Add MP Net 2 (#9004)

README.md

@@ -213,6 +213,7 @@ Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
 1. **[LXMERT](https://huggingface.co/transformers/model_doc/lxmert.html)** (from UNC Chapel Hill) released with the paper [LXMERT: Learning Cross-Modality Encoder Representations from Transformers for Open-Domain Question Answering](https://arxiv.org/abs/1908.07490) by Hao Tan and Mohit Bansal.
 1. **[MarianMT](https://huggingface.co/transformers/model_doc/marian.html)** Machine translation models trained using [OPUS](http://opus.nlpl.eu/) data by Jörg Tiedemann. The [Marian Framework](https://marian-nmt.github.io/) is being developed by the Microsoft Translator Team.
 1. **[MBart](https://huggingface.co/transformers/model_doc/mbart.html)** (from Facebook) released with the paper [Multilingual Denoising Pre-training for Neural Machine Translation](https://arxiv.org/abs/2001.08210) by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer.
+1. **[MPNet](https://huggingface.co/transformers/model_doc/mpnet.html)** (from Microsoft Research) released with the paper [MPNet: Masked and Permuted Pre-training for Language Understanding](https://arxiv.org/abs/2004.09297) by Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu.
 1. **[MT5](https://huggingface.co/transformers/model_doc/mt5.html)** (from Google AI) released with the paper [mT5: A massively multilingual pre-trained text-to-text transformer](https://arxiv.org/abs/2010.11934) by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel.
 1. **[Pegasus](https://huggingface.co/transformers/model_doc/pegasus.html)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777)> by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
 1. **[ProphetNet](https://huggingface.co/transformers/model_doc/prophetnet.html)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.

docs/source/index.rst

@@ -151,41 +151,44 @@ and conversion utilities for the following models:
 22. :doc:`MBart <model_doc/mbart>` (from Facebook) released with the paper `Multilingual Denoising Pre-training for
     Neural Machine Translation <https://arxiv.org/abs/2001.08210>`__ by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li,
     Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer.
-23. :doc:`MT5 <model_doc/mt5>` (from Google AI) released with the paper `mT5: A massively multilingual pre-trained
+23. :doc:`MPNet <model_doc/mpnet>` (from Microsoft Research) released with the paper `MPNet: Masked and Permuted
+    Pre-training for Language Understanding <https://arxiv.org/abs/2004.09297>`__ by Kaitao Song, Xu Tan, Tao Qin,
+    Jianfeng Lu, Tie-Yan Liu.
+24. :doc:`MT5 <model_doc/mt5>` (from Google AI) released with the paper `mT5: A massively multilingual pre-trained
     text-to-text transformer <https://arxiv.org/abs/2010.11934>`__ by Linting Xue, Noah Constant, Adam Roberts, Mihir
     Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel.
-24. :doc:`Pegasus <model_doc/pegasus>` (from Google) released with the paper `PEGASUS: Pre-training with Extracted
+25. :doc:`Pegasus <model_doc/pegasus>` (from Google) released with the paper `PEGASUS: Pre-training with Extracted
     Gap-sentences for Abstractive Summarization <https://arxiv.org/abs/1912.08777>`__> by Jingqing Zhang, Yao Zhao,
     Mohammad Saleh and Peter J. Liu.
-25. :doc:`ProphetNet <model_doc/prophetnet>` (from Microsoft Research) released with the paper `ProphetNet: Predicting
+26. :doc:`ProphetNet <model_doc/prophetnet>` (from Microsoft Research) released with the paper `ProphetNet: Predicting
     Future N-gram for Sequence-to-Sequence Pre-training <https://arxiv.org/abs/2001.04063>`__ by Yu Yan, Weizhen Qi,
     Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
-26. :doc:`Reformer <model_doc/reformer>` (from Google Research) released with the paper `Reformer: The Efficient
+27. :doc:`Reformer <model_doc/reformer>` (from Google Research) released with the paper `Reformer: The Efficient
     Transformer <https://arxiv.org/abs/2001.04451>`__ by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.
-27. :doc:`RoBERTa <model_doc/roberta>` (from Facebook), released together with the paper a `Robustly Optimized BERT
+28. :doc:`RoBERTa <model_doc/roberta>` (from Facebook), released together with the paper a `Robustly Optimized BERT
     Pretraining Approach <https://arxiv.org/abs/1907.11692>`__ by Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar
     Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov. ultilingual BERT into `DistilmBERT
     <https://github.com/huggingface/transformers/tree/master/examples/distillation>`__ and a German version of
     DistilBERT.
-28. :doc:`SqueezeBert <model_doc/squeezebert>` released with the paper `SqueezeBERT: What can computer vision teach NLP
+29. :doc:`SqueezeBert <model_doc/squeezebert>` released with the paper `SqueezeBERT: What can computer vision teach NLP
     about efficient neural networks? <https://arxiv.org/abs/2006.11316>`__ by Forrest N. Iandola, Albert E. Shaw, Ravi
     Krishna, and Kurt W. Keutzer.
-29. :doc:`T5 <model_doc/t5>` (from Google AI) released with the paper `Exploring the Limits of Transfer Learning with a
+30. :doc:`T5 <model_doc/t5>` (from Google AI) released with the paper `Exploring the Limits of Transfer Learning with a
     Unified Text-to-Text Transformer <https://arxiv.org/abs/1910.10683>`__ by Colin Raffel and Noam Shazeer and Adam
     Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
-30. :doc:`Transformer-XL <model_doc/transformerxl>` (from Google/CMU) released with the paper `Transformer-XL:
+31. :doc:`Transformer-XL <model_doc/transformerxl>` (from Google/CMU) released with the paper `Transformer-XL:
     Attentive Language Models Beyond a Fixed-Length Context <https://arxiv.org/abs/1901.02860>`__ by Zihang Dai*,
     Zhilin Yang*, Yiming Yang, Jaime Carbonell, Quoc V. Le, Ruslan Salakhutdinov.
-31. :doc:`XLM <model_doc/xlm>` (from Facebook) released together with the paper `Cross-lingual Language Model
+32. :doc:`XLM <model_doc/xlm>` (from Facebook) released together with the paper `Cross-lingual Language Model
     Pretraining <https://arxiv.org/abs/1901.07291>`__ by Guillaume Lample and Alexis Conneau.
-32. :doc:`XLM-ProphetNet <model_doc/xlmprophetnet>` (from Microsoft Research) released with the paper `ProphetNet:
+33. :doc:`XLM-ProphetNet <model_doc/xlmprophetnet>` (from Microsoft Research) released with the paper `ProphetNet:
     Predicting Future N-gram for Sequence-to-Sequence Pre-training <https://arxiv.org/abs/2001.04063>`__ by Yu Yan,
     Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
-33. :doc:`XLM-RoBERTa <model_doc/xlmroberta>` (from Facebook AI), released together with the paper `Unsupervised
+34. :doc:`XLM-RoBERTa <model_doc/xlmroberta>` (from Facebook AI), released together with the paper `Unsupervised
     Cross-lingual Representation Learning at Scale <https://arxiv.org/abs/1911.02116>`__ by Alexis Conneau*, Kartikay
     Khandelwal*, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke
     Zettlemoyer and Veselin Stoyanov.
-34. :doc:`XLNet <model_doc/xlnet>` (from Google/CMU) released with the paper `​XLNet: Generalized Autoregressive
+35. :doc:`XLNet <model_doc/xlnet>` (from Google/CMU) released with the paper `​XLNet: Generalized Autoregressive
     Pretraining for Language Understanding <https://arxiv.org/abs/1906.08237>`__ by Zhilin Yang*, Zihang Dai*, Yiming
     Yang, Jaime Carbonell, Ruslan Salakhutdinov, Quoc V. Le.
 
@@ -240,6 +243,8 @@ TensorFlow and/or Flax.
 +-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
 |         Longformer          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 +-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
+|            MPNet            |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
++-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
 |           Marian            |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
 +-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
 |         MobileBERT          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
@@ -279,7 +284,6 @@ TensorFlow and/or Flax.
 |             mT5             |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 +-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
 
-
 .. toctree::
     :maxdepth: 2
     :caption: Get started
@@ -366,6 +370,7 @@ TensorFlow and/or Flax.
     model_doc/marian
     model_doc/mbart
     model_doc/mobilebert
+    model_doc/mpnet
     model_doc/mt5
     model_doc/gpt
     model_doc/gpt2

docs/source/model_doc/mpnet.rst

+MPNet
+-----------------------------------------------------------------------------------------------------------------------
+
+Overview
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The MPNet model was proposed in `MPNet: Masked and Permuted Pre-training for Language Understanding
+<https://arxiv.org/abs/2004.09297>`__ by Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu.
+
+MPNet adopts a novel pre-training method, named masked and permuted language modeling, to inherit the advantages of
+masked language modeling and permuted language modeling for natural language understanding.
+
+The abstract from the paper is the following:
+
+*BERT adopts masked language modeling (MLM) for pre-training and is one of the most successful pre-training models.
+Since BERT neglects dependency among predicted tokens, XLNet introduces permuted language modeling (PLM) for
+pre-training to address this problem. However, XLNet does not leverage the full position information of a sentence and
+thus suffers from position discrepancy between pre-training and fine-tuning. In this paper, we propose MPNet, a novel
+pre-training method that inherits the advantages of BERT and XLNet and avoids their limitations. MPNet leverages the
+dependency among predicted tokens through permuted language modeling (vs. MLM in BERT), and takes auxiliary position
+information as input to make the model see a full sentence and thus reducing the position discrepancy (vs. PLM in
+XLNet). We pre-train MPNet on a large-scale dataset (over 160GB text corpora) and fine-tune on a variety of
+down-streaming tasks (GLUE, SQuAD, etc). Experimental results show that MPNet outperforms MLM and PLM by a large
+margin, and achieves better results on these tasks compared with previous state-of-the-art pre-trained methods (e.g.,
+BERT, XLNet, RoBERTa) under the same model setting.*
+
+Tips:
+
+- MPNet doesn't have :obj:`token_type_ids`, you don't need to indicate which token belongs to which segment. just
+  separate your segments with the separation token :obj:`tokenizer.sep_token` (or :obj:`[sep]`).
+
+The original code can be found `here <https://github.com/microsoft/MPNet>`__.
+
+MPNetConfig
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetConfig
+    :members:
+
+
+MPNetTokenizer
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetTokenizer
+    :members: build_inputs_with_special_tokens, get_special_tokens_mask,
+        create_token_type_ids_from_sequences, save_vocabulary
+
+
+MPNetTokenizerFast
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetTokenizerFast
+    :members:
+
+
+MPNetModel
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetModel
+    :members: forward
+
+
+MPNetForMaskedLM
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetForMaskedLM
+    :members: forward
+
+
+MPNetForSequenceClassification
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetForSequenceClassification
+    :members: forward
+
+
+MPNetForMultipleChoice
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetForMultipleChoice
+    :members: forward
+
+
+MPNetForTokenClassification
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetForTokenClassification
+    :members: forward
+
+
+MPNetForQuestionAnswering
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.MPNetForQuestionAnswering
+    :members: forward
+
+
+TFMPNetModel
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.TFMPNetModel
+    :members: call
+
+
+TFMPNetForMaskedLM
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.TFMPNetForMaskedLM
+    :members: call
+
+
+TFMPNetForSequenceClassification
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.TFMPNetForSequenceClassification
+    :members: call
+
+
+TFMPNetForMultipleChoice
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.TFMPNetForMultipleChoice
+    :members: call
+
+
+TFMPNetForTokenClassification
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.TFMPNetForTokenClassification
+    :members: call
+
+
+TFMPNetForQuestionAnswering
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.TFMPNetForQuestionAnswering
+    :members: call

src/transformers/__init__.py

@@ -152,6 +152,7 @@ from .models.marian import MarianConfig
 from .models.mbart import MBartConfig
 from .models.mmbt import MMBTConfig
 from .models.mobilebert import MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MobileBertConfig, MobileBertTokenizer
+from .models.mpnet import MPNET_PRETRAINED_CONFIG_ARCHIVE_MAP, MPNetConfig, MPNetTokenizer
 from .models.mt5 import MT5Config
 from .models.openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig, OpenAIGPTTokenizer
 from .models.pegasus import PegasusConfig
@@ -255,6 +256,7 @@ if is_tokenizers_available():
     from .models.lxmert import LxmertTokenizerFast
     from .models.mbart import MBartTokenizerFast
     from .models.mobilebert import MobileBertTokenizerFast
+    from .models.mpnet import MPNetTokenizerFast
     from .models.mt5 import MT5TokenizerFast
     from .models.openai import OpenAIGPTTokenizerFast
     from .models.pegasus import PegasusTokenizerFast
@@ -530,6 +532,17 @@ if is_torch_available():
         MobileBertPreTrainedModel,
         load_tf_weights_in_mobilebert,
     )
+    from .models.mpnet import (
+        MPNET_PRETRAINED_MODEL_ARCHIVE_LIST,
+        MPNetForMaskedLM,
+        MPNetForMultipleChoice,
+        MPNetForQuestionAnswering,
+        MPNetForSequenceClassification,
+        MPNetForTokenClassification,
+        MPNetLayer,
+        MPNetModel,
+        MPNetPreTrainedModel,
+    )
     from .models.mt5 import MT5EncoderModel, MT5ForConditionalGeneration, MT5Model
     from .models.openai import (
         OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST,
@@ -830,6 +843,17 @@ if is_tf_available():
         TFMobileBertModel,
         TFMobileBertPreTrainedModel,
     )
+    from .models.mpnet import (
+        TF_MPNET_PRETRAINED_MODEL_ARCHIVE_LIST,
+        TFMPNetForMaskedLM,
+        TFMPNetForMultipleChoice,
+        TFMPNetForQuestionAnswering,
+        TFMPNetForSequenceClassification,
+        TFMPNetForTokenClassification,
+        TFMPNetMainLayer,
+        TFMPNetModel,
+        TFMPNetPreTrainedModel,
+    )
     from .models.mt5 import TFMT5EncoderModel, TFMT5ForConditionalGeneration, TFMT5Model
     from .models.openai import (
         TF_OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST,

src/transformers/data/processors/squad.py

@@ -28,7 +28,7 @@ from .utils import DataProcessor
 
 
 # Store the tokenizers which insert 2 separators tokens
-MULTI_SEP_TOKENS_TOKENIZERS_SET = {"roberta", "camembert", "bart"}
+MULTI_SEP_TOKENS_TOKENIZERS_SET = {"roberta", "camembert", "bart", "mpnet"}
 
 
 if is_torch_available():

src/transformers/models/auto/configuration_auto.py

@@ -40,6 +40,7 @@ from ..lxmert.configuration_lxmert import LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
 from ..marian.configuration_marian import MarianConfig
 from ..mbart.configuration_mbart import MBART_PRETRAINED_CONFIG_ARCHIVE_MAP, MBartConfig
 from ..mobilebert.configuration_mobilebert import MobileBertConfig
+from ..mpnet.configuration_mpnet import MPNET_PRETRAINED_CONFIG_ARCHIVE_MAP, MPNetConfig
 from ..mt5.configuration_mt5 import MT5Config
 from ..openai.configuration_openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig
 from ..pegasus.configuration_pegasus import PegasusConfig
@@ -93,6 +94,7 @@ ALL_PRETRAINED_CONFIG_ARCHIVE_MAP = dict(
         SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
         XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
         PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        MPNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ]
     for key, value, in pretrained_map.items()
 )
@@ -113,6 +115,7 @@ CONFIG_MAPPING = OrderedDict(
         ("pegasus", PegasusConfig),
         ("marian", MarianConfig),
         ("mbart", MBartConfig),
+        ("mpnet", MPNetConfig),
         ("bart", BartConfig),
         ("blenderbot", BlenderbotConfig),
         ("reformer", ReformerConfig),
@@ -181,6 +184,7 @@ MODEL_NAMES_MAPPING = OrderedDict(
         ("xlm-prophetnet", "XLMProphetNet"),
         ("prophetnet", "ProphetNet"),
         ("mt5", "mT5"),
+        ("mpnet", "MPNet"),
     ]
 )
 

src/transformers/models/auto/modeling_auto.py

@@ -122,6 +122,14 @@ from ..mobilebert.modeling_mobilebert import (
     MobileBertForTokenClassification,
     MobileBertModel,
 )
+from ..mpnet.modeling_mpnet import (
+    MPNetForMaskedLM,
+    MPNetForMultipleChoice,
+    MPNetForQuestionAnswering,
+    MPNetForSequenceClassification,
+    MPNetForTokenClassification,
+    MPNetModel,
+)
 from ..mt5.modeling_mt5 import MT5ForConditionalGeneration, MT5Model
 from ..openai.modeling_openai import OpenAIGPTForSequenceClassification, OpenAIGPTLMHeadModel, OpenAIGPTModel
 from ..pegasus.modeling_pegasus import PegasusForConditionalGeneration
@@ -212,6 +220,7 @@ from .configuration_auto import (
     MarianConfig,
     MBartConfig,
     MobileBertConfig,
+    MPNetConfig,
     MT5Config,
     OpenAIGPTConfig,
     PegasusConfig,
@@ -267,6 +276,7 @@ MODEL_MAPPING = OrderedDict(
         (DPRConfig, DPRQuestionEncoder),
         (XLMProphetNetConfig, XLMProphetNetModel),
         (ProphetNetConfig, ProphetNetModel),
+        (MPNetConfig, MPNetModel),
     ]
 )
 
@@ -297,6 +307,7 @@ MODEL_FOR_PRETRAINING_MAPPING = OrderedDict(
         (ElectraConfig, ElectraForPreTraining),
         (LxmertConfig, LxmertForPreTraining),
         (FunnelConfig, FunnelForPreTraining),
+        (MPNetConfig, MPNetForMaskedLM),
     ]
 )
 
@@ -328,6 +339,7 @@ MODEL_WITH_LM_HEAD_MAPPING = OrderedDict(
         (EncoderDecoderConfig, EncoderDecoderModel),
         (ReformerConfig, ReformerModelWithLMHead),
         (FunnelConfig, FunnelForMaskedLM),
+        (MPNetConfig, MPNetForMaskedLM),
     ]
 )
 
@@ -373,6 +385,7 @@ MODEL_FOR_MASKED_LM_MAPPING = OrderedDict(
         (ElectraConfig, ElectraForMaskedLM),
         (ReformerConfig, ReformerForMaskedLM),
         (FunnelConfig, FunnelForMaskedLM),
+        (MPNetConfig, MPNetForMaskedLM),
     ]
 )
 
@@ -417,6 +430,7 @@ MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING = OrderedDict(
         (ReformerConfig, ReformerForSequenceClassification),
         (CTRLConfig, CTRLForSequenceClassification),
         (TransfoXLConfig, TransfoXLForSequenceClassification),
+        (MPNetConfig, MPNetForSequenceClassification),
     ]
 )
 
@@ -440,6 +454,7 @@ MODEL_FOR_QUESTION_ANSWERING_MAPPING = OrderedDict(
         (ReformerConfig, ReformerForQuestionAnswering),
         (FunnelConfig, FunnelForQuestionAnswering),
         (LxmertConfig, LxmertForQuestionAnswering),
+        (MPNetConfig, MPNetForQuestionAnswering),
     ]
 )
 
@@ -462,6 +477,7 @@ MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = OrderedDict(
         (ElectraConfig, ElectraForTokenClassification),
         (FlaubertConfig, FlaubertForTokenClassification),
         (FunnelConfig, FunnelForTokenClassification),
+        (MPNetConfig, MPNetForTokenClassification),
     ]
 )
 
@@ -482,6 +498,7 @@ MODEL_FOR_MULTIPLE_CHOICE_MAPPING = OrderedDict(
         (XLMConfig, XLMForMultipleChoice),
         (FlaubertConfig, FlaubertForMultipleChoice),
         (FunnelConfig, FunnelForMultipleChoice),
+        (MPNetConfig, MPNetForMultipleChoice),
     ]
 )
 

src/transformers/models/auto/modeling_tf_auto.py

@@ -111,6 +111,14 @@ from ..mobilebert.modeling_tf_mobilebert import (
     TFMobileBertForTokenClassification,
     TFMobileBertModel,
 )
+from ..mpnet.modeling_tf_mpnet import (
+    TFMPNetForMaskedLM,
+    TFMPNetForMultipleChoice,
+    TFMPNetForQuestionAnswering,
+    TFMPNetForSequenceClassification,
+    TFMPNetForTokenClassification,
+    TFMPNetModel,
+)
 from ..mt5.modeling_tf_mt5 import TFMT5ForConditionalGeneration, TFMT5Model
 from ..openai.modeling_tf_openai import TFOpenAIGPTLMHeadModel, TFOpenAIGPTModel
 from ..pegasus.modeling_tf_pegasus import TFPegasusForConditionalGeneration
@@ -167,6 +175,7 @@ from .configuration_auto import (
     MarianConfig,
     MBartConfig,
     MobileBertConfig,
+    MPNetConfig,
     MT5Config,
     OpenAIGPTConfig,
     PegasusConfig,
@@ -208,6 +217,7 @@ TF_MODEL_MAPPING = OrderedDict(
         (ElectraConfig, TFElectraModel),
         (FunnelConfig, TFFunnelModel),
         (DPRConfig, TFDPRQuestionEncoder),
+        (MPNetConfig, TFMPNetModel),
     ]
 )
 
@@ -233,6 +243,7 @@ TF_MODEL_FOR_PRETRAINING_MAPPING = OrderedDict(
         (CTRLConfig, TFCTRLLMHeadModel),
         (ElectraConfig, TFElectraForPreTraining),
         (FunnelConfig, TFFunnelForPreTraining),
+        (MPNetConfig, TFMPNetForMaskedLM),
     ]
 )
 
@@ -259,6 +270,7 @@ TF_MODEL_WITH_LM_HEAD_MAPPING = OrderedDict(
         (CTRLConfig, TFCTRLLMHeadModel),
         (ElectraConfig, TFElectraForMaskedLM),
         (FunnelConfig, TFFunnelForMaskedLM),
+        (MPNetConfig, TFMPNetForMaskedLM),
     ]
 )
 
@@ -293,6 +305,7 @@ TF_MODEL_FOR_MASKED_LM_MAPPING = OrderedDict(
         (XLMConfig, TFXLMWithLMHeadModel),
         (ElectraConfig, TFElectraForMaskedLM),
         (FunnelConfig, TFFunnelForMaskedLM),
+        (MPNetConfig, TFMPNetForMaskedLM),
     ]
 )
 
@@ -327,6 +340,7 @@ TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING = OrderedDict(
         (ElectraConfig, TFElectraForSequenceClassification),
         (FunnelConfig, TFFunnelForSequenceClassification),
         (GPT2Config, TFGPT2ForSequenceClassification),
+        (MPNetConfig, TFMPNetForSequenceClassification),
     ]
 )
 
@@ -346,6 +360,7 @@ TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING = OrderedDict(
         (XLMConfig, TFXLMForQuestionAnsweringSimple),
         (ElectraConfig, TFElectraForQuestionAnswering),
         (FunnelConfig, TFFunnelForQuestionAnswering),
+        (MPNetConfig, TFMPNetForQuestionAnswering),
     ]
 )
 
@@ -365,6 +380,7 @@ TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = OrderedDict(
         (XLNetConfig, TFXLNetForTokenClassification),
         (ElectraConfig, TFElectraForTokenClassification),
         (FunnelConfig, TFFunnelForTokenClassification),
+        (MPNetConfig, TFMPNetForTokenClassification),
     ]
 )
 
@@ -384,6 +400,7 @@ TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING = OrderedDict(
         (AlbertConfig, TFAlbertForMultipleChoice),
         (ElectraConfig, TFElectraForMultipleChoice),
         (FunnelConfig, TFFunnelForMultipleChoice),
+        (MPNetConfig, TFMPNetForMultipleChoice),
     ]
 )
 

src/transformers/models/auto/tokenization_auto.py

@@ -39,6 +39,7 @@ from ..layoutlm.tokenization_layoutlm import LayoutLMTokenizer
 from ..longformer.tokenization_longformer import LongformerTokenizer
 from ..lxmert.tokenization_lxmert import LxmertTokenizer
 from ..mobilebert.tokenization_mobilebert import MobileBertTokenizer
+from ..mpnet.tokenization_mpnet import MPNetTokenizer
 from ..openai.tokenization_openai import OpenAIGPTTokenizer
 from ..phobert.tokenization_phobert import PhobertTokenizer
 from ..prophetnet.tokenization_prophetnet import ProphetNetTokenizer
@@ -72,6 +73,7 @@ from .configuration_auto import (
     MarianConfig,
     MBartConfig,
     MobileBertConfig,
+    MPNetConfig,
     MT5Config,
     OpenAIGPTConfig,
     PegasusConfig,
@@ -137,6 +139,7 @@ if is_tokenizers_available():
     from ..lxmert.tokenization_lxmert_fast import LxmertTokenizerFast
     from ..mbart.tokenization_mbart_fast import MBartTokenizerFast
     from ..mobilebert.tokenization_mobilebert_fast import MobileBertTokenizerFast
+    from ..mpnet.tokenization_mpnet_fast import MPNetTokenizerFast
     from ..mt5 import MT5TokenizerFast
     from ..openai.tokenization_openai_fast import OpenAIGPTTokenizerFast
     from ..pegasus.tokenization_pegasus_fast import PegasusTokenizerFast
@@ -164,6 +167,7 @@ else:
     LxmertTokenizerFast = None
     MBartTokenizerFast = None
     MobileBertTokenizerFast = None
+    MPNetTokenizerFast = None
     MT5TokenizerFast = None
     OpenAIGPTTokenizerFast = None
     PegasusTokenizerFast = None
@@ -218,6 +222,7 @@ TOKENIZER_MAPPING = OrderedDict(
         (RagConfig, (RagTokenizer, None)),
         (XLMProphetNetConfig, (XLMProphetNetTokenizer, None)),
         (ProphetNetConfig, (ProphetNetTokenizer, None)),
+        (MPNetConfig, (MPNetTokenizer, MPNetTokenizerFast)),
     ]
 )
 

src/transformers/models/mpnet/__init__.py

+# flake8: noqa
+# There's no way to ignore "F401 '...' imported but unused" warnings in this
+# module, but to preserve other warnings. So, don't check this module at all.
+
+from ...file_utils import is_flax_available, is_tf_available, is_tokenizers_available, is_torch_available
+from .configuration_mpnet import MPNET_PRETRAINED_CONFIG_ARCHIVE_MAP, MPNetConfig
+from .tokenization_mpnet import MPNetTokenizer
+
+
+if is_tokenizers_available():
+    from .tokenization_mpnet_fast import MPNetTokenizerFast
+
+if is_torch_available():
+    from .modeling_mpnet import (
+        MPNET_PRETRAINED_MODEL_ARCHIVE_LIST,
+        MPNetForMaskedLM,
+        MPNetForMultipleChoice,
+        MPNetForQuestionAnswering,
+        MPNetForSequenceClassification,
+        MPNetForTokenClassification,
+        MPNetLayer,
+        MPNetModel,
+        MPNetPreTrainedModel,
+    )
+
+if is_tf_available():
+    from .modeling_tf_mpnet import (
+        TF_MPNET_PRETRAINED_MODEL_ARCHIVE_LIST,
+        TFMPNetEmbeddings,
+        TFMPNetForMaskedLM,
+        TFMPNetForMultipleChoice,
+        TFMPNetForQuestionAnswering,
+        TFMPNetForSequenceClassification,
+        TFMPNetForTokenClassification,
+        TFMPNetMainLayer,
+        TFMPNetModel,
+        TFMPNetPreTrainedModel,
+    )

src/transformers/models/mpnet/configuration_mpnet.py

+# coding=utf-8
+# Copyright 2018 The HuggingFace Inc. team, Microsoft Corporation.
+# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" MPNet model configuration """
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+MPNET_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "microsoft/mpnet-base": "https://huggingface.co/microsoft/mpnet-base/resolve/main/config.json",
+}
+
+
+class MPNetConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a :class:`~transformers.MPNetModel` or a
+    :class:`~transformers.TFMPNetModel`. It is used to instantiate a MPNet model according to the specified arguments,
+    defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration
+    to that of the MPNet `mpnet-base <https://huggingface.co/mpnet-base>`__ architecture.
+
+    Configuration objects inherit from :class:`~transformers.PretrainedConfig` and can be used to control the model
+    outputs. Read the documentation from :class:`~transformers.PretrainedConfig` for more information.
+
+    Args:
+        vocab_size (:obj:`int`, `optional`, defaults to 30527):
+            Vocabulary size of the MPNet model. Defines the number of different tokens that can be represented by the
+            :obj:`inputs_ids` passed when calling :class:`~transformers.MPNetModel` or
+            :class:`~transformers.TFMPNetModel`.
+        hidden_size (:obj:`int`, `optional`, defaults to 768):
+            Dimensionality of the encoder layers and the pooler layer.
+        num_hidden_layers (:obj:`int`, `optional`, defaults to 12):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (:obj:`int`, `optional`, defaults to 12):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        intermediate_size (:obj:`int`, `optional`, defaults to 3072):
+            Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
+        hidden_act (:obj:`str` or :obj:`Callable`, `optional`, defaults to :obj:`"gelu"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string,
+            :obj:`"gelu"`, :obj:`"relu"`, :obj:`"silu"` and :obj:`"gelu_new"` are supported.
+        hidden_dropout_prob (:obj:`float`, `optional`, defaults to 0.1):
+            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
+        attention_probs_dropout_prob (:obj:`float`, `optional`, defaults to 0.1):
+            The dropout ratio for the attention probabilities.
+        max_position_embeddings (:obj:`int`, `optional`, defaults to 512):
+            The maximum sequence length that this model might ever be used with. Typically set this to something large
+            just in case (e.g., 512 or 1024 or 2048).
+        initializer_range (:obj:`float`, `optional`, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        layer_norm_eps (:obj:`float`, `optional`, defaults to 1e-12):
+            The epsilon used by the layer normalization layers.
+        relative_attention_num_buckets (:obj:`int`, `optional`, defaults to 32):
+            The number of buckets to use for each attention layer.
+
+    Examples::
+
+        >>> from transformers import MPNetModel, MPNetConfig
+
+        >>> # Initializing a MPNet mpnet-base style configuration
+        >>> configuration = MPNetConfig()
+
+        >>> # Initializing a model from the mpnet-base style configuration
+        >>> model = MPNetModel(configuration)
+
+        >>> # Accessing the model configuration
+        >>> configuration = model.config
+    """
+    model_type = "mpnet"
+
+    def __init__(
+        self,
+        vocab_size=30527,
+        hidden_size=768,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        intermediate_size=3072,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=512,
+        initializer_range=0.02,
+        layer_norm_eps=1e-12,
+        relative_attention_num_buckets=32,
+        pad_token_id=1,
+        bos_token_id=0,
+        eos_token_id=2,
+        **kwargs,
+    ):
+        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
+
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.hidden_act = hidden_act
+        self.intermediate_size = intermediate_size
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.initializer_range = initializer_range
+        self.layer_norm_eps = layer_norm_eps
+        self.relative_attention_num_buckets = relative_attention_num_buckets

src/transformers/models/mpnet/tokenization_mpnet_fast.py

+# coding=utf-8
+# Copyright 2018 The HuggingFace Inc. team, Microsoft Corporation.
+# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Fast Tokenization classes for MPNet."""
+
+import json
+from typing import List, Optional, Tuple
+
+from tokenizers import normalizers
+
+from ...tokenization_utils import AddedToken
+from ...tokenization_utils_fast import PreTrainedTokenizerFast
+from ...utils import logging
+from .tokenization_mpnet import MPNetTokenizer
+
+
+logger = logging.get_logger(__name__)
+
+VOCAB_FILES_NAMES = {"vocab_file": "vocab.txt", "tokenizer_file": "tokenizer.json"}
+
+PRETRAINED_VOCAB_FILES_MAP = {
+    "vocab_file": {
+        "microsoft/mpnet-base": "https://huggingface.co/microsoft/mpnet-base/resolve/main/vocab.txt",
+    },
+    "tokenizer_file": {
+        "microsoft/mpnet-base": "https://huggingface.co/microsoft/mpnet-base/resolve/main/tokenizer.json",
+    },
+}
+
+PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
+    "microsoft/mpnet-base": 512,
+}
+
+PRETRAINED_INIT_CONFIGURATION = {
+    "microsoft/mpnet-base": {"do_lower_case": True},
+}
+
+
+class MPNetTokenizerFast(PreTrainedTokenizerFast):
+    r"""
+    Construct a "fast" MPNet tokenizer (backed by HuggingFace's `tokenizers` library). Based on WordPiece.
+
+    This tokenizer inherits from :class:`~transformers.PreTrainedTokenizerFast` which contains most of the main
+    methods. Users should refer to this superclass for more information regarding those methods.
+
+    Args:
+        vocab_file (:obj:`str`):
+            File containing the vocabulary.
+        do_lower_case (:obj:`bool`, `optional`, defaults to :obj:`True`):
+            Whether or not to lowercase the input when tokenizing.
+        bos_token (:obj:`str`, `optional`, defaults to :obj:`"<s>"`):
+            The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.
+
+            .. note::
+
+                When building a sequence using special tokens, this is not the token that is used for the beginning of
+                sequence. The token used is the :obj:`cls_token`.
+        eos_token (:obj:`str`, `optional`, defaults to :obj:`"</s>"`):
+            The end of sequence token.
+
+            .. note::
+
+                When building a sequence using special tokens, this is not the token that is used for the end of
+                sequence. The token used is the :obj:`sep_token`.
+        sep_token (:obj:`str`, `optional`, defaults to :obj:`"</s>"`):
+            The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences for
+            sequence classification or for a text and a question for question answering. It is also used as the last
+            token of a sequence built with special tokens.
+        cls_token (:obj:`str`, `optional`, defaults to :obj:`"<s>"`):
+            The classifier token which is used when doing sequence classification (classification of the whole sequence
+            instead of per-token classification). It is the first token of the sequence when built with special tokens.
+        unk_token (:obj:`str`, `optional`, defaults to :obj:`"[UNK]"`):
+            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
+            token instead.
+        pad_token (:obj:`str`, `optional`, defaults to :obj:`"<pad>"`):
+            The token used for padding, for example when batching sequences of different lengths.
+        mask_token (:obj:`str`, `optional`, defaults to :obj:`"<mask>"`):
+            The token used for masking values. This is the token used when training this model with masked language
+            modeling. This is the token which the model will try to predict.
+        tokenize_chinese_chars (:obj:`bool`, `optional`, defaults to :obj:`True`):
+            Whether or not to tokenize Chinese characters. This should likely be deactivated for Japanese (see `this
+            issue <https://github.com/huggingface/transformers/issues/328>`__).
+        strip_accents: (:obj:`bool`, `optional`):
+            Whether or not to strip all accents. If this option is not specified, then it will be determined by the
+            value for :obj:`lowercase` (as in the original BERT).
+    """
+
+    vocab_files_names = VOCAB_FILES_NAMES
+    pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
+    pretrained_init_configuration = PRETRAINED_INIT_CONFIGURATION
+    max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
+    slow_tokenizer_class = MPNetTokenizer
+
+    def __init__(
+        self,
+        vocab_file,
+        tokenizer_file=None,
+        do_lower_case=True,
+        bos_token="<s>",
+        eos_token="</s>",
+        sep_token="</s>",
+        cls_token="<s>",
+        unk_token="[UNK]",
+        pad_token="<pad>",
+        mask_token="<mask>",
+        tokenize_chinese_chars=True,
+        strip_accents=None,
+        **kwargs
+    ):
+        super().__init__(
+            vocab_file,
+            tokenizer_file=tokenizer_file,
+            do_lower_case=do_lower_case,
+            bos_token=bos_token,
+            eos_token=eos_token,
+            sep_token=sep_token,
+            cls_token=cls_token,
+            unk_token=unk_token,
+            pad_token=pad_token,
+            mask_token=mask_token,
+            tokenize_chinese_chars=tokenize_chinese_chars,
+            strip_accents=strip_accents,
+            **kwargs,
+        )
+
+        pre_tok_state = json.loads(self.backend_tokenizer.normalizer.__getstate__())
+        if (
+            pre_tok_state.get("do_lower_case", do_lower_case) != do_lower_case
+            or pre_tok_state.get("strip_accents", strip_accents) != strip_accents
+        ):
+            pre_tok_class = getattr(normalizers, pre_tok_state.pop("type"))
+            pre_tok_state["do_lower_case"] = do_lower_case
+            pre_tok_state["strip_accents"] = strip_accents
+            self.backend_tokenizer.normalizer = pre_tok_class(**pre_tok_state)
+
+        self.do_lower_case = do_lower_case
+
+    @property
+    def mask_token(self) -> str:
+        """
+        :obj:`str`: Mask token, to use when training a model with masked-language modeling. Log an error if used while
+        not having been set.
+
+        MPNet tokenizer has a special mask token to be usble in the fill-mask pipeline. The mask token will greedily
+        comprise the space before the `<mask>`.
+        """
+        if self._mask_token is None and self.verbose:
+            logger.error("Using mask_token, but it is not set yet.")
+            return None
+        return str(self._mask_token)
+
+    @mask_token.setter
+    def mask_token(self, value):
+        """
+        Overriding the default behavior of the mask token to have it eat the space before it.
+
+        This is needed to preserve backward compatibility with all the previously used models based on MPNet.
+        """
+        # Mask token behave like a normal word, i.e. include the space before it
+        # So we set lstrip to True
+        value = AddedToken(value, lstrip=True, rstrip=False) if isinstance(value, str) else value
+        self._mask_token = value
+
+    def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
+        output = [self.bos_token_id] + token_ids_0 + [self.eos_token_id]
+        if token_ids_1 is None:
+            return output
+
+        return output + [self.eos_token_id] + token_ids_1 + [self.eos_token_id]
+
+    def create_token_type_ids_from_sequences(
+        self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
+    ) -> List[int]:
+        """
+        Creates a mask from the two sequences passed to be used in a sequence-pair classification task. MPNet does not
+        make use of token type ids, therefore a list of zeros is returned
+
+        Args:
+            token_ids_0 (:obj:`List[int]`):
+                List of ids.
+            token_ids_1 (:obj:`List[int]`, `optional`):
+                Optional second list of IDs for sequence pairs
+
+        Returns:
+            :obj:`List[int]`: List of zeros.
+        """
+        sep = [self.sep_token_id]
+        cls = [self.cls_token_id]
+
+        if token_ids_1 is None:
+            return len(cls + token_ids_0 + sep) * [0]
+        return len(cls + token_ids_0 + sep + sep + token_ids_1 + sep) * [0]
+
+    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        files = self._tokenizer.model.save(save_directory, name=filename_prefix)
+        return tuple(files)

src/transformers/utils/dummy_pt_objects.py

@@ -1380,6 +1380,77 @@ def load_tf_weights_in_mobilebert(*args, **kwargs):
     requires_pytorch(load_tf_weights_in_mobilebert)
 
 
+MPNET_PRETRAINED_MODEL_ARCHIVE_LIST = None
+
+
+class MPNetForMaskedLM:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+class MPNetForMultipleChoice:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+class MPNetForQuestionAnswering:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+class MPNetForSequenceClassification:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+class MPNetForTokenClassification:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+class MPNetLayer:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+class MPNetModel:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
+class MPNetPreTrainedModel:
+    def __init__(self, *args, **kwargs):
+        requires_pytorch(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_pytorch(self)
+
+
 class MT5EncoderModel:
     def __init__(self, *args, **kwargs):
         requires_pytorch(self)

src/transformers/utils/dummy_tf_objects.py

@@ -1006,6 +1006,77 @@ class TFMobileBertPreTrainedModel:
         requires_tf(self)
 
 
+TF_MPNET_PRETRAINED_MODEL_ARCHIVE_LIST = None
+
+
+class TFMPNetForMaskedLM:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFMPNetForMultipleChoice:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFMPNetForQuestionAnswering:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFMPNetForSequenceClassification:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFMPNetForTokenClassification:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFMPNetMainLayer:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFMPNetModel:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tf(self)
+
+
+class TFMPNetPreTrainedModel:
+    def __init__(self, *args, **kwargs):
+        requires_tf(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tf(self)
+
+
 class TFMT5EncoderModel:
     def __init__(self, *args, **kwargs):
         requires_tf(self)

src/transformers/utils/dummy_tokenizers_objects.py

@@ -164,6 +164,15 @@ class MobileBertTokenizerFast:
         requires_tokenizers(self)
 
 
+class MPNetTokenizerFast:
+    def __init__(self, *args, **kwargs):
+        requires_tokenizers(self)
+
+    @classmethod
+    def from_pretrained(self, *args, **kwargs):
+        requires_tokenizers(self)
+
+
 class MT5TokenizerFast:
     def __init__(self, *args, **kwargs):
         requires_tokenizers(self)

tests/test_modeling_mpnet.py

+# coding=utf-8
+# Copyright 2020 The HuggingFace Inc. team, Microsoft Corporation.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import unittest
+
+from transformers import is_torch_available
+from transformers.testing_utils import require_torch, slow, torch_device
+
+from .test_configuration_common import ConfigTester
+from .test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask
+
+
+if is_torch_available():
+    import torch
+
+    from transformers import (
+        MPNetConfig,
+        MPNetForMaskedLM,
+        MPNetForMultipleChoice,
+        MPNetForQuestionAnswering,
+        MPNetForSequenceClassification,
+        MPNetForTokenClassification,
+        MPNetModel,
+    )
+
+
+class MPNetModelTester:
+    def __init__(
+        self,
+        parent,
+        batch_size=13,
+        seq_length=7,
+        is_training=True,
+        use_input_mask=True,
+        use_token_type_ids=False,
+        use_labels=True,
+        vocab_size=99,
+        hidden_size=64,
+        num_hidden_layers=5,
+        num_attention_heads=4,
+        intermediate_size=64,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=512,
+        type_vocab_size=16,
+        type_sequence_label_size=2,
+        initializer_range=0.02,
+        num_labels=3,
+        num_choices=4,
+        scope=None,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.seq_length = seq_length
+        self.is_training = is_training
+        self.use_input_mask = use_input_mask
+        self.use_token_type_ids = use_token_type_ids
+        self.use_labels = use_labels
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.hidden_act = hidden_act
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.type_sequence_label_size = type_sequence_label_size
+        self.initializer_range = initializer_range
+        self.num_labels = num_labels
+        self.num_choices = num_choices
+        self.scope = scope
+
+    def get_large_model_config(self):
+        return MPNetConfig.from_pretrained("microsoft/mpnet-base")
+
+    def prepare_config_and_inputs(self):
+        input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
+
+        input_mask = None
+        if self.use_input_mask:
+            input_mask = random_attention_mask([self.batch_size, self.seq_length])
+
+        sequence_labels = None
+        token_labels = None
+        choice_labels = None
+        if self.use_labels:
+            sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
+            token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
+            choice_labels = ids_tensor([self.batch_size], self.num_choices)
+
+        config = MPNetConfig(
+            vocab_size=self.vocab_size,
+            hidden_size=self.hidden_size,
+            num_hidden_layers=self.num_hidden_layers,
+            num_attention_heads=self.num_attention_heads,
+            intermediate_size=self.intermediate_size,
+            hidden_act=self.hidden_act,
+            hidden_dropout_prob=self.hidden_dropout_prob,
+            attention_probs_dropout_prob=self.attention_probs_dropout_prob,
+            max_position_embeddings=self.max_position_embeddings,
+            initializer_range=self.initializer_range,
+        )
+        return config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+
+    def create_and_check_mpnet_model(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = MPNetModel(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, input_mask)
+        result = model(input_ids)
+        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
+        self.parent.assertEqual(result.pooler_output.shape, (self.batch_size, self.hidden_size))
+
+    def create_and_check_mpnet_for_question_answering(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = MPNetForQuestionAnswering(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(
+            input_ids,
+            attention_mask=input_mask,
+            start_positions=sequence_labels,
+            end_positions=sequence_labels,
+        )
+        self.parent.assertEqual(result.start_logits.shape, (self.batch_size, self.seq_length))
+        self.parent.assertEqual(result.end_logits.shape, (self.batch_size, self.seq_length))
+
+    def create_and_check_mpnet_for_sequence_classification(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = MPNetForSequenceClassification(config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, attention_mask=input_mask, labels=sequence_labels)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_labels))
+
+    def create_and_check_mpnet_for_multiple_choice(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_choices = self.num_choices
+        model = MPNetForMultipleChoice(config=config)
+        model.to(torch_device)
+        model.eval()
+        multiple_choice_inputs_ids = input_ids.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
+        multiple_choice_input_mask = input_mask.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
+        result = model(
+            multiple_choice_inputs_ids,
+            attention_mask=multiple_choice_input_mask,
+            labels=choice_labels,
+        )
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_choices))
+
+    def create_and_check_mpnet_for_token_classification(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = MPNetForTokenClassification(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, attention_mask=input_mask, labels=token_labels)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        (config, input_ids, input_mask, sequence_labels, token_labels, choice_labels) = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask}
+        return config, inputs_dict
+
+
+@require_torch
+class MPNetModelTest(ModelTesterMixin, unittest.TestCase):
+
+    all_model_classes = (
+        (
+            MPNetForMaskedLM,
+            MPNetForMultipleChoice,
+            MPNetForQuestionAnswering,
+            MPNetForSequenceClassification,
+            MPNetForTokenClassification,
+            MPNetModel,
+        )
+        if is_torch_available()
+        else ()
+    )
+    test_pruning = False
+    test_torchscript = True
+    test_resize_embeddings = True
+
+    def setUp(self):
+        self.model_tester = MPNetModelTester(self)
+        self.config_tester = ConfigTester(self, config_class=MPNetConfig, hidden_size=37)
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    def test_mpnet_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_model(*config_and_inputs)
+
+    def test_for_sequence_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_sequence_classification(*config_and_inputs)
+
+    def test_for_multiple_choice(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_multiple_choice(*config_and_inputs)
+
+    def test_for_token_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_token_classification(*config_and_inputs)
+
+    def test_for_question_answering(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_question_answering(*config_and_inputs)
+
+
+@require_torch
+class MPNetModelIntegrationTest(unittest.TestCase):
+    @slow
+    def test_inference_no_head(self):
+        model = MPNetModel.from_pretrained("microsoft/mpnet-base")
+        input_ids = torch.tensor([[0, 345, 232, 328, 740, 140, 1695, 69, 6078, 1588, 2]])
+        output = model(input_ids)[0]
+        expected_shape = torch.Size((1, 11, 768))
+        self.assertEqual(output.shape, expected_shape)
+        expected_slice = torch.tensor(
+            [[[-0.0550, 0.1943, -0.0740], [-0.0562, 0.2211, -0.0579], [-0.0437, 0.3337, -0.0641]]]
+        )
+        # compare the actual values for a slice.
+        self.assertTrue(torch.allclose(output[:, :3, :3], expected_slice, atol=1e-4))

tests/test_modeling_tf_mpnet.py

+# coding=utf-8
+# Copyright 2020 The HuggingFace Inc. team, Microsoft Corporation.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import unittest
+
+from transformers import MPNetConfig, is_tf_available
+from transformers.testing_utils import require_tf, slow
+
+from .test_configuration_common import ConfigTester
+from .test_modeling_tf_common import TFModelTesterMixin, ids_tensor
+
+
+if is_tf_available():
+    import tensorflow as tf
+
+    from transformers.models.mpnet.modeling_tf_mpnet import (
+        TFMPNetForMaskedLM,
+        TFMPNetForMultipleChoice,
+        TFMPNetForQuestionAnswering,
+        TFMPNetForSequenceClassification,
+        TFMPNetForTokenClassification,
+        TFMPNetModel,
+    )
+
+
+class TFMPNetModelTester:
+    def __init__(
+        self,
+        parent,
+        batch_size=13,
+        seq_length=7,
+        is_training=True,
+        use_input_mask=True,
+        use_token_type_ids=False,
+        use_labels=True,
+        vocab_size=99,
+        hidden_size=64,
+        num_hidden_layers=5,
+        num_attention_heads=4,
+        intermediate_size=64,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=512,
+        type_vocab_size=16,
+        type_sequence_label_size=2,
+        initializer_range=0.02,
+        num_labels=3,
+        num_choices=4,
+        scope=None,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.seq_length = seq_length
+        self.is_training = is_training
+        self.use_input_mask = use_input_mask
+        self.use_token_type_ids = use_token_type_ids
+        self.use_labels = use_labels
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.hidden_act = hidden_act
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.type_sequence_label_size = type_sequence_label_size
+        self.initializer_range = initializer_range
+        self.num_labels = num_labels
+        self.num_choices = num_choices
+        self.scope = scope
+
+    def prepare_config_and_inputs(self):
+        input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
+
+        input_mask = None
+        if self.use_input_mask:
+            input_mask = ids_tensor([self.batch_size, self.seq_length], vocab_size=2)
+
+        sequence_labels = None
+        token_labels = None
+        choice_labels = None
+        if self.use_labels:
+            sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
+            token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
+            choice_labels = ids_tensor([self.batch_size], self.num_choices)
+
+        config = MPNetConfig(
+            vocab_size=self.vocab_size,
+            hidden_size=self.hidden_size,
+            num_hidden_layers=self.num_hidden_layers,
+            num_attention_heads=self.num_attention_heads,
+            intermediate_size=self.intermediate_size,
+            hidden_act=self.hidden_act,
+            hidden_dropout_prob=self.hidden_dropout_prob,
+            attention_probs_dropout_prob=self.attention_probs_dropout_prob,
+            max_position_embeddings=self.max_position_embeddings,
+            initializer_range=self.initializer_range,
+        )
+        return config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+
+    def create_and_check_mpnet_model(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = TFMPNetModel(config=config)
+        inputs = {"input_ids": input_ids, "attention_mask": input_mask}
+        result = model(inputs)
+        inputs = [input_ids, input_mask]
+        result = model(inputs)
+        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
+
+    def create_and_check_mpnet_for_masked_lm(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = TFMPNetForMaskedLM(config=config)
+        inputs = {"input_ids": input_ids, "attention_mask": input_mask}
+        result = model(inputs)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
+
+    def create_and_check_mpnet_for_question_answering(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = TFMPNetForQuestionAnswering(config=config)
+        inputs = {
+            "input_ids": input_ids,
+            "attention_mask": input_mask,
+        }
+        result = model(inputs)
+        self.parent.assertEqual(result.start_logits.shape, (self.batch_size, self.seq_length))
+        self.parent.assertEqual(result.end_logits.shape, (self.batch_size, self.seq_length))
+
+    def create_and_check_mpnet_for_sequence_classification(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = TFMPNetForSequenceClassification(config)
+        inputs = {"input_ids": input_ids, "attention_mask": input_mask}
+        result = model(inputs)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_labels))
+
+    def create_and_check_mpnet_for_multiple_choice(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_choices = self.num_choices
+        model = TFMPNetForMultipleChoice(config)
+        multiple_choice_inputs_ids = tf.tile(tf.expand_dims(input_ids, 1), (1, self.num_choices, 1))
+        multiple_choice_input_mask = tf.tile(tf.expand_dims(input_mask, 1), (1, self.num_choices, 1))
+        inputs = {
+            "input_ids": multiple_choice_inputs_ids,
+            "attention_mask": multiple_choice_input_mask,
+        }
+        result = model(inputs)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_choices))
+
+    def create_and_check_mpnet_for_token_classification(
+        self, config, input_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = TFMPNetForTokenClassification(config)
+        inputs = {"input_ids": input_ids, "attention_mask": input_mask}
+        result = model(inputs)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        (config, input_ids, input_mask, sequence_labels, token_labels, choice_labels) = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask}
+        return config, inputs_dict
+
+
+@require_tf
+class TFMPNetModelTest(TFModelTesterMixin, unittest.TestCase):
+
+    all_model_classes = (
+        (
+            TFMPNetForMaskedLM,
+            TFMPNetForMultipleChoice,
+            TFMPNetForQuestionAnswering,
+            TFMPNetForSequenceClassification,
+            TFMPNetForTokenClassification,
+            TFMPNetModel,
+        )
+        if is_tf_available()
+        else ()
+    )
+
+    def setUp(self):
+        self.model_tester = TFMPNetModelTester(self)
+        self.config_tester = ConfigTester(self, config_class=MPNetConfig, hidden_size=37)
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    def test_mpnet_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_model(*config_and_inputs)
+
+    def test_for_masked_lm(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_masked_lm(*config_and_inputs)
+
+    def test_for_question_answering(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_question_answering(*config_and_inputs)
+
+    def test_for_sequence_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_sequence_classification(*config_and_inputs)
+
+    def test_for_multiple_choice(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_multiple_choice(*config_and_inputs)
+
+    def test_for_token_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_mpnet_for_token_classification(*config_and_inputs)
+
+    @slow
+    def test_model_from_pretrained(self):
+        for model_name in ["microsoft/mpnet-base"]:
+            model = TFMPNetModel.from_pretrained(model_name)
+            self.assertIsNotNone(model)