @@ class TFLEDModel(TFLEDPreTrainedModel):
     @add_start_docstrings_to_model_forward(LED_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/led-base-16384",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFLEDSeq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/longformer/modeling_longformer.py

@@ -44,6 +44,7 @@ from .configuration_longformer import LongformerConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "allenai/longformer-base-4096"
 _CONFIG_FOR_DOC = "LongformerConfig"
 _TOKENIZER_FOR_DOC = "LongformerTokenizer"
 
@@ -1811,7 +1812,7 @@ class LongformerForSequenceClassification(LongformerPreTrainedModel):
     @add_start_docstrings_to_model_forward(LONGFORMER_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/longformer-base-4096",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=LongformerSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2059,7 +2060,7 @@ class LongformerForTokenClassification(LongformerPreTrainedModel):
     @add_start_docstrings_to_model_forward(LONGFORMER_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/longformer-base-4096",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=LongformerTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2150,7 +2151,7 @@ class LongformerForMultipleChoice(LongformerPreTrainedModel):
     )
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/longformer-base-4096",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=LongformerMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/longformer/modeling_tf_longformer.py

@@ -46,6 +46,7 @@ from .configuration_longformer import LongformerConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "allenai/longformer-base-4096"
 _CONFIG_FOR_DOC = "LongformerConfig"
 _TOKENIZER_FOR_DOC = "LongformerTokenizer"
 
@@ -2081,7 +2082,7 @@ class TFLongformerForMaskedLM(TFLongformerPreTrainedModel, TFMaskedLanguageModel
     @add_start_docstrings_to_model_forward(LONGFORMER_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/longformer-base-4096",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFLongformerMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2358,7 +2359,7 @@ class TFLongformerForSequenceClassification(TFLongformerPreTrainedModel, TFSeque
     @add_start_docstrings_to_model_forward(LONGFORMER_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/longformer-base-4096",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFLongformerSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2484,7 +2485,7 @@ class TFLongformerForMultipleChoice(TFLongformerPreTrainedModel, TFMultipleChoic
     )
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/longformer-base-4096",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFLongformerMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2637,7 +2638,7 @@ class TFLongformerForTokenClassification(TFLongformerPreTrainedModel, TFTokenCla
     @add_start_docstrings_to_model_forward(LONGFORMER_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="allenai/longformer-base-4096",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFLongformerTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/lxmert/modeling_lxmert.py

@@ -40,6 +40,7 @@ from .configuration_lxmert import LxmertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "unc-nlp/lxmert-base-uncased"
 _CONFIG_FOR_DOC = "LxmertConfig"
 _TOKENIZER_FOR_DOC = "LxmertTokenizer"
 
@@ -901,7 +902,7 @@ class LxmertModel(LxmertPreTrainedModel):
     @add_start_docstrings_to_model_forward(LXMERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="unc-nlp/lxmert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=LxmertModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1382,7 +1383,7 @@ class LxmertForQuestionAnswering(LxmertPreTrainedModel):
     @add_start_docstrings_to_model_forward(LXMERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="unc-nlp/lxmert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=LxmertForQuestionAnsweringOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/lxmert/modeling_tf_lxmert.py

@@ -37,7 +37,7 @@ from .configuration_lxmert import LxmertConfig
 
 logger = logging.get_logger(__name__)
 
-
+_CHECKPOINT_FOR_DOC = "unc-nlp/lxmert-base-uncased"
 _CONFIG_FOR_DOC = "LxmertConfig"
 _TOKENIZER_FOR_DOC = "LxmertTokenizer"
 
@@ -951,7 +951,7 @@ class TFLxmertModel(TFLxmertPreTrainedModel):
     @add_start_docstrings_to_model_forward(LXMERT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="unc-nlp/lxmert-base-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFLxmertModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/marian/modeling_tf_marian.py

@@ -53,6 +53,7 @@ from .configuration_marian import MarianConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "Helsinki-NLP/opus-mt-en-de"
 _CONFIG_FOR_DOC = "MarianConfig"
 _TOKENIZER_FOR_DOC = "MarianTokenizer"
 
@@ -1203,7 +1204,7 @@ class TFMarianModel(TFMarianPreTrainedModel):
     @add_start_docstrings_to_model_forward(MARIAN_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="Helsinki-NLP/opus-mt-en-de",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSeq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/mbart/modeling_mbart.py

@@ -48,6 +48,7 @@ from .configuration_mbart import MBartConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "facebook/mbart-large-cc25"
 _CONFIG_FOR_DOC = "MBartConfig"
 _TOKENIZER_FOR_DOC = "MBartTokenizer"
 
@@ -1121,7 +1122,7 @@ class MBartModel(MBartPreTrainedModel):
     @add_start_docstrings_to_model_forward(MBART_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/mbart-large-cc25",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1377,7 +1378,7 @@ class MBartForSequenceClassification(MBartPreTrainedModel):
     @add_start_docstrings_to_model_forward(MBART_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/mbart-large-cc25",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1483,7 +1484,7 @@ class MBartForQuestionAnswering(MBartPreTrainedModel):
     @add_start_docstrings_to_model_forward(MBART_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/mbart-large-cc25",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=Seq2SeqQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/mbart/modeling_tf_mbart.py

@@ -52,6 +52,7 @@ from .configuration_mbart import MBartConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "facebook/mbart-large-cc25"
 _CONFIG_FOR_DOC = "MBartConfig"
 _TOKENIZER_FOR_DOC = "MBartTokenizer"
 
@@ -1189,7 +1190,7 @@ class TFMBartModel(TFMBartPreTrainedModel):
     @add_start_docstrings_to_model_forward(MBART_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="facebook/mbart-large-cc25",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSeq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/mobilebert/modeling_tf_mobilebert.py

@@ -59,6 +59,7 @@ from .configuration_mobilebert import MobileBertConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "google/mobilebert-uncased"
 _CONFIG_FOR_DOC = "MobileBertConfig"
 _TOKENIZER_FOR_DOC = "MobileBertTokenizer"
 
@@ -935,7 +936,7 @@ class TFMobileBertModel(TFMobileBertPreTrainedModel):
     @add_start_docstrings_to_model_forward(MOBILEBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/mobilebert-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutputWithPooling,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1127,7 +1128,7 @@ class TFMobileBertForMaskedLM(TFMobileBertPreTrainedModel, TFMaskedLanguageModel
     @add_start_docstrings_to_model_forward(MOBILEBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/mobilebert-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1348,7 +1349,7 @@ class TFMobileBertForSequenceClassification(TFMobileBertPreTrainedModel, TFSeque
     @add_start_docstrings_to_model_forward(MOBILEBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/mobilebert-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1456,7 +1457,7 @@ class TFMobileBertForQuestionAnswering(TFMobileBertPreTrainedModel, TFQuestionAn
     @add_start_docstrings_to_model_forward(MOBILEBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/mobilebert-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1591,7 +1592,7 @@ class TFMobileBertForMultipleChoice(TFMobileBertPreTrainedModel, TFMultipleChoic
     )
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/mobilebert-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1739,7 +1740,7 @@ class TFMobileBertForTokenClassification(TFMobileBertPreTrainedModel, TFTokenCla
     @add_start_docstrings_to_model_forward(MOBILEBERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/mobilebert-uncased",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/mpnet/modeling_mpnet.py

@@ -40,6 +40,7 @@ from .configuration_mpnet import MPNetConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "microsoft/mpnet-base"
 _CONFIG_FOR_DOC = "MPNetConfig"
 _TOKENIZER_FOR_DOC = "MPNetTokenizer"
 
@@ -511,7 +512,7 @@ class MPNetModel(MPNetPreTrainedModel):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("(batch_size, sequence_length)"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithPooling,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -593,7 +594,7 @@ class MPNetForMaskedLM(MPNetPreTrainedModel):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -695,7 +696,7 @@ class MPNetForSequenceClassification(MPNetPreTrainedModel):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -777,7 +778,7 @@ class MPNetForMultipleChoice(MPNetPreTrainedModel):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -869,7 +870,7 @@ class MPNetForTokenClassification(MPNetPreTrainedModel):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("(batch_size, sequence_length)"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -977,7 +978,7 @@ class MPNetForQuestionAnswering(MPNetPreTrainedModel):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=QuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/mpnet/modeling_tf_mpnet.py

@@ -55,6 +55,7 @@ from .configuration_mpnet import MPNetConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "microsoft/mpnet-base"
 _CONFIG_FOR_DOC = "MPNetConfig"
 _TOKENIZER_FOR_DOC = "MPNetTokenizer"
 
@@ -684,7 +685,7 @@ class TFMPNetModel(TFMPNetPreTrainedModel):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -814,7 +815,7 @@ class TFMPNetForMaskedLM(TFMPNetPreTrainedModel, TFMaskedLanguageModelingLoss):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -935,7 +936,7 @@ class TFMPNetForSequenceClassification(TFMPNetPreTrainedModel, TFSequenceClassif
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1041,7 +1042,7 @@ class TFMPNetForMultipleChoice(TFMPNetPreTrainedModel, TFMultipleChoiceLoss):
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1173,7 +1174,7 @@ class TFMPNetForTokenClassification(TFMPNetPreTrainedModel, TFTokenClassificatio
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1272,7 +1273,7 @@ class TFMPNetForQuestionAnswering(TFMPNetPreTrainedModel, TFQuestionAnsweringLos
     @add_start_docstrings_to_model_forward(MPNET_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="microsoft/mpnet-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/openai/modeling_openai.py

@@ -48,6 +48,7 @@ from .configuration_openai import OpenAIGPTConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "openai-gpt"
 _CONFIG_FOR_DOC = "OpenAIGPTConfig"
 _TOKENIZER_FOR_DOC = "OpenAIGPTTokenizer"
 
@@ -433,7 +434,7 @@ class OpenAIGPTModel(OpenAIGPTPreTrainedModel):
     @add_start_docstrings_to_model_forward(OPENAI_GPT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="openai-gpt",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -552,7 +553,7 @@ class OpenAIGPTLMHeadModel(OpenAIGPTPreTrainedModel):
     @add_start_docstrings_to_model_forward(OPENAI_GPT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="openai-gpt",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=CausalLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -756,7 +757,7 @@ class OpenAIGPTForSequenceClassification(OpenAIGPTPreTrainedModel):
     @add_start_docstrings_to_model_forward(OPENAI_GPT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="openai-gpt",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/openai/modeling_tf_openai.py

@@ -47,6 +47,7 @@ from .configuration_openai import OpenAIGPTConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "openai-gpt"
 _CONFIG_FOR_DOC = "OpenAIGPTConfig"
 _TOKENIZER_FOR_DOC = "OpenAIGPTTokenizer"
 
@@ -522,7 +523,7 @@ class TFOpenAIGPTModel(TFOpenAIGPTPreTrainedModel):
     @add_start_docstrings_to_model_forward(OPENAI_GPT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="openai-gpt",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -598,7 +599,7 @@ class TFOpenAIGPTLMHeadModel(TFOpenAIGPTPreTrainedModel, TFCausalLanguageModelin
     @add_start_docstrings_to_model_forward(OPENAI_GPT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="openai-gpt",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFCausalLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -857,7 +858,7 @@ class TFOpenAIGPTForSequenceClassification(TFOpenAIGPTPreTrainedModel, TFSequenc
     @add_start_docstrings_to_model_forward(OPENAI_GPT_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="openai-gpt",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/pegasus/modeling_tf_pegasus.py

@@ -53,6 +53,7 @@ from .configuration_pegasus import PegasusConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "google/pegasus-large"
 _CONFIG_FOR_DOC = "PegasusConfig"
 _TOKENIZER_FOR_DOC = "PegasusTokenizer"
 
@@ -1216,7 +1217,7 @@ class TFPegasusModel(TFPegasusPreTrainedModel):
     @add_start_docstrings_to_model_forward(PEGASUS_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/pegasus-large",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSeq2SeqModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/reformer/modeling_reformer.py

@@ -45,6 +45,7 @@ from .configuration_reformer import ReformerConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "google/reformer-crime-and-punishment"
 _CONFIG_FOR_DOC = "ReformerConfig"
 _TOKENIZER_FOR_DOC = "ReformerTokenizer"
 
@@ -1997,7 +1998,7 @@ class ReformerModel(ReformerPreTrainedModel):
     @add_start_docstrings_to_model_forward(REFORMER_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/reformer-crime-and-punishment",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=ReformerModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2204,7 +2205,7 @@ class ReformerModelWithLMHead(ReformerPreTrainedModel):
     @add_start_docstrings_to_model_forward(REFORMER_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/reformer-crime-and-punishment",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=CausalLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2319,7 +2320,7 @@ class ReformerForMaskedLM(ReformerPreTrainedModel):
     @add_start_docstrings_to_model_forward(REFORMER_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/reformer-crime-and-punishment",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2399,7 +2400,7 @@ class ReformerForSequenceClassification(ReformerPreTrainedModel):
     @add_start_docstrings_to_model_forward(REFORMER_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/reformer-crime-and-punishment",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -2501,7 +2502,7 @@ class ReformerForQuestionAnswering(ReformerPreTrainedModel):
     @add_start_docstrings_to_model_forward(REFORMER_INPUTS_DOCSTRING)
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="google/reformer-crime-and-punishment",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=QuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/roberta/modeling_roberta.py

@@ -51,6 +51,7 @@ from .configuration_roberta import RobertaConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "roberta-base"
 _CONFIG_FOR_DOC = "RobertaConfig"
 _TOKENIZER_FOR_DOC = "RobertaTokenizer"
 
@@ -709,7 +710,7 @@ class RobertaModel(RobertaPreTrainedModel):
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("(batch_size, sequence_length)"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=BaseModelOutputWithPoolingAndCrossAttentions,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1012,7 +1013,7 @@ class RobertaForMaskedLM(RobertaPreTrainedModel):
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
         mask="<mask>",
@@ -1122,7 +1123,7 @@ class RobertaForSequenceClassification(RobertaPreTrainedModel):
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=SequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1205,7 +1206,7 @@ class RobertaForMultipleChoice(RobertaPreTrainedModel):
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=MultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1299,7 +1300,7 @@ class RobertaForTokenClassification(RobertaPreTrainedModel):
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1408,7 +1409,7 @@ class RobertaForQuestionAnswering(RobertaPreTrainedModel):
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=QuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )

src/transformers/models/roberta/modeling_tf_roberta.py

@@ -57,6 +57,7 @@ from .configuration_roberta import RobertaConfig
 
 logger = logging.get_logger(__name__)
 
+_CHECKPOINT_FOR_DOC = "roberta-base"
 _CONFIG_FOR_DOC = "RobertaConfig"
 _TOKENIZER_FOR_DOC = "RobertaTokenizer"
 
@@ -707,7 +708,7 @@ class TFRobertaModel(TFRobertaPreTrainedModel):
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFBaseModelOutputWithPooling,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -841,7 +842,7 @@ class TFRobertaForMaskedLM(TFRobertaPreTrainedModel, TFMaskedLanguageModelingLos
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMaskedLMOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -965,7 +966,7 @@ class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel, TFSequenceCla
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFSequenceClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1076,7 +1077,7 @@ class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss)
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFMultipleChoiceModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1211,7 +1212,7 @@ class TFRobertaForTokenClassification(TFRobertaPreTrainedModel, TFTokenClassific
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFTokenClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
     )
@@ -1312,7 +1313,7 @@ class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnswerin
     @add_start_docstrings_to_model_forward(ROBERTA_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
         tokenizer_class=_TOKENIZER_FOR_DOC,
-        checkpoint="roberta-base",
+        checkpoint=_CHECKPOINT_FOR_DOC,
         output_type=TFQuestionAnsweringModelOutput,
         config_class=_CONFIG_FOR_DOC,
     )