ProphetNet (#7157)

* add new model prophetnet

prophetnet modified

modify codes as suggested v1

add prophetnet test files

* still bugs, because of changed output formats of encoder and decoder

* move prophetnet into the latest version

* clean integration tests

* clean tokenizers

* add xlm config to init

* correct typo in init

* further refactoring

* continue refactor

* save parallel

* add decoder_attention_mask

* fix use_cache vs. past_key_values

* fix common tests

* change decoder output logits

* fix xlm tests

* make common tests pass

* change model architecture

* add tokenizer tests

* finalize model structure

* no weight mapping

* correct n-gram stream attention mask as discussed with qweizhen

* remove unused import

* fix index.rst

* fix tests

* delete unnecessary code

* add fast integration test

* rename weights

* final weight remapping

* save intermediate

* Descriptions for Prophetnet Config File

* finish all models

* finish new model outputs

* delete unnecessary files

* refactor encoder layer

* add dummy docs

* code quality

* fix tests

* add model pages to doctree

* further refactor

* more refactor, more tests

* finish code refactor and tests

* remove unnecessary files

* further clean up

* add docstring template

* finish tokenizer doc

* finish prophetnet

* fix copies

* fix typos

* fix tf tests

* fix fp16

* fix tf test 2nd try

* fix code quality

* add test for each model

* merge new tests to branch

* Update model_cards/microsoft/prophetnet-large-uncased-cnndm/README.md
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>

* Update model_cards/microsoft/prophetnet-large-uncased-cnndm/README.md
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>

* Update src/transformers/modeling_prophetnet.py
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>

* Update utils/check_repo.py
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>

* apply sams and sylvains comments

* make style

* remove unnecessary code

* Update README.md
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update README.md
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update src/transformers/configuration_prophetnet.py
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>

* implement lysandres comments

* correct docs

* fix isort

* fix tokenizers

* fix copies
Co-authored-by: weizhen <weizhen@mail.ustc.edu.cn>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
Co-authored-by: Sam Shleifer <sshleifer@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>

README.md

@@ -182,6 +182,7 @@ Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
 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. **[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/master/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.
 1. **[Reformer](https://huggingface.co/transformers/model_doc/reformer.html)** (from Google Research) released with the paper [Reformer: The Efficient Transformer](https://arxiv.org/abs/2001.04451) by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.
 1. **[RoBERTa](https://huggingface.co/transformers/model_doc/roberta.html)** (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.
@@ -189,6 +190,7 @@ ultilingual BERT into [DistilmBERT](https://github.com/huggingface/transformers/
 1. **[T5](https://huggingface.co/transformers/model_doc/t5.html)** (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.
 1. **[Transformer-XL](https://huggingface.co/transformers/model_doc/transformerxl.html)** (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.
 1. **[XLM](https://huggingface.co/transformers/model_doc/xlm.html)** (from Facebook) released together with the paper [Cross-lingual Language Model Pretraining](https://arxiv.org/abs/1901.07291) by Guillaume Lample and Alexis Conneau.
+1. **[XLM-ProphetNet](https://huggingface.co/transformers/master/model_doc/xlmprophetnet.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.
 1. **[XLM-RoBERTa](https://huggingface.co/transformers/model_doc/xlmroberta.html)** (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.
 1. **[XLNet](https://huggingface.co/transformers/model_doc/xlnet.html)** (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.
 1. **[Other community models](https://huggingface.co/models)**, contributed by the [community](https://huggingface.co/users).

docs/source/index.rst

@@ -129,32 +129,40 @@ conversion utilities for the following models:
 22. :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.
-23. :doc:`Reformer <model_doc/reformer>` (from Google Research) released with the paper `Reformer: The Efficient
+23. `ProphetNet <https://huggingface.co/transformers/master/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.
+24. :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.
-24. :doc:`RoBERTa <model_doc/roberta>` (from Facebook), released together with the paper a `Robustly Optimized BERT
+25. :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.
-25. `SqueezeBert <https://huggingface.co/transformers/master/model_doc/squeezebert.html>`__ released with the paper
+26. `SqueezeBert <https://huggingface.co/transformers/master/model_doc/squeezebert.html>`__ 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.
-26. :doc:`T5 <model_doc/t5>` (from Google AI) released with the paper `Exploring the Limits of Transfer Learning with a
+27. :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.
-27. :doc:`Transformer-XL <model_doc/transformerxl>` (from Google/CMU) released with the paper `Transformer-XL:
+28. :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.
-28. :doc:`XLM <model_doc/xlm>` (from Facebook) released together with the paper `Cross-lingual Language Model
+29. :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.
-29. :doc:`XLM-RoBERTa <model_doc/xlmroberta>` (from Facebook AI), released together with the paper `Unsupervised
+30. `XLM-ProphetNet <https://huggingface.co/transformers/master/model_doc/xlmprophetnet.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.
+31. :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.
-30. :doc:`XLNet <model_doc/xlnet>` (from Google/CMU) released with the paper `​XLNet: Generalized Autoregressive
+32. :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.
-31. `Other community models <https://huggingface.co/models>`__, contributed by the `community
+33. `Other community models <https://huggingface.co/models>`__, contributed by the `community
     <https://huggingface.co/users>`__.
 
 .. toctree::
@@ -245,6 +253,7 @@ conversion utilities for the following models:
     model_doc/gpt
     model_doc/gpt2
     model_doc/pegasus
+    model_doc/prophetnet
     model_doc/rag
     model_doc/reformer
     model_doc/retribert
@@ -253,6 +262,7 @@ conversion utilities for the following models:
     model_doc/t5
     model_doc/transformerxl
     model_doc/xlm
+    model_doc/xlmprophetnet
     model_doc/xlmroberta
     model_doc/xlnet
 

docs/source/model_doc/encoderdecoder.rst

@@ -27,4 +27,4 @@ EncoderDecoderModel
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 .. autoclass:: transformers.EncoderDecoderModel
-    :members: forward
+    :members: forward, from_encoder_decoder_pretrained

docs/source/model_doc/prophetnet.rst

+ProphetNet
+-----------------------------------------------------------------------------------------------------------------------
+
+**DISCLAIMER:** If you see something strange, file a `Github Issue
+<https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title>`__ and assign
+@patrickvonplaten
+
+Overview
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The ProphetNet model was proposed in `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, Ming Zhou on 13 Jan, 2020.
+
+ProphetNet is an encoder-decoder model and can predict n-future tokens for "ngram" language modeling instead of just the next token.
+
+The abstract from the paper is the following:
+
+*In this paper, we present a new sequence-to-sequence pre-training model called ProphetNet, which introduces a novel self-supervised objective named future n-gram prediction and the proposed n-stream self-attention mechanism. Instead of the optimization of one-step ahead prediction in traditional sequence-to-sequence model, the ProphetNet is optimized by n-step ahead prediction which predicts the next n tokens simultaneously based on previous context tokens at each time step. The future n-gram prediction explicitly encourages the model to plan for the future tokens and prevent overfitting on strong local correlations. We pre-train ProphetNet using a base scale dataset (16GB) and a large scale dataset (160GB) respectively. Then we conduct experiments on CNN/DailyMail, Gigaword, and SQuAD 1.1 benchmarks for abstractive summarization and question generation tasks. Experimental results show that ProphetNet achieves new state-of-the-art results on all these datasets compared to the models using the same scale pre-training corpus.*
+
+The Authors' code can be found `here <https://github.com/microsoft/ProphetNet>`__.
+
+
+ProphetNetConfig
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.ProphetNetConfig
+    :members:
+
+
+ProphetNetTokenizer
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.ProphetNetTokenizer
+    :members:
+
+
+ProphetNet specific outputs
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.modeling_prophetnet.ProphetNetSeq2SeqLMOutput
+    :members:
+
+.. autoclass:: transformers.modeling_prophetnet.ProphetNetSeq2SeqModelOutput
+    :members:
+
+.. autoclass:: transformers.modeling_prophetnet.ProphetNetDecoderModelOutput
+    :members:
+
+.. autoclass:: transformers.modeling_prophetnet.ProphetNetDecoderLMOutput
+    :members:
+
+ProphetNetModel
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.ProphetNetModel
+    :members: forward
+
+
+ProphetNetEncoder
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.ProphetNetEncoder
+    :members: forward
+
+
+ProphetNetDecoder
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.ProphetNetDecoder
+    :members: forward
+
+
+ProphetNetForConditionalGeneration
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.ProphetNetForConditionalGeneration
+    :members: forward
+
+
+ProphetNetForCausalLM
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.ProphetNetForCausalLM
+    :members: forward

docs/source/model_doc/xlmprophetnet.rst

+XLM-ProphetNet
+-----------------------------------------------------------------------------------------------------------------------
+
+**DISCLAIMER:** If you see something strange, file a `Github Issue
+<https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title>`__ and assign
+@patrickvonplaten
+
+
+Overview
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The XLM-ProphetNet model was proposed in `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, Ming Zhou on 13 Jan, 2020.
+
+XLM-ProphetNet is an encoder-decoder model and can predict n-future tokens for "ngram" language modeling instead of just the next token. Its architecture is identical to ProhpetNet, but the model was trained on the multi-lingual "wiki100" Wikipedia dump.
+
+The abstract from the paper is the following:
+
+*In this paper, we present a new sequence-to-sequence pre-training model called ProphetNet, which introduces a novel self-supervised objective named future n-gram prediction and the proposed n-stream self-attention mechanism. Instead of the optimization of one-step ahead prediction in traditional sequence-to-sequence model, the ProphetNet is optimized by n-step ahead prediction which predicts the next n tokens simultaneously based on previous context tokens at each time step. The future n-gram prediction explicitly encourages the model to plan for the future tokens and prevent overfitting on strong local correlations. We pre-train ProphetNet using a base scale dataset (16GB) and a large scale dataset (160GB) respectively. Then we conduct experiments on CNN/DailyMail, Gigaword, and SQuAD 1.1 benchmarks for abstractive summarization and question generation tasks. Experimental results show that ProphetNet achieves new state-of-the-art results on all these datasets compared to the models using the same scale pre-training corpus.*
+
+The Authors' code can be found `here <https://github.com/microsoft/ProphetNet>`__.
+
+XLMProphetNetConfig
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.XLMProphetNetConfig
+    :members:
+
+
+XLMProphetNetTokenizer
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.XLMProphetNetTokenizer
+    :members:
+
+
+XLMProphetNetModel
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.XLMProphetNetModel
+
+
+XLMProphetNetEncoder
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.XLMProphetNetEncoder
+
+
+XLMProphetNetDecoder
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.XLMProphetNetDecoder
+
+
+XLMProphetNetForConditionalGeneration
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.XLMProphetNetForConditionalGeneration
+
+
+XLMProphetNetForCausalLM
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.XLMProphetNetForCausalLM

model_cards/microsoft/prophetnet-large-uncased-cnndm/README.md

+## prophetnet-large-uncased-cnndm
+Fine-tuned weights(converted from [original fairseq version repo](https://github.com/microsoft/ProphetNet)) for [ProphetNet](https://arxiv.org/abs/2001.04063) on summarization task CNN/DailyMail.  
+ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction.  
+ProphetNet is able to predict more future tokens with a n-stream decoder. The original implementation is Fairseq version at [github repo](https://github.com/microsoft/ProphetNet).   
+
+### Usage
+```
+from transformers import ProphetNetTokenizer, ProphetNetForConditionalGeneration, ProphetNetConfig
+
+model = ProphetNetForConditionalGeneration.from_pretrained('microsoft/prophetnet-large-uncased-cnndm')
+tokenizer = ProphetNetTokenizer.from_pretrained('microsoft/prophetnet-large-uncased-cnndm')
+
+ARTICLE_TO_SUMMARIZE = "USTC was founded in Beijing by the Chinese Academy of Sciences (CAS) in September 1958. The Director of CAS, Mr. Guo Moruo was appointed the first president of USTC. USTC's founding mission was to develop a high-level science and technology workforce, as deemed critical for development of China's economy, defense, and science and technology education. The establishment was hailed as \"A Major Event in the History of Chinese Education and Science.\" CAS has supported USTC by combining most of its institutes with the departments of the university. USTC is listed in the top 16 national key universities, becoming the youngest national key university.".lower()
+inputs = tokenizer([ARTICLE_TO_SUMMARIZE], max_length=100, return_tensors='pt')
+
+# Generate Summary
+summary_ids = model.generate(inputs['input_ids'], num_beams=4, max_length=512, early_stopping=True)
+tokenizer.batch_decode(summary_ids.tolist())
+```
+Here, [X_SEP] is used as a special token to seperate sentences.
+### Citation
+```bibtex
+@article{yan2020prophetnet,
+  title={Prophetnet: Predicting future n-gram for sequence-to-sequence pre-training},
+  author={Yan, Yu and Qi, Weizhen and Gong, Yeyun and Liu, Dayiheng and Duan, Nan and Chen, Jiusheng and Zhang, Ruofei and Zhou, Ming},
+  journal={arXiv preprint arXiv:2001.04063},
+  year={2020}
+}
+```

model_cards/microsoft/prophetnet-large-uncased/README.md

+## prophetnet-large-uncased
+Pretrained weights for [ProphetNet](https://arxiv.org/abs/2001.04063).  
+ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction.  
+ProphetNet is able to predict more future tokens with a n-stream decoder. The original implementation is Fairseq version at [github repo](https://github.com/microsoft/ProphetNet).   
+
+### Usage
+Please see [the official repository](https://github.com/microsoft/ProphetNet) for details.
+
+### Citation
+```bibtex
+@article{yan2020prophetnet,
+  title={Prophetnet: Predicting future n-gram for sequence-to-sequence pre-training},
+  author={Yan, Yu and Qi, Weizhen and Gong, Yeyun and Liu, Dayiheng and Duan, Nan and Chen, Jiusheng and Zhang, Ruofei and Zhou, Ming},
+  journal={arXiv preprint arXiv:2001.04063},
+  year={2020}
+}
+```

model_cards/microsoft/xprophetnet-large-wiki100-cased-xglue-ntg/README.md

+## xprophetnet-large-wiki100-cased-xglue-ntg
+Cross-lingual version [ProphetNet](https://arxiv.org/abs/2001.04063), pretrained on [wiki100 xGLUE dataset](https://arxiv.org/abs/2004.01401) and finetuned on xGLUE cross-lingual News Titles Generation task.  
+ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction.  
+ProphetNet is able to predict more future tokens with a n-stream decoder. The original implementation is Fairseq version at [github repo](https://github.com/microsoft/ProphetNet).   
+
+xProphetNet is also served as the baseline model for xGLUE cross-lingual natural language generation tasks.  
+For xGLUE corss-lingual NLG tasks, xProphetNet is finetuned with English data, but inference with both English and other zero-shot language data.  
+### Usage
+A quick usage is like: 
+```
+from transformers import ProphetNetTokenizer, ProphetNetForConditionalGeneration, ProphetNetConfig
+
+model = ProphetNetForConditionalGeneration.from_pretrained('microsoft/xprophetnet-large-wiki100-cased-xglue-ntg')
+tokenizer = ProphetNetTokenizer.from_pretrained('microsoft/xprophetnet-large-wiki100-cased-xglue-ntg')
+
+EN_SENTENCE = "Microsoft Corporation intends to officially end free support for the Windows 7 operating system after January 14, 2020, according to the official portal of the organization. From that day, users of this system will not be able to receive security updates, which could make their computers vulnerable to cyber attacks."
+RU_SENTENCE = "орпорация Microsoft намерена официально прекратить бесплатную поддержку операционной системы Windows 7 после 14 января 2020 года, сообщается на официальном портале организации . С указанного дня пользователи этой системы не смогут получать обновления безопасности, из-за чего их компьютеры могут стать уязвимыми к кибератакам."
+ZH_SENTENCE = "根据该组织的官方门户网站，微软公司打算在2020年1月14日之后正式终止对Windows 7操作系统的免费支持。从那时起，该系统的用户将无法接收安全更新，这可能会使他们的计算机容易受到网络攻击。"
+inputs = tokenizer([EN_SENTENCE, RU_SENTENCE, ZH_SENTENCE], padding=True, max_length=256, return_tensors='pt')
+
+summary_ids = model.generate(inputs['input_ids'], num_beams=4, max_length=100, early_stopping=True)
+print([tokenizer.decode(g) for g in summary_ids])  
+```
+### Citation
+```bibtex
+@article{yan2020prophetnet,
+  title={Prophetnet: Predicting future n-gram for sequence-to-sequence pre-training},
+  author={Yan, Yu and Qi, Weizhen and Gong, Yeyun and Liu, Dayiheng and Duan, Nan and Chen, Jiusheng and Zhang, Ruofei and Zhou, Ming},
+  journal={arXiv preprint arXiv:2001.04063},
+  year={2020}
+}
+```

model_cards/microsoft/xprophetnet-large-wiki100-cased-xglue-qg/README.md

+## xprophetnet-large-wiki100-cased-xglue-ntg
+Cross-lingual version [ProphetNet](https://arxiv.org/abs/2001.04063), pretrained on [wiki100 xGLUE dataset](https://arxiv.org/abs/2004.01401) and finetuned on xGLUE cross-lingual Question Generation task.  
+ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction.  
+ProphetNet is able to predict more future tokens with a n-stream decoder. The original implementation is Fairseq version at [github repo](https://github.com/microsoft/ProphetNet).   
+
+xProphetNet is also served as the baseline model for xGLUE cross-lingual natural language generation tasks.  
+For xGLUE corss-lingual NLG tasks, xProphetNet is finetuned with English data, but inference with both English and other zero-shot language data.  
+### Usage
+A quick usage is like: 
+```
+from transformers import ProphetNetTokenizer, ProphetNetForConditionalGeneration, ProphetNetConfig
+
+model = ProphetNetForConditionalGeneration.from_pretrained('microsoft/xprophetnet-large-wiki100-cased-xglue-qg')
+tokenizer = ProphetNetTokenizer.from_pretrained('microsoft/xprophetnet-large-wiki100-cased-xglue-qg')
+
+EN_SENTENCE = "Google left China in 2010"
+ZH_SENTENCE = "Google在2010年离开中国"
+inputs = tokenizer([EN_SENTENCE, ZH_SENTENCE], padding=True, max_length=256, return_tensors='pt')
+
+summary_ids = model.generate(inputs['input_ids'], num_beams=4, max_length=100, early_stopping=True)
+print([tokenizer.decode(g) for g in summary_ids])  
+```
+### Citation
+```bibtex
+@article{yan2020prophetnet,
+  title={Prophetnet: Predicting future n-gram for sequence-to-sequence pre-training},
+  author={Yan, Yu and Qi, Weizhen and Gong, Yeyun and Liu, Dayiheng and Duan, Nan and Chen, Jiusheng and Zhang, Ruofei and Zhou, Ming},
+  journal={arXiv preprint arXiv:2001.04063},
+  year={2020}
+}
+```

model_cards/microsoft/xprophetnet-large-wiki100-cased/README.md

+## xprophetnet-large-wiki100-cased
+Cross-lingual version [ProphetNet](https://arxiv.org/abs/2001.04063), pretrained on [wiki100 xGLUE dataset](https://arxiv.org/abs/2004.01401).  
+ProphetNet is a new pre-trained language model for sequence-to-sequence learning with a novel self-supervised objective called future n-gram prediction.  
+ProphetNet is able to predict more future tokens with a n-stream decoder. The original implementation is Fairseq version at [github repo](https://github.com/microsoft/ProphetNet).   
+
+xProphetNet is also served as the baseline model for xGLUE cross-lingual natural language generation tasks.  
+For xGLUE corss-lingual NLG tasks, xProphetNet is finetuned with English data, but inference with both English and other zero-shot language data.  
+### Usage
+Please see [the official repository](https://github.com/microsoft/ProphetNet/tree/master/xProphetNet) for details.
+
+### Citation
+```bibtex
+@article{yan2020prophetnet,
+  title={Prophetnet: Predicting future n-gram for sequence-to-sequence pre-training},
+  author={Yan, Yu and Qi, Weizhen and Gong, Yeyun and Liu, Dayiheng and Duan, Nan and Chen, Jiusheng and Zhang, Ruofei and Zhou, Ming},
+  journal={arXiv preprint arXiv:2001.04063},
+  year={2020}
+}
+```

src/transformers/__init__.py

@@ -54,6 +54,7 @@ from .configuration_mmbt import MMBTConfig
 from .configuration_mobilebert import MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MobileBertConfig
 from .configuration_openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig
 from .configuration_pegasus import PegasusConfig
+from .configuration_prophetnet import PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, ProphetNetConfig
 from .configuration_rag import RagConfig
 from .configuration_reformer import REFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, ReformerConfig
 from .configuration_retribert import RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, RetriBertConfig
@@ -63,6 +64,7 @@ from .configuration_t5 import T5_PRETRAINED_CONFIG_ARCHIVE_MAP, T5Config
 from .configuration_transfo_xl import TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP, TransfoXLConfig
 from .configuration_utils import PretrainedConfig
 from .configuration_xlm import XLM_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMConfig
+from .configuration_xlm_prophetnet import XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMProphetNetConfig
 from .configuration_xlm_roberta import XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMRobertaConfig
 from .configuration_xlnet import XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLNetConfig
 from .data import (
@@ -183,6 +185,7 @@ from .tokenization_lxmert import LxmertTokenizer
 from .tokenization_mobilebert import MobileBertTokenizer
 from .tokenization_openai import OpenAIGPTTokenizer
 from .tokenization_phobert import PhobertTokenizer
+from .tokenization_prophetnet import ProphetNetTokenizer
 from .tokenization_rag import RagTokenizer
 from .tokenization_retribert import RetriBertTokenizer
 from .tokenization_roberta import RobertaTokenizer
@@ -210,6 +213,7 @@ if is_sentencepiece_available():
     from .tokenization_pegasus import PegasusTokenizer
     from .tokenization_reformer import ReformerTokenizer
     from .tokenization_t5 import T5Tokenizer
+    from .tokenization_xlm_prophetnet import XLMProphetNetTokenizer
     from .tokenization_xlm_roberta import XLMRobertaTokenizer
     from .tokenization_xlnet import XLNetTokenizer
 else:
@@ -251,7 +255,6 @@ if is_tokenizers_available():
 else:
     from .utils.dummy_tokenizers_objects import *
 
-
 # Trainer
 from .trainer_callback import (
     DefaultFlowCallback,
@@ -490,6 +493,15 @@ if is_torch_available():
         load_tf_weights_in_openai_gpt,
     )
     from .modeling_pegasus import PegasusForConditionalGeneration
+    from .modeling_prophetnet import (
+        PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST,
+        ProphetNetDecoder,
+        ProphetNetEncoder,
+        ProphetNetForCausalLM,
+        ProphetNetForConditionalGeneration,
+        ProphetNetModel,
+        ProphetNetPreTrainedModel,
+    )
     from .modeling_rag import RagModel, RagSequenceForGeneration, RagTokenForGeneration
     from .modeling_reformer import (
         REFORMER_PRETRAINED_MODEL_ARCHIVE_LIST,
@@ -550,6 +562,14 @@ if is_torch_available():
         XLMPreTrainedModel,
         XLMWithLMHeadModel,
     )
+    from .modeling_xlm_prophetnet import (
+        XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST,
+        XLMProphetNetDecoder,
+        XLMProphetNetEncoder,
+        XLMProphetNetForCausalLM,
+        XLMProphetNetForConditionalGeneration,
+        XLMProphetNetModel,
+    )
     from .modeling_xlm_roberta import (
         XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST,
         XLMRobertaForCausalLM,

src/transformers/configuration_auto.py

@@ -41,6 +41,7 @@ from .configuration_mbart import MBART_PRETRAINED_CONFIG_ARCHIVE_MAP, MBartConfi
 from .configuration_mobilebert import MobileBertConfig
 from .configuration_openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig
 from .configuration_pegasus import PegasusConfig
+from .configuration_prophetnet import PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, ProphetNetConfig
 from .configuration_rag import RagConfig
 from .configuration_reformer import ReformerConfig
 from .configuration_retribert import RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, RetriBertConfig
@@ -50,6 +51,7 @@ from .configuration_t5 import T5_PRETRAINED_CONFIG_ARCHIVE_MAP, T5Config
 from .configuration_transfo_xl import TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP, TransfoXLConfig
 from .configuration_utils import PretrainedConfig
 from .configuration_xlm import XLM_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMConfig
+from .configuration_xlm_prophetnet import XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMProphetNetConfig
 from .configuration_xlm_roberta import XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, XLMRobertaConfig
 from .configuration_xlnet import XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP, XLNetConfig
 
@@ -84,6 +86,8 @@ ALL_PRETRAINED_CONFIG_ARCHIVE_MAP = dict(
         DPR_PRETRAINED_CONFIG_ARCHIVE_MAP,
         DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
         SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
+        PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ]
     for key, value, in pretrained_map.items()
 )
@@ -116,6 +120,8 @@ CONFIG_MAPPING = OrderedDict(
         ("gpt2", GPT2Config),
         ("transfo-xl", TransfoXLConfig),
         ("xlnet", XLNetConfig),
+        ("xlm-prophetnet", XLMProphetNetConfig),
+        ("prophetnet", ProphetNetConfig),
         ("xlm", XLMConfig),
         ("ctrl", CTRLConfig),
         ("electra", ElectraConfig),
@@ -164,6 +170,8 @@ MODEL_NAMES_MAPPING = OrderedDict(
         ("layoutlm", "LayoutLM"),
         ("dpr", "DPR"),
         ("rag", "RAG"),
+        ("xlm-prophetnet", "XLMProphetNet"),
+        ("prophetnet", "ProphetNet"),
     ]
 )
 

src/transformers/configuration_lxmert.py

@@ -15,12 +15,11 @@
 """ LXMERT model configuration """
 
 
-import logging
-
 from .configuration_utils import PretrainedConfig
+from .utils import logging
 
 
-logger = logging.getLogger(__name__)
+logger = logging.get_logger(__name__)
 
 LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP = {
     "unc-nlp/lxmert-base-uncased": "",

src/transformers/configuration_prophetnet.py

+# coding=utf-8
+# Copyright 2020 The Microsoft Authors and The HuggingFace Inc. team.
+#
+# 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.
+""" ProphetNet model configuration """
+
+
+from .configuration_utils import PretrainedConfig
+from .utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "microsoft/prophetnet-large-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/microsoft/prophetnet-large-uncased/config.json",
+}
+
+
+class ProphetNetConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a :class:`~transformers.ProphetNetModel`. It is used to
+    instantiate a ProphetNet model according to the specified arguments, defining the model 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:
+        activation_dropout (:obj:`float`, `optional`, defaults to 0.1):
+            The dropout ratio for activations inside the fully connected layer.
+        activation_function (:obj:`str` or :obj:`function`, `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:`"swish"` and :obj:`"gelu_new"` are supported.
+        vocab_size (:obj:`int`, `optional`, defaults to 30522):
+            Vocabulary size of the ProphetNET model. Defines the number of different tokens that can be represented by the
+            :obj:`inputs_ids` passed when calling :class:`~transformers.ProphetNetModel`.
+        hidden_size (:obj:`int`, `optional`, defaults to 1024):
+            Dimensionality of the layers and the pooler layer.
+        encoder_ffn_dim (:obj:`int`, `optional`, defaults to 4096):
+            Dimensionality of the "intermediate" (often named feed-forward) layer in decoder.
+        num_encoder_layers (:obj:`int`, `optional`, defaults to 12):
+            Number of encoder layers.
+        num_encoder_attention_heads (:obj:`int`, `optional`, defaults to 16):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        decoder_ffn_dim (:obj:`int`, `optional`, defaults to 4096):
+            Dimensionality of the ``intermediate`` (often named feed-forward) layer in decoder.
+        num_decoder_layers (:obj:`int`, `optional`, defaults to 12):
+            Number of decoder layers.
+        num_decoder_attention_heads (:obj:`int`, `optional`, defaults to 16):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        attention_dropout (:obj:`float`, `optional`, defaults to 0.1):
+            The dropout ratio for the attention probabilities.
+        dropout (:obj:`float`, `optional`, defaults to 0.1):
+            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
+        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).
+        init_std (:obj:`float`, `optional`, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        add_cross_attention (:obj:`bool`, `optional`, defaults to :obj:`True`):
+            Whether cross-attention layers should be added to the model.
+        is_encoder_decoder (:obj:`bool`, `optional`, defaults to :obj:`True`):
+            Whether this is an encoder/decoder model.
+        pad_token_id (:obj:`int`, `optional`, defaults to 1)
+            Padding token id.
+        bos_token_id (:obj:`int`, `optional`, defaults to 0)
+            Beginning of stream token id.
+        eos_token_id (:obj:`int`, `optional`, defaults to 2)
+            End of stream token id.
+        ngram (:obj:`int`, `optional`, defaults to 2)
+            Number of future tokens to predict.
+            Set to 1 to be same as traditional Language model to predict next first token.
+        num_buckets (:obj:`int`, `optional`, defaults to 32)
+            The number of buckets to use for each attention layer.
+            This is for relative position calculation. See the `T5 paper
+            <see https://arxiv.org/abs/1910.10683>`__ for more details.
+        relative_max_distance (:obj:`int`, `optional`, defaults to 128)
+            Relative distances greater than this number will be put into the last same bucket.
+            This is for relative position calculation. See the `T5 paper
+            <see https://arxiv.org/abs/1910.10683>`__ for more details.
+        disable_ngram_loss (:obj:`bool`, `optional`, defaults to :obj:`False`):
+            Whether be trained predicting only the next first token.
+        eps (:obj:`float`, `optional`, defaults to 0.0):
+            Controls the ``epsilon`` parameter value for label
+            smoothing in the loss calculation. If set to 0, no label smoothing is performed.
+    """
+    model_type = "prophetnet"
+
+    def __init__(
+        self,
+        activation_dropout=0.1,
+        activation_function="gelu",
+        vocab_size=30522,
+        hidden_size=1024,
+        encoder_ffn_dim=4096,
+        num_encoder_layers=12,
+        num_encoder_attention_heads=16,
+        decoder_ffn_dim=4096,
+        num_decoder_layers=12,
+        num_decoder_attention_heads=16,
+        attention_dropout=0.1,
+        dropout=0.1,
+        max_position_embeddings=512,
+        init_std=0.02,
+        is_encoder_decoder=True,
+        add_cross_attention=True,
+        pad_token_id=0,
+        bos_token_id=1,
+        eos_token_id=2,
+        decoder_start_token_id=0,
+        ngram=2,
+        num_buckets=32,
+        relative_max_distance=128,
+        disable_ngram_loss=False,
+        eps=0.0,
+        **kwargs
+    ):
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            is_encoder_decoder=is_encoder_decoder,
+            add_cross_attention=add_cross_attention,
+            decoder_start_token_id=decoder_start_token_id,
+            **kwargs,
+        )
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.encoder_ffn_dim = encoder_ffn_dim
+        self.num_encoder_layers = num_encoder_layers
+        self.num_encoder_attention_heads = num_encoder_attention_heads
+        self.decoder_ffn_dim = decoder_ffn_dim
+        self.num_decoder_layers = num_decoder_layers
+        self.num_decoder_attention_heads = num_decoder_attention_heads
+        self.max_position_embeddings = max_position_embeddings
+        self.init_std = init_std  # Normal(0, this parameter)
+        self.activation_function = activation_function
+
+        # parameters for prophetnet
+        self.ngram = ngram
+        self.num_buckets = num_buckets
+        self.relative_max_distance = relative_max_distance
+        self.disable_ngram_loss = disable_ngram_loss
+        self.eps = eps
+
+        # 3 Types of Dropout
+        self.attention_dropout = attention_dropout
+        self.activation_dropout = activation_dropout
+        self.dropout = dropout
+
+    @property
+    def num_attention_heads(self) -> int:
+        return self.num_encoder_attention_heads
+
+    @property
+    def num_hidden_layers(self) -> int:
+        return self.num_encoder_layers + self.num_decoder_layers

src/transformers/configuration_xlm_prophetnet.py

+# coding=utf-8
+# Copyright 2020 The Microsoft Authors and The HuggingFace Inc. team.
+#
+# 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.
+""" XLM-ProphetNet model configuration """
+
+
+from .configuration_prophetnet import ProphetNetConfig
+from .utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "microsoft/xprophetnet-large-wiki100-cased": "https://s3.amazonaws.com/models.huggingface.co/bert/microsoft/xprophetnet-large-wiki100-cased/config.json",
+}
+
+
+class XLMProphetNetConfig(ProphetNetConfig):
+    """
+    This class overrides :class:`~transformers.ProphetNetConfig`. Please check the
+    superclass for the appropriate documentation alongside usage examples.
+    """
+
+    model_type = "xlm-prophetnet"

src/transformers/convert_prophetnet_original_pytorch_checkpoint_to_pytorch.py

+# coding=utf-8
+# Copyright 2020 The HuggingFace Inc. team.
+#
+# 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.
+"""Convert ProphetNet checkpoint."""
+
+
+import argparse
+
+import torch
+
+from transformers import logging
+from transformers.modeling_prophetnet import ProphetNetForConditionalGeneration
+from transformers.modeling_xlm_prophetnet import XLMProphetNetForConditionalGeneration
+
+# transformers_old should correspond to branch `save_old_prophetnet_model_structure` here
+# original prophetnet_checkpoints are saved under `patrickvonplaten/..._old` respectively
+from transformers_old.modeling_prophetnet import (
+    ProphetNetForConditionalGeneration as ProphetNetForConditionalGenerationOld,
+)
+from transformers_old.modeling_xlm_prophetnet import (
+    XLMProphetNetForConditionalGeneration as XLMProphetNetForConditionalGenerationOld,
+)
+
+
+logger = logging.get_logger(__name__)
+logging.set_verbosity_info()
+
+
+def convert_prophetnet_checkpoint_to_pytorch(prophetnet_checkpoint_path: str, pytorch_dump_folder_path: str):
+    """
+    Copy/paste/tweak prohpetnet's weights to our prophetnet structure.
+    """
+    if "xprophetnet" in prophetnet_checkpoint_path:
+        prophet_old = XLMProphetNetForConditionalGenerationOld.from_pretrained(prophetnet_checkpoint_path)
+        prophet, loading_info = XLMProphetNetForConditionalGeneration.from_pretrained(
+            prophetnet_checkpoint_path, output_loading_info=True
+        )
+    else:
+        prophet_old = ProphetNetForConditionalGenerationOld.from_pretrained(prophetnet_checkpoint_path)
+        prophet, loading_info = ProphetNetForConditionalGeneration.from_pretrained(
+            prophetnet_checkpoint_path, output_loading_info=True
+        )
+
+    special_keys = ["key_proj", "value_proj", "query_proj"]
+
+    mapping = {
+        "self_attn": "ngram_self_attn",
+        "cross_attn": "encoder_attn",
+        "cross_attn_layer_norm": "encoder_attn_layer_norm",
+        "feed_forward_layer_norm": "final_layer_norm",
+        "feed_forward": "",
+        "intermediate": "fc1",
+        "output": "fc2",
+        "key_proj": "k_proj",
+        "query_proj": "q_proj",
+        "value_proj": "v_proj",
+        "word_embeddings": "embed_tokens",
+        "embeddings_layer_norm": "emb_layer_norm",
+        "relative_pos_embeddings": "relative_linear",
+        "ngram_embeddings": "ngram_input_embed",
+        "position_embeddings": "embed_positions",
+    }
+
+    for key in loading_info["missing_keys"]:
+        attributes = key.split(".")
+
+        if attributes[0] == "lm_head":
+            model = prophet
+            old_model = prophet_old
+        else:
+            model = prophet.prophetnet
+            old_model = prophet_old.model
+
+        is_key_init = False
+        for attribute in attributes:
+            if attribute in mapping:
+                old_attribute = mapping[attribute]
+                if not hasattr(old_model, old_attribute) and len(old_attribute) > 0:
+                    old_attribute = attribute
+            elif hasattr(old_model, attribute):
+                old_attribute = attribute
+
+            if attribute == "weight":
+                assert old_model.weight.shape == model.weight.shape, "Shapes have to match!"
+                model.weight = old_model.weight
+                logger.info(f"{attribute} is initialized.")
+                is_key_init = True
+                break
+            elif attribute == "bias":
+                assert old_model.bias.shape == model.bias.shape, "Shapes have to match!"
+                model.bias = old_model.bias
+                logger.info(f"{attribute} is initialized")
+                is_key_init = True
+                break
+            elif attribute in special_keys and hasattr(old_model, "in_proj_weight"):
+                embed_dim = old_model.in_proj_weight.shape[0] // 3
+                param = getattr(model, attribute)
+                param.weight.shape == old_model.in_proj_weight[:embed_dim, :].shape, "Shapes have to match"
+                param.bias.shape == old_model.in_proj_bias[:embed_dim].shape, "Shapes have to match"
+                if attribute == "query_proj":
+                    model.query_proj.weight = torch.nn.Parameter(old_model.in_proj_weight[:embed_dim, :])
+                    model.query_proj.bias = torch.nn.Parameter(old_model.in_proj_bias[:embed_dim])
+
+                elif attribute == "key_proj":
+                    model.key_proj.weight = torch.nn.Parameter(old_model.in_proj_weight[embed_dim : 2 * embed_dim, :])
+                    model.key_proj.bias = torch.nn.Parameter(old_model.in_proj_bias[embed_dim : 2 * embed_dim])
+                elif attribute == "value_proj":
+                    model.value_proj.weight = torch.nn.Parameter(old_model.in_proj_weight[2 * embed_dim :, :])
+                    model.value_proj.bias = torch.nn.Parameter(old_model.in_proj_bias[2 * embed_dim :])
+                is_key_init = True
+                break
+            elif attribute == "position_embeddings":
+                assert (
+                    model.position_embeddings.weight.shape[-1] == old_model.embed_positions.weight.shape[-1]
+                ), "Hidden size has to match"
+                assert model.position_embeddings.weight.shape[0] == 512, "We want 512 position_embeddings."
+                model.position_embeddings.weight = torch.nn.Parameter(old_model.embed_positions.weight[:512, :])
+                is_key_init = True
+                break
+
+            if attribute.isdigit():
+                model = model[int(attribute)]
+                old_model = old_model[int(old_attribute)]
+            else:
+                model = getattr(model, attribute)
+
+                if old_attribute == "":
+                    old_model = old_model
+                else:
+                    if not hasattr(old_model, old_attribute):
+                        raise ValueError(f"{old_model} does not have {old_attribute}")
+                    old_model = getattr(old_model, old_attribute)
+
+        if not is_key_init:
+            raise ValueError(f"{key} was not correctly initialized!")
+
+    print(f"Saving model to {pytorch_dump_folder_path}")
+    prophet.save_pretrained(pytorch_dump_folder_path)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    # Required parameters
+    parser.add_argument(
+        "--prophetnet_checkpoint_path", default=None, type=str, required=True, help="Path the official PyTorch dump."
+    )
+    parser.add_argument(
+        "--pytorch_dump_folder_path", default=None, type=str, required=True, help="Path to the output PyTorch model."
+    )
+    args = parser.parse_args()
+    convert_prophetnet_checkpoint_to_pytorch(args.prophetnet_checkpoint_path, args.pytorch_dump_folder_path)

src/transformers/modeling_auto.py

@@ -43,6 +43,7 @@ from .configuration_auto import (
     MobileBertConfig,
     OpenAIGPTConfig,
     PegasusConfig,
+    ProphetNetConfig,
     ReformerConfig,
     RetriBertConfig,
     RobertaConfig,
@@ -50,6 +51,7 @@ from .configuration_auto import (
     T5Config,
     TransfoXLConfig,
     XLMConfig,
+    XLMProphetNetConfig,
     XLMRobertaConfig,
     XLNetConfig,
     replace_list_option_in_docstrings,
@@ -155,6 +157,7 @@ from .modeling_mobilebert import (
 )
 from .modeling_openai import OpenAIGPTForSequenceClassification, OpenAIGPTLMHeadModel, OpenAIGPTModel
 from .modeling_pegasus import PegasusForConditionalGeneration
+from .modeling_prophetnet import ProphetNetForCausalLM, ProphetNetForConditionalGeneration, ProphetNetModel
 from .modeling_rag import (  # noqa: F401 - need to import all RagModels to be in globals() function
     RagModel,
     RagSequenceForGeneration,
@@ -194,6 +197,11 @@ from .modeling_xlm import (
     XLMModel,
     XLMWithLMHeadModel,
 )
+from .modeling_xlm_prophetnet import (
+    XLMProphetNetForCausalLM,
+    XLMProphetNetForConditionalGeneration,
+    XLMProphetNetModel,
+)
 from .modeling_xlm_roberta import (
     XLMRobertaForCausalLM,
     XLMRobertaForMaskedLM,
@@ -247,6 +255,8 @@ MODEL_MAPPING = OrderedDict(
         (BertGenerationConfig, BertGenerationEncoder),
         (DebertaConfig, DebertaModel),
         (DPRConfig, DPRQuestionEncoder),
+        (XLMProphetNetConfig, XLMProphetNetModel),
+        (ProphetNetConfig, ProphetNetModel),
     ]
 )
 
@@ -325,6 +335,8 @@ MODEL_FOR_CAUSAL_LM_MAPPING = OrderedDict(
         (CTRLConfig, CTRLLMHeadModel),
         (ReformerConfig, ReformerModelWithLMHead),
         (BertGenerationConfig, BertGenerationDecoder),
+        (ProphetNetConfig, XLMProphetNetForCausalLM),
+        (ProphetNetConfig, ProphetNetForCausalLM),
     ]
 )
 
@@ -359,6 +371,8 @@ MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING = OrderedDict(
         (BartConfig, BartForConditionalGeneration),
         (FSMTConfig, FSMTForConditionalGeneration),
         (EncoderDecoderConfig, EncoderDecoderModel),
+        (XLMProphetNetConfig, XLMProphetNetForConditionalGeneration),
+        (ProphetNetConfig, ProphetNetForConditionalGeneration),
     ]
 )
 

src/transformers/modeling_lxmert.py

@@ -15,7 +15,6 @@
 """ PyTorch LXMERT model. """
 
 
-import logging
 import math
 import os
 from dataclasses import dataclass
@@ -35,9 +34,10 @@ from .file_utils import (
     replace_return_docstrings,
 )
 from .modeling_utils import PreTrainedModel
+from .utils import logging
 
 
-logger = logging.getLogger(__name__)
+logger = logging.get_logger(__name__)
 
 _CONFIG_FOR_DOC = "LxmertConfig"
 _TOKENIZER_FOR_DOC = "LxmertTokenizer"

src/transformers/modeling_prophetnet.py

+# coding=utf-8
+# Copyright 2020 The Microsoft Authors and The HuggingFace Inc. team.
+#
+# 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.
+""" PyTorch ProphetNet model, ported from ProphetNet repo(fairsequery_states version). """
+
+import copy
+import math
+from dataclasses import dataclass
+from typing import Dict, Optional, Tuple
+
+import torch
+import torch.nn.functional as F
+from torch import Tensor, nn
+
+from .activations import ACT2FN
+from .configuration_prophetnet import ProphetNetConfig
+from .file_utils import ModelOutput, add_start_docstrings, add_start_docstrings_to_callable, replace_return_docstrings
+from .modeling_outputs import BaseModelOutput
+from .modeling_utils import PreTrainedModel
+from .utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+_CONFIG_FOR_DOC = "ProphenetConfig"
+_TOKENIZER_FOR_DOC = "ProphetNetTokenizer"
+
+PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST = [
+    "microsoft/prophetnet-large-uncased",
+    # See all ProphetNet models at https://huggingface.co/models?filter=prophetnet
+]
+
+
+PROPHETNET_START_DOCSTRING = r"""
+    This model inherits from :class:`~transformers.PreTrainedModel`. Check the superclass documentation for the generic
+    methods the library implements for all its model (such as downloading or saving, resizing the input embeddings,
+    pruning heads etc.)
+
+    Original ProphetNet code can be found at <https://github.com/microsoft/ProphetNet> .
+    Checkpoints were converted from original Fairseq checkpoints. For more information on
+    the checkpoint conversion, please take a look at the file ``convert_prophetnet_original_pytorch_checkpoint_to_pytorch.py``.
+
+    This model is a PyTorch `torch.nn.Module <https://pytorch.org/docs/stable/nn.html#torch.nn.Module>`_ sub-class. Use it as a regular PyTorch Module and
+    refer to the PyTorch documentation for all matters related to general usage and behavior.
+
+    Parameters:
+        config (:class:`~transformers.ProphetNetConfig`): Model configuration class with all the parameters of the model.
+            Initializing with a config file does not load the weights associated with the model, only the configuration.
+            Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
+"""
+
+PROPHETNET_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using :class:`~transformers.ProphetNetTokenizer`.
+            See :meth:`transformers.PreTrainedTokenizer.encode` and
+            :meth:`transformers.PreTrainedTokenizer.__call__` for details.
+
+            `What are input IDs? <../glossary.html#input-ids>`__
+        attention_mask (:obj:`torch.Tensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
+            Mask to avoid performing attention on padding token indices.
+            Mask values selected in ``[0, 1]``:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            `What are attention masks? <../glossary.html#attention-mask>`__
+        decoder_input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, target_sequence_length)`, `optional`):
+            Provide for translation and summarization training. By default, the model will create this tensor by
+            shifting the :obj:`input_ids` to the right, following the paper.
+        decoder_attention_mask (:obj:`torch.BoolTensor` of shape :obj:`(batch_size, tgt_seq_len)`, `optional`):
+            Default behavior: generate a tensor that ignores pad tokens in :obj:`decoder_input_ids`. Causal mask will
+            also be used by default.
+
+            If you want to change padding behavior, you should read :func:`modeling_bart._prepare_decoder_inputs` and
+            modify to your needs. See diagram 1 in `the paper <https://arxiv.org/abs/1910.13461>`__ for more
+            information on the default strategy.
+        encoder_outputs (:obj:`tuple(tuple(torch.FloatTensor)`, `optional`):
+            Tuple consists of (:obj:`last_hidden_state`, `optional`: :obj:`hidden_states`, `optional`: :obj:`attentions`)
+            :obj:`last_hidden_state` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`) is a
+            sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of
+            the decoder.
+        past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
+            Contains precomputed key and value hidden-states of the attention blocks. Can be used to speed up decoding.
+
+            If :obj:`past_key_values` are used, the user can optionally input only the last
+            ``decoder_input_ids`` (those that don't have their past key value states given to this model) of shape
+            :obj:`(batch_size, 1)` instead of all ``decoder_input_ids`` of shape :obj:`(batch_size, sequence_length)`.
+        use_cache (:obj:`bool`, `optional`):
+            If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up
+            decoding (see :obj:`past_key_values`).
+        output_attentions (:obj:`bool`, `optional`):
+            Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
+            tensors for more detail.
+        output_hidden_states (:obj:`bool`, `optional`):
+            Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
+            more detail.
+        return_dict (:obj:`bool`, `optional`):
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+"""
+
+PROPHETNET_STANDALONE_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using :class:`~transformers.ProphetNetTokenizer`.
+            See :meth:`transformers.PreTrainedTokenizer.encode` and
+            :meth:`transformers.PreTrainedTokenizer.__call__` for details.
+
+            `What are input IDs? <../glossary.html#input-ids>`__
+        attention_mask (:obj:`torch.Tensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
+            Mask to avoid performing attention on padding token indices.
+            Mask values selected in ``[0, 1]``:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            `What are attention masks? <../glossary.html#attention-mask>`__
+        output_attentions (:obj:`bool`, `optional`):
+            Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
+            tensors for more detail.
+        output_hidden_states (:obj:`bool`, `optional`):
+            Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
+            more detail.
+        return_dict (:obj:`bool`, `optional`):
+            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
+"""
+
+
+def softmax(hidden_state, dim, onnx_trace=False):
+    if onnx_trace:
+        return F.softmax(hidden_state.float(), dim=dim)
+    else:
+        return F.softmax(hidden_state, dim=dim, dtype=torch.float32)
+
+
+def ngram_attention_bias(sequence_length, ngram, device, dtype):
+    """
+    This function computes the bias for the predict stream
+    """
+    bias = torch.ones((ngram, sequence_length, 2 * sequence_length), device=device, dtype=dtype) * float("-inf")
+    # create bias
+    for stream_idx in range(ngram):
+        for i in range(sequence_length):
+            bias[stream_idx, i, sequence_length + i] = 0
+            bias[stream_idx, i, : max(i - stream_idx, 0) + 1] = 0
+    return bias
+
+
+def compute_relative_buckets(num_buckets, max_distance, relative_positions, is_bidirectional=False):
+    """
+    This function computes individual parts of the relative position buckets. For more detail, see paper.
+    """
+    inv_relative_positions = -relative_positions
+    rel_positions_bucket = 0
+
+    if is_bidirectional:
+        num_buckets = num_buckets // 2
+        rel_positions_bucket = (
+            rel_positions_bucket
+            + torch.lt(inv_relative_positions, torch.zeros_like(inv_relative_positions)).int() * num_buckets
+        )
+        inv_relative_positions = torch.abs(inv_relative_positions)
+    else:
+        inv_relative_positions = torch.max(inv_relative_positions, torch.zeros_like(inv_relative_positions))
+
+    max_exact = num_buckets // 2
+    is_small = torch.lt(inv_relative_positions, max_exact)
+    val_if_large = max_exact + torch.log(inv_relative_positions.float() / max_exact) / math.log(
+        max_distance / max_exact
+    ) * (num_buckets - max_exact)
+    val_if_large = torch.min(val_if_large, torch.ones_like(val_if_large) * (num_buckets - 1)).int()
+    rel_positions_bucket = rel_positions_bucket + torch.where(is_small, inv_relative_positions.int(), val_if_large)
+    return rel_positions_bucket
+
+
+def compute_all_stream_relative_buckets(num_buckets, max_distance, position_ids):
+    """
+    This function computes both main and predict relative position buckets. For more detail, see paper.
+    """
+    # main stream
+    main_stream_relative_positions = position_ids.unsqueeze(1).repeat(1, position_ids.size(-1), 1)
+    main_stream_relative_positions = main_stream_relative_positions - position_ids.unsqueeze(-1)
+
+    # predicting stream
+    predicting_stream_relative_positions = torch.cat((position_ids - 1, position_ids), dim=-1).unsqueeze(1)
+    predicting_stream_relative_positions = predicting_stream_relative_positions.repeat(1, position_ids.size(-1), 1)
+    predicting_stream_relative_positions = predicting_stream_relative_positions - position_ids.unsqueeze(-1)
+
+    # get both position buckets
+    main_relative_position_buckets = compute_relative_buckets(
+        num_buckets, max_distance, main_stream_relative_positions, is_bidirectional=False
+    )
+    predict_relative_position_buckets = compute_relative_buckets(
+        num_buckets, max_distance, predicting_stream_relative_positions, is_bidirectional=False
+    )
+    return main_relative_position_buckets, predict_relative_position_buckets
+
+
+@dataclass
+class ProphetNetSeq2SeqLMOutput(ModelOutput):
+    """
+    Base class for sequence-to-sequence language models outputs.
+
+    Args:
+        loss (:obj:`torch.FloatTensor` of shape :obj:`(1,)`, `optional`, returned when :obj:`labels` is provided):
+            Languaged modeling loss.
+        logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, decoder_sequence_length, config.vocab_size)`):
+            Prediction scores of the main stream language modeling head
+            (scores for each vocabulary token before SoftMax).
+        logits_ngram (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, ngram * decoder_sequence_length, config.vocab_size)`):
+            Prediction scores of the predict stream language modeling head (scores for
+            each vocabulary token before SoftMax).
+        past_key_values (:obj:`List[torch.FloatTensor]`, `optional`, returned when ``use_cache=True`` is passed or when ``config.use_cache=True``):
+            List of :obj:`torch.FloatTensor` of length :obj:`config.n_layers`,  with each tensor of shape
+            :obj:`(2, batch_size, num_attn_heads, decoder_sequence_length, embed_size_per_head)`).
+
+            Contains pre-computed hidden-states (key and values in the attention blocks)
+            of the decoder that can be
+            used (see :obj:`past_key_values` input) to speed up sequential decoding.
+        decoder_hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings
+            + one for the output of each layer)
+            of shape :obj:`(batch_size, decoder_sequence_length, hidden_size)`.
+
+            Hidden-states of main stream of the decoder at the output of each layer plus the initial embedding outputs.
+        decoder_ngram_hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, ngram * decoder_sequence_length, hidden_size)`.
+
+            Hidden-states of the predict stream of the decoder at the
+            output of each layer plus the initial embedding outputs.
+        decoder_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, decoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the decoder, after the attention softmax, used to compute the weighted average in the
+            self-attention heads.
+        decoder_ngram_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, decoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the predict stream of the decoder, after the attention softmax,
+            used to compute the weighted average in the self-attention heads.
+        decoder_cross_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, encoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the cross-attention layer of the decoder,
+            after the attention softmax, used to compute the weighted average in the
+        encoder_last_hidden_state (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, encoder_sequence_length, hidden_size)`, `optional`):
+            Sequence of hidden-states at the output of the last layer of the
+            encoder of the model.
+        encoder_hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, encoder_sequence_length, hidden_size)`.
+
+            Hidden-states of the encoder at the output of each layer plus the initial embedding outputs.
+        encoder_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, encoder_sequence_length, encoder_sequence_length)`.
+            Attentions weights of the encoder, after the attention softmax, used
+            to compute the weighted average in the
+            self-attention heads.
+    """
+
+    loss: Optional[torch.FloatTensor] = None
+    logits: torch.FloatTensor = None
+    logits_ngram: Optional[torch.FloatTensor] = None
+    past_key_values: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_ngram_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_ngram_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_cross_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    encoder_last_hidden_state: Optional[torch.FloatTensor] = None
+    encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    encoder_attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+@dataclass
+class ProphetNetSeq2SeqModelOutput(ModelOutput):
+    """
+    Base class for model encoder's outputs that also contains : pre-computed hidden states that can speed up sequential
+    decoding.
+
+    Args:
+        last_hidden_state (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, decoder_sequence_length, hidden_size)`):
+            Sequence of main stream hidden-states at the output of the last layer of the decoder of the model.
+
+            If :obj:`past_key_values` is used only the last hidden-state of
+            the sequences of shape :obj:`(batch_size, 1, hidden_size)` is output.
+        last_hidden_state_ngram (:obj:`torch.FloatTensor` of shape :obj:`(batch_size,ngram * decoder_sequence_length, config.vocab_size)`):
+            Sequence of predict stream hidden-states at the output of the last layer of the decoder of the model.
+        past_key_values (:obj:`List[torch.FloatTensor]`, `optional`, returned when ``use_cache=True`` is passed or when ``config.use_cache=True``):
+            List of :obj:`torch.FloatTensor` of length :obj:`config.n_layers`,  with each tensor of shape
+            :obj:`(2, batch_size, num_attn_heads, decoder_sequence_length, embed_size_per_head)`).
+
+            Contains pre-computed hidden-states (key and values in the attention blocks)
+            of the decoder that can be
+            used (see :obj:`past_key_values` input) to speed up sequential decoding.
+        decoder_hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, decoder_sequence_length, hidden_size)`.
+
+            Hidden-states of main stream of the decoder at the output of each layer plus the initial embedding outputs.
+        decoder_ngram_hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, ngram * decoder_sequence_length, hidden_size)`.
+
+            Hidden-states of the predict stream of the decoder at the
+            output of each layer plus the initial embedding outputs.
+        decoder_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, decoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the decoder, after the attention softmax, used to compute the weighted average in the
+            self-attention heads.
+        decoder_ngram_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, decoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the predict stream of the decoder,
+            after the attention softmax, used to compute the weighted average in the
+        decoder_cross_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, encoder_sequence_length,
+            decoder_sequence_length)`.
+
+            Attentions weights of the cross-attention layer of the decoder,
+            after the attention softmax, used to compute the weighted average in the
+        encoder_last_hidden_state (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, encoder_sequence_length, hidden_size)`, `optional`):
+            Sequence of hidden-states at the output of the last layer of the encoder of the model.
+        encoder_hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, encoder_sequence_length, hidden_size)`.
+
+            Hidden-states of the encoder at the output of each layer plus the initial embedding outputs.
+        encoder_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, encoder_sequence_length, encoder_sequence_length)`.
+
+            Attentions weights of the encoder, after the attention softmax, used to compute the weighted average in the
+            self-attention heads.
+    """
+
+    last_hidden_state: torch.FloatTensor
+    last_hidden_state_ngram: Optional[torch.FloatTensor] = None
+    past_key_values: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_ngram_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_ngram_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    decoder_cross_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    encoder_last_hidden_state: Optional[torch.FloatTensor] = None
+    encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    encoder_attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+@dataclass
+class ProphetNetDecoderModelOutput(ModelOutput):
+    """
+    Base class for model's outputs that may also contain a past key/values (to speed up sequential decoding).
+
+    Args:
+        last_hidden_state (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, decoder_sequence_length, hidden_size)`):
+            Sequence of main stream hidden-states at the output of the last layer of the decoder of the model.
+
+            If :obj:`past_key_values` is used only the last hidden-state
+            of the sequences of shape :obj:`(batch_size, 1, hidden_size)` is output.
+        last_hidden_state_ngram (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, ngram * decoder_sequence_length, config.vocab_size)`):
+            Sequence of predict stream hidden-states at the output of the last layer of the decoder of the model.
+        past_key_values (:obj:`List[torch.FloatTensor]`, `optional`, returned when ``use_cache=True`` is passed or when ``config.use_cache=True``):
+            List of :obj:`torch.FloatTensor` of length :obj:`config.n_layers`,  with each tensor of shape
+            :obj:`(2, batch_size, num_attn_heads, decoder_sequence_length, embed_size_per_head)`).
+
+            Contains pre-computed hidden-states (key and values in the attention blocks) of the decoder that can be
+            used (see :obj:`past_key_values` input) to speed up sequential decoding.
+        hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, decoder_sequence_length, hidden_size)`.
+
+            Hidden-states of main stream of the decoder at the output
+            of each layer plus the initial embedding outputs.
+        ngram_hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, ngram * decoder_sequence_length, hidden_size)`.
+
+            Hidden-states of the predict stream of the decoder at the
+            output of each layer plus the initial embedding outputs.
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, decoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the decoder, after the attention softmax,
+            used to compute the weighted average in the
+            self-attention heads.
+        ngram_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, decoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the predict stream of the decoder,
+            after the attention softmax, used to compute the weighted average in the
+        cross_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, encoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the cross-attention layer of the decoder, after the attention softmax, used to compute the weighted average in the
+    """
+
+    last_hidden_state: torch.FloatTensor
+    last_hidden_state_ngram: Optional[torch.FloatTensor] = None
+    past_key_values: Optional[Tuple[torch.FloatTensor]] = None
+    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    hidden_states_ngram: Optional[Tuple[torch.FloatTensor]] = None
+    attentions: Optional[Tuple[torch.FloatTensor]] = None
+    ngram_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    cross_attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+@dataclass
+class ProphetNetDecoderLMOutput(ModelOutput):
+    """
+    Base class for model's outputs that may also contain a past key/values (to speed up sequential decoding).
+
+    Args:
+        loss (:obj:`torch.FloatTensor` of shape :obj:`(1,)`, `optional`, returned when :obj:`labels` is provided):
+            Languaged modeling loss.
+        logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, decoder_sequence_length, config.vocab_size)`):
+            Prediction scores of the main stream language modeling head (scores for each vocabulary token before SoftMax).
+        logits_ngram (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, ngram * decoder_sequence_length, config.vocab_size)`):
+            Prediction scores of the predict stream language modeling head (scores for each vocabulary token before SoftMax).
+        past_key_values (:obj:`List[torch.FloatTensor]`, `optional`, returned when ``use_cache=True`` is passed or when ``config.use_cache=True``):
+            List of :obj:`torch.FloatTensor` of length :obj:`config.n_layers`,  with each tensor of shape
+            :obj:`(2, batch_size, num_attn_heads, decoder_sequence_length, embed_size_per_head)`).
+
+            Contains pre-computed hidden-states (key and values in the attention blocks) of the decoder that can be
+            used (see :obj:`past_key_values` input) to speed up sequential decoding.
+        hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, decoder_sequence_length, hidden_size)`.
+
+            Hidden-states of main stream of the decoder at the output of each layer plus the initial embedding outputs.
+        ngram_hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, ngram * decoder_sequence_length, hidden_size)`.
+
+            Hidden-states of the predict stream of the decoder at the output of each layer plus the initial embedding outputs.
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, decoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the decoder, after the attention softmax, used to compute the weighted average in the
+            self-attention heads.
+        ngram_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, decoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the predict stream of the decoder, after the attention softmax, used to compute the weighted average in the
+        cross_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_attn_heads, encoder_sequence_length, decoder_sequence_length)`.
+
+            Attentions weights of the cross-attention layer of the decoder, after the attention softmax, used to compute the weighted average in the
+    """
+
+    loss: Optional[torch.FloatTensor] = None
+    logits: torch.FloatTensor = None
+    logits_ngram: Optional[torch.FloatTensor] = None
+    past_key_values: Optional[Tuple[torch.FloatTensor]] = None
+    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    hidden_states_ngram: Optional[Tuple[torch.FloatTensor]] = None
+    attentions: Optional[Tuple[torch.FloatTensor]] = None
+    ngram_attentions: Optional[Tuple[torch.FloatTensor]] = None
+    cross_attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+def ProphetNetLayerNorm(normalized_shape, eps=1e-5, elementwise_affine=True):
+    if torch.cuda.is_available():
+        try:
+            from apex.normalization import FusedProphetNetLayerNorm
+
+            return FusedProphetNetLayerNorm(normalized_shape, eps, elementwise_affine)
+        except ImportError:
+            pass
+    return torch.nn.LayerNorm(normalized_shape, eps, elementwise_affine)
+
+
+class ProphetNetPreTrainedModel(PreTrainedModel):
+    config_class = ProphetNetConfig
+    base_model_prefix = "prophetnet"
+
+    def _init_weights(self, module):
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=self.config.init_std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=self.config.init_std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+
+    def _shift_right(self, input_ids):
+        decoder_start_token_id = self.config.decoder_start_token_id
+        pad_token_id = self.config.pad_token_id
+
+        assert (
+            decoder_start_token_id is not None
+        ), "self.model.config.decoder_start_token_id has to be defined. In ProphetNet it is usually set to the pad_token_id. See ProphetNet docs for more information"
+
+        # shift inputs to the right
+        shifted_input_ids = input_ids.new_zeros(input_ids.shape)
+        shifted_input_ids[..., 1:] = input_ids[..., :-1].clone()
+        shifted_input_ids[..., 0] = decoder_start_token_id
+
+        assert pad_token_id is not None, "self.model.config.pad_token_id has to be defined."
+        # replace possible -100 values in labels by `pad_token_id`
+        shifted_input_ids.masked_fill_(shifted_input_ids == -100, pad_token_id)
+
+        assert torch.all(shifted_input_ids >= 0).item(), "Verify that `shifted_input_ids` has only positive values"
+
+        return shifted_input_ids
+
+
+class ProhpetNetPositionalEmbeddings(nn.Embedding):
+    """
+    This module learns positional embeddings up to a fixed maximum size.
+    Padding ids are ignored by either offsetting based on padding_idx
+    or by setting padding_idx to None and ensuring that the appropriate
+    position ids are passed to the forward function.
+    """
+
+    def __init__(self, config: ProphetNetConfig):
+        super().__init__(config.max_position_embeddings, config.hidden_size, config.pad_token_id)
+
+    def forward(self, inputs_shape, device, attention_mask=None, past_key_values=None, position_ids=None):
+        assert (position_ids is None) or (
+            self.padding_idx is None
+        ), "If position_ids is pre-computed then padding_idx should not be set."
+
+        if position_ids is None:
+            if past_key_values is not None:
+                # position_ids is the same for every token when decoding a single step
+                # Without the int() cast, it doesn't work in some cases when exporting to ONNX
+                prev_num_input_ids = past_key_values[0]["self"]["prev_key_states"].shape[2]
+                num_input_ids = inputs_shape[1] + prev_num_input_ids
+                position_ids = torch.ones((1, 1), dtype=torch.long, device=device) * (
+                    int(self.padding_idx + num_input_ids)
+                )
+            else:
+                if attention_mask is None:
+                    attention_mask = torch.ones(inputs_shape, dtype=torch.long, device=device)
+
+                # retrieve position_ids from input_ids / attention_mask
+                position_ids = (
+                    torch.cumsum(attention_mask, dim=1).type_as(attention_mask) * attention_mask
+                ).long() + self.padding_idx
+
+        return super().forward(position_ids), position_ids
+
+    def _forward(self, position_ids):
+        return super().forward(position_ids)
+
+
+class ProphetNetSelfAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(
+        self,
+        config: ProphetNetConfig,
+        num_attn_heads: int,
+    ):
+        super().__init__()
+        hidden_size = config.hidden_size
+
+        self.attention_dropout = config.attention_dropout
+        self.dropout = config.dropout
+        self.num_attn_heads = num_attn_heads
+        self.head_dim = hidden_size // num_attn_heads
+
+        assert (
+            self.head_dim * num_attn_heads == hidden_size
+        ), "`config.hidden_size` must be divisible by `config.num_encoder_attention_heads` and `config.num_decoder_attention_heads`"
+
+        self.key_proj = nn.Linear(hidden_size, hidden_size)
+        self.value_proj = nn.Linear(hidden_size, hidden_size)
+        self.query_proj = nn.Linear(hidden_size, hidden_size)
+
+        self.out_proj = nn.Linear(hidden_size, hidden_size)
+
+    def _reshape(self, tensor, first_dim, batch_size):
+        return tensor.reshape(first_dim, batch_size * self.num_attn_heads, self.head_dim).transpose(0, 1)
+
+    def forward(
+        self,
+        hidden_states,
+        key_value_states: Optional[Tensor] = None,
+        attention_mask: Optional[Tensor] = None,
+        layer_state: Optional[Dict[str, Optional[Tensor]]] = None,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+
+        sequence_length, batch_size, hidden_size = hidden_states.size()
+
+        # if key_value_states are provided this layer is used as a cross-attention layer
+        # for the decoder
+        is_cross_attention = key_value_states is not None
+        cache_key = "cross_attention" if is_cross_attention else "self"
+        assert list(hidden_states.size()) == [
+            sequence_length,
+            batch_size,
+            hidden_size,
+        ], f"Size of hidden states should be {sequence_length, batch_size, hidden_size}, but is {hidden_states.size()}"
+
+        # previous time steps are cached - no need to recompute key and value if they are static
+        if layer_state is not None:
+            saved_state = layer_state.get(cache_key, None)
+
+        query_states = self.query_proj(hidden_states) / (self.head_dim ** 0.5)
+        query_states = self._reshape(query_states, sequence_length, batch_size)
+
+        if not is_cross_attention:
+            # self-attention
+            key_states = self.key_proj(hidden_states)
+            key_states = self._reshape(key_states, -1, batch_size)
+            value_states = self.value_proj(hidden_states)
+            value_states = self._reshape(value_states, -1, batch_size)
+        elif saved_state is None:
+            # cross-attention without layer state
+            key_states = self.key_proj(key_value_states)
+            key_states = self._reshape(key_states, -1, batch_size)
+            value_states = self.value_proj(key_value_states)
+            value_states = self._reshape(value_states, -1, batch_size)
+        else:
+            key_states = saved_state["prev_key_states"].view(batch_size * self.num_attn_heads, -1, self.head_dim)
+            value_states = saved_state["prev_value_states"].view(batch_size * self.num_attn_heads, -1, self.head_dim)
+
+        # Update cache
+        if is_cross_attention:
+            layer_state[cache_key] = {
+                "prev_key_states": key_states.view(batch_size, self.num_attn_heads, -1, self.head_dim),
+                "prev_value_states": value_states.view(batch_size, self.num_attn_heads, -1, self.head_dim),
+            }
+
+        key_sequence_length = key_states.size(1)
+        attn_weights = torch.bmm(query_states, key_states.transpose(1, 2))
+        assert attn_weights.size() == (
+            batch_size * self.num_attn_heads,
+            sequence_length,
+            key_sequence_length,
+        ), f"`attn_weights` should be of size {batch_size * self.num_attn_heads, sequence_length, key_sequence_length}, but is of size {attn_weights.shape}"
+
+        # This is part of a workaround to get around fork/join parallelism not supporting Optional types.
+        if attention_mask is not None and attention_mask.dim() == 0:
+            attention_mask = None
+        assert attention_mask is None or attention_mask.size() == (
+            self.num_attn_heads * batch_size,
+            1,
+            key_sequence_length,
+        ), f"`attention_mask` should be `None` or of shape attention_mask.size() == {batch_size * self.num_attn_heads, 1, key_sequence_length}, but is {attention_mask.shape}"
+
+        if attention_mask is not None:  # don't attend to padding symbols
+            attn_weights = attn_weights + attention_mask
+
+        attn_weights = F.softmax(attn_weights, dim=-1)
+        attn_probs = F.dropout(
+            attn_weights,
+            p=self.attention_dropout,
+            training=self.training,
+        )
+
+        attn_output = torch.bmm(attn_probs, value_states)
+        assert attn_output.size() == (
+            batch_size * self.num_attn_heads,
+            sequence_length,
+            self.head_dim,
+        ), "`attn_output` should be of shape {batch_size * self.num_attn_heads, sequence_length, self.head_dim}, but is of shape {attn_output.size()}"
+        attn_output = attn_output.transpose(0, 1).contiguous().view(sequence_length, batch_size, hidden_size)
+
+        attn_output = self.out_proj(attn_output)
+
+        attn_weights = attn_weights.view(batch_size, self.num_attn_heads, sequence_length, key_sequence_length)
+        attn_output = F.dropout(attn_output, p=self.dropout, training=self.training)
+        return attn_output, attn_weights
+
+
+class ProhpetNetFeedForward(nn.Module):
+    """
+    This is the residual two feed-forward layer block based on the original
+    Transformer implementation.
+    """
+
+    def __init__(self, config: ProphetNetConfig, ffn_dim: int):
+        super().__init__()
+        self.activation_fn = ACT2FN[config.activation_function]
+        self.intermediate = nn.Linear(config.hidden_size, ffn_dim)
+        self.output = nn.Linear(ffn_dim, config.hidden_size)
+        self.activation_dropout = config.activation_dropout
+        self.dropout = config.dropout
+
+    def forward(self, hidden_states):
+        hidden_states = self.intermediate(hidden_states)
+        hidden_states = self.activation_fn(hidden_states)
+
+        hidden_states = F.dropout(hidden_states, p=self.activation_dropout, training=self.training)
+        hidden_states = self.output(hidden_states)
+        hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+        return hidden_states
+
+
+class ProphetNetNgramProphetNetSelfAttention(nn.Module):
+    def __init__(self, config: ProphetNetConfig):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.num_buckets = config.num_buckets
+        self.relative_max_distance = config.relative_max_distance
+        self.num_attn_heads = config.num_attention_heads
+        self.dropout = config.dropout
+        self.attention_dropout = config.attention_dropout
+        self.head_dim = config.hidden_size // self.num_attn_heads
+        self.ngram = config.ngram
+
+        assert (
+            self.head_dim * self.num_attn_heads == config.hidden_size
+        ), "config.hidden_size must be divisible by num_attn_heads"
+        # key, value, query projection
+        self.key_proj = nn.Linear(config.hidden_size, config.hidden_size)
+        self.value_proj = nn.Linear(config.hidden_size, config.hidden_size)
+        self.query_proj = nn.Linear(config.hidden_size, config.hidden_size)
+
+        # out projection
+        self.out_proj = nn.Linear(config.hidden_size, config.hidden_size)
+
+        # rel position embeddings
+        self.relative_pos_embeddings = nn.Linear(config.hidden_size, self.num_buckets * self.num_attn_heads)
+
+        # for onnx runtime
+        self.onnx_trace = False
+
+    def _reshape(self, tensor, first_dim, batch_size):
+        return tensor.reshape(first_dim, batch_size * self.num_attn_heads, self.head_dim).transpose(0, 1)
+
+    def prepare_for_onnx_export_(self):
+        self.onnx_trace = True
+
+    def forward(
+        self,
+        hidden_states,
+        layer_state=None,
+        attention_mask=None,
+        extended_predict_attention_mask=None,
+        main_relative_position_buckets=None,
+        predict_relative_position_buckets=None,
+        position_ids=None,
+    ):
+        sequence_length, batch_size, hidden_size = hidden_states.size()
+
+        assert list(hidden_states.size()) == [
+            sequence_length,
+            batch_size,
+            hidden_size,
+        ], f"`hidden_states` should be of shape {sequence_length, batch_size, hidden_size}, but is of shape {hidden_states.shape}"
+
+        # key and value of previous time steps are cached
+        saved_state = layer_state.get("self", None)
+
+        # project
+        query_states = self.query_proj(hidden_states)
+        key_states = self.key_proj(hidden_states)
+        value_states = self.value_proj(hidden_states)
+
+        # normalize
+        query_states = query_states / (self.head_dim ** 0.5)
+
+        # reshape
+        query_states = self._reshape(query_states, sequence_length, batch_size)
+        key_states = self._reshape(key_states, -1, batch_size)
+        value_states = self._reshape(value_states, -1, batch_size)
+
+        # chunk into main stream and predict stream
+        hidden_states_list = hidden_states.chunk(1 + self.ngram, dim=0)
+
+        query_states_list = query_states.chunk(1 + self.ngram, dim=1)
+        key_states_list = key_states.chunk(1 + self.ngram, dim=1)
+        value_states_list = value_states.chunk(1 + self.ngram, dim=1)
+
+        main_hidden_states, hidden_states_predict_list = hidden_states_list[0], hidden_states_list[1:]
+        main_query_states, predict_query_states_list = query_states_list[0], query_states_list[1:]
+        main_key_states, predict_key_states_list = key_states_list[0], key_states_list[1:]
+        main_value_states, predict_value_states_list = value_states_list[0], value_states_list[1:]
+
+        # saved states are stored with shape (batch_size, num_attn_heads, seq_len, head_dim)
+        if saved_state is not None:
+            prev_main_key_states = saved_state["prev_key_states"].view(
+                batch_size * self.num_attn_heads, -1, self.head_dim
+            )
+            main_key_states = torch.cat((prev_main_key_states, main_key_states), dim=1)
+            prev_main_value_states = saved_state["prev_value_states"].view(
+                batch_size * self.num_attn_heads, -1, self.head_dim
+            )
+            main_value_states = torch.cat((prev_main_value_states, main_value_states), dim=1)
+
+        # Update cache
+        layer_state["self"] = {
+            "prev_key_states": main_key_states.view(batch_size, self.num_attn_heads, -1, self.head_dim),
+            "prev_value_states": main_value_states.view(batch_size, self.num_attn_heads, -1, self.head_dim),
+        }
+
+        # get seq_length of main stream only
+        main_sequence_length = sequence_length // (1 + self.ngram)
+
+        # MAIN-STREAM
+        # main attn weights
+        main_attn_weights = torch.bmm(main_query_states, main_key_states.transpose(1, 2))
+
+        # retrieve relative position embeddings for each layer -> see paper for more details
+        main_relative_pos_embeddings = self.get_main_relative_pos_embeddings(
+            main_hidden_states, main_attn_weights, position_ids, main_relative_position_buckets
+        )
+        main_attn_weights = main_attn_weights + main_relative_pos_embeddings
+
+        if attention_mask is not None:
+            main_attn_weights = main_attn_weights + attention_mask
+
+        main_attn_probs = softmax(
+            main_attn_weights,
+            dim=-1,
+            onnx_trace=self.onnx_trace,
+        ).type_as(main_attn_weights)
+
+        main_attn_probs = F.dropout(main_attn_probs, p=self.attention_dropout, training=self.training)
+
+        # project to attn_output
+        main_attn_output = torch.bmm(main_attn_probs, main_value_states)
+        main_attn_output = (
+            main_attn_output.transpose(0, 1).contiguous().view(1, main_sequence_length, batch_size, hidden_size)
+        )
+        main_attn_output = self.out_proj(main_attn_output)
+
+        # PREDICT-STREAM
+        # [ngram, B*head, T, c]
+        predict_query_states = torch.cat(predict_query_states_list, 0).view(
+            self.ngram, -1, main_sequence_length, self.head_dim
+        )
+        # [ngram, B*head, 2*T, c]
+        predict_key_states = torch.cat(
+            [torch.cat([main_key_states, key], 1).unsqueeze(0) for key in predict_key_states_list], 0
+        )
+
+        # [ngram, T, B, C]
+        predict_hidden_states = torch.cat(hidden_states_predict_list, 0).view(
+            self.ngram, main_sequence_length, batch_size, hidden_size
+        )
+
+        # [ngram, B*head, 2*T, c]
+        predict_value_states = torch.cat(
+            [torch.cat([main_value_states, v_p], 1).unsqueeze(0) for v_p in predict_value_states_list], 0
+        )
+        # [ngram, B*head, T, 2*T]
+        predict_attn_weights = torch.einsum("nbtc,nbsc->nbts", (predict_query_states, predict_key_states))
+
+        # [ngram, B*head, T, S]
+        # retrieve relative position embeddings for each layer -> see paper for more details
+        predict_relative_pos_embeddings = self.get_predict_relative_pos_embeddings(
+            predict_hidden_states, predict_attn_weights, position_ids, predict_relative_position_buckets
+        )
+
+        # [ngram, B*head, T, 2*T]
+        predict_attn_weights = predict_attn_weights + predict_relative_pos_embeddings
+
+        if extended_predict_attention_mask is not None:
+            predict_attn_weights = predict_attn_weights + extended_predict_attention_mask
+
+        predict_attn_probs = softmax(
+            predict_attn_weights,
+            dim=-1,
+            onnx_trace=self.onnx_trace,
+        ).type_as(predict_attn_weights)
+        predict_attn_probs = F.dropout(predict_attn_probs, p=self.attention_dropout, training=self.training)
+
+        # project to attention output
+        # [ngram, B*head, T, c]
+        predict_attn_output = torch.einsum("nbts,nbsc->nbtc", (predict_attn_probs, predict_value_states))
+        # [ngram, T, B, C]
+        predict_attn_output = (
+            predict_attn_output.transpose(1, 2)
+            .contiguous()
+            .view(self.ngram, main_sequence_length, batch_size, hidden_size)
+        )
+        predict_attn_output = self.out_proj(predict_attn_output)
+
+        # concat to single attn output
+        # [1+ngram*T, B, C]
+        attn_output = torch.cat([main_attn_output, predict_attn_output], 0).view(-1, batch_size, hidden_size)
+
+        # reshape into better form for `config.output_attentions`
+        main_attn_probs = main_attn_probs.view(batch_size, self.num_attn_heads, main_sequence_length, -1)
+        predict_attn_probs = predict_attn_probs.view(
+            self.ngram, batch_size, self.num_attn_heads, main_sequence_length, -1
+        ).transpose(0, 1)
+
+        attn_output = F.dropout(attn_output, p=self.dropout, training=self.training)
+        return attn_output, main_attn_probs, predict_attn_probs
+
+    def get_main_relative_pos_embeddings(
+        self, hidden_states, attn_weights, position_ids, main_relative_position_buckets
+    ):
+        # input hidden_states [T,B,C], input attn_weights [T*head,T,S], input position_ids [B,T] or [1,1]
+
+        if main_relative_position_buckets is None:
+            batch_size, sequence_length = hidden_states.shape[:2]
+            relative_positions = (
+                torch.arange(1, attn_weights.shape[-1] + 1)
+                .unsqueeze(0)
+                .unsqueeze(0)
+                .repeat(batch_size, sequence_length, 1)
+                .to(position_ids.device)
+            )
+            relative_positions = relative_positions - position_ids.unsqueeze(0).repeat(
+                batch_size, sequence_length, 1
+            )  # [B, T, s]
+            main_relative_position_buckets = compute_relative_buckets(
+                self.num_buckets, self.relative_max_distance, relative_positions, False
+            )
+
+        hidden_states = hidden_states.transpose(0, 1)  # [B,T,C]
+        rel_pos_embeddings = self.relative_pos_embeddings(hidden_states)  # [B,T,Buckets*head]
+        rel_pos_embeddings = rel_pos_embeddings.view(
+            rel_pos_embeddings.shape[:2] + (self.num_buckets, self.num_attn_heads)
+        ).permute(
+            0, 3, 1, 2
+        )  # [B,T,Buckets,head]
+        rel_pos_embeddings = rel_pos_embeddings.reshape(attn_weights.shape[:2] + (-1,))  # [B*head,T,Buckets]
+
+        main_relative_position_buckets = (
+            main_relative_position_buckets.repeat(1, self.num_attn_heads, 1)
+            .view(-1, main_relative_position_buckets.shape[-1])
+            .long()
+        )  # [B*head*T, T]
+        rel_pos_embeddings = rel_pos_embeddings.reshape(-1, rel_pos_embeddings.size(-1))  # [B*head*T,Buckets]
+
+        main_relative_pos_embeddings = torch.gather(
+            rel_pos_embeddings, dim=1, index=main_relative_position_buckets
+        ).view(attn_weights.shape[:2] + (-1,))
+
+        return main_relative_pos_embeddings
+
+    def get_predict_relative_pos_embeddings(
+        self, hidden_states, attn_weights, position_ids, predict_relative_position_buckets
+    ):
+        # input hidden_states [ngram, T,B,C], input attn_weights [ngram, B*head,T,S], input position_ids [B,T] or [1,1], input predict_relative_position_buckets [B,T, 2*T] or None
+
+        sequence_length, batch_size = hidden_states.shape[1:3]
+
+        if predict_relative_position_buckets is None:
+            key_sequence_length = attn_weights.shape[-1]
+            assert (
+                position_ids[0][0] == key_sequence_length - 1
+            ), "`position_ids` are incorrect. They should be of the format 1 2 3 4 5 ... (key_sequence_length - 1)"
+            relative_positions = (
+                torch.arange(0, key_sequence_length)
+                .unsqueeze(0)
+                .unsqueeze(0)
+                .repeat(batch_size, sequence_length, 1)
+                .to(position_ids.device)
+            )
+
+            relative_positions = relative_positions - position_ids.unsqueeze(0).repeat(batch_size, sequence_length, 1)
+            predict_relative_position_buckets = compute_relative_buckets(
+                self.num_buckets, self.relative_max_distance, relative_positions, False
+            )
+
+        hidden_states = hidden_states.transpose(1, 2)  # [ngram, B, T, C]
+        rel_pos_embeddings = self.relative_pos_embeddings(hidden_states).view(
+            hidden_states.shape[:-1] + (self.num_buckets, self.num_attn_heads)
+        )  # [ngram, B, T, bucket, head]
+        rel_pos_embeddings = rel_pos_embeddings.permute(0, 1, 4, 2, 3).reshape(
+            self.ngram * batch_size * self.num_attn_heads, sequence_length, -1
+        )  # [ngram*B*head, T, bucket]
+
+        predict_relative_position_buckets = predict_relative_position_buckets.unsqueeze(0).repeat(
+            self.ngram, 1, self.num_attn_heads, 1
+        )  # [ngram, B, head*T, S]
+
+        rel_pos_embeddings = rel_pos_embeddings.reshape(-1, rel_pos_embeddings.size(-1))
+        predict_relative_position_buckets = predict_relative_position_buckets.view(
+            -1, predict_relative_position_buckets.size(-1)
+        ).long()  # [ngram*B*head*T, S]
+
+        predict_relative_pos_embeddings = torch.gather(
+            rel_pos_embeddings, dim=1, index=predict_relative_position_buckets
+        ).view(
+            self.ngram, batch_size * self.num_attn_heads, sequence_length, -1
+        )  # [ngram, B*head, T, S]
+
+        return predict_relative_pos_embeddings
+
+
+class ProphetNetEncoderLayer(nn.Module):
+    """
+    Encoder block for Prophetnet
+    """
+
+    def __init__(self, config: ProphetNetConfig):
+        super().__init__()
+        # 1st residual block
+        self.self_attn = ProphetNetSelfAttention(config, config.num_encoder_attention_heads)
+        self.self_attn_layer_norm = ProphetNetLayerNorm(config.hidden_size)
+
+        # 2nd residual block
+        self.feed_forward = ProhpetNetFeedForward(config, config.encoder_ffn_dim)
+        self.feed_forward_layer_norm = ProphetNetLayerNorm(config.hidden_size)
+
+    def forward(self, hidden_states, attention_mask):
+        # 1st residual block
+        attention_output, attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+        )
+        hidden_states = self.self_attn_layer_norm(attention_output + hidden_states)
+
+        # 2nd residual block
+        feed_forward_output = self.feed_forward(hidden_states)
+        hidden_states = self.feed_forward_layer_norm(feed_forward_output + hidden_states)
+        return hidden_states, attn_weights
+
+
+class ProphetNetDecoderLayer(nn.Module):
+    """
+    Decoder block for Prophetnet
+    """
+
+    def __init__(self, config: ProphetNetConfig):
+        super().__init__()
+        # 1st residual block
+        self.self_attn = ProphetNetNgramProphetNetSelfAttention(config)
+        self.self_attn_layer_norm = ProphetNetLayerNorm(config.hidden_size)
+
+        # 2nd residual block
+        if config.add_cross_attention:
+            self.cross_attn = ProphetNetSelfAttention(config, config.num_decoder_attention_heads)
+            self.cross_attn_layer_norm = ProphetNetLayerNorm(config.hidden_size)
+
+        # 3rd residual block
+        self.feed_forward = ProhpetNetFeedForward(config, config.decoder_ffn_dim)
+        self.feed_forward_layer_norm = ProphetNetLayerNorm(config.hidden_size)
+
+    def forward(
+        self,
+        hidden_states,
+        encoder_hidden_states=None,
+        encoder_attn_mask=None,
+        layer_state=None,
+        attention_mask=None,
+        extended_predict_attention_mask=None,
+        main_relative_position_buckets=None,
+        predict_relative_position_buckets=None,
+        position_ids=None,
+    ):
+        layer_state = layer_state if layer_state is not None else {}
+
+        # 1st residual block
+        ngram_attention_output, self_attn_weights, self_attn_weights_ngram = self.self_attn(
+            hidden_states=hidden_states,
+            layer_state=layer_state,
+            attention_mask=attention_mask,
+            extended_predict_attention_mask=extended_predict_attention_mask,
+            main_relative_position_buckets=main_relative_position_buckets,
+            predict_relative_position_buckets=predict_relative_position_buckets,
+            position_ids=position_ids,
+        )
+        hidden_states = self.self_attn_layer_norm(hidden_states + ngram_attention_output)
+
+        cross_attn_weights = None
+        if encoder_hidden_states is not None:
+            # 2nd residual block
+            attention_output, cross_attn_weights = self.cross_attn(
+                hidden_states=hidden_states,
+                key_value_states=encoder_hidden_states,
+                attention_mask=encoder_attn_mask,
+                layer_state=layer_state,  # mutates layer state
+            )
+            hidden_states = self.cross_attn_layer_norm(attention_output + hidden_states)
+
+        # 3rd residual block
+        feed_forward_output = self.feed_forward(hidden_states)
+        hidden_states = self.feed_forward_layer_norm(feed_forward_output + hidden_states)
+
+        return (
+            hidden_states,
+            self_attn_weights,
+            self_attn_weights_ngram,
+            cross_attn_weights,
+            layer_state,
+        )  # just self_attn weights for now, following t5, layer_state = cache for decoding
+
+
+@add_start_docstrings(
+    "The standalone encoder part of the ProphetNetModel.",
+    PROPHETNET_START_DOCSTRING,
+)
+class ProphetNetEncoder(ProphetNetPreTrainedModel):
+    r"""
+    word_embeddings  (:obj:`torch.nn.Embeddings` of shape :obj:`(config.vocab_size, config.hidden_size)`, `optional`):
+        The word embedding parameters. This can be used to initialize :class:`~transformers.ProphetNetEncoder` with pre-defined
+        word embeddings instead of randomely initialized word embeddings.
+    """
+
+    def __init__(self, config: ProphetNetConfig, word_embeddings: nn.Embedding = None):
+        super().__init__(config)
+
+        self.word_embeddings = (
+            word_embeddings
+            if word_embeddings is not None
+            else nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
+        )
+        self.position_embeddings = ProhpetNetPositionalEmbeddings(config)
+        self.embeddings_layer_norm = ProphetNetLayerNorm(config.hidden_size)
+
+        self.layers = nn.ModuleList([ProphetNetEncoderLayer(config) for _ in range(config.num_encoder_layers)])
+
+        self.init_weights()
+
+    def get_input_embeddings(self):
+        return self.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.word_embeddings = value
+
+    @add_start_docstrings_to_callable(PROPHETNET_STANDALONE_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=BaseModelOutput, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        inputs_embeds=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+    ):
+        r"""
+        Returns:
+
+        Example::
+
+            >>> from transformers import ProphetNetTokenizer, ProphetNetEncoder
+            >>> import torch
+
+            >>> tokenizer = ProphetNetTokenizer.from_pretrained('microsoft/prophetnet-large-uncased')
+            >>> model = ProphetNetEncoder.from_pretrained('patrickvonplaten/prophetnet-large-uncased-standalone', return_dict=True)
+            >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
+            >>> outputs = model(**inputs)
+
+            >>> last_hidden_states = outputs.last_hidden_state
+        """
+
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if input_ids is None and inputs_embeds is None:
+            raise ValueError("Either input_ids or inputs_embeds has to be passed.")
+        elif input_ids is not None and inputs_embeds is not None:
+            raise ValueError("Make sure to only pass input_ids or inputs_embeds.")
+        elif input_ids is not None and inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+
+        # prepare attention mask
+        if attention_mask is not None:
+            extended_attention_mask = (
+                1.0 - attention_mask[:, None, :].repeat(self.config.num_attention_heads, 1, 1)
+            ) * -10000.0
+            extended_attention_mask = extended_attention_mask.to(inputs_embeds.dtype)
+        else:
+            extended_attention_mask = None
+
+        position_embeddings, position_ids = self.position_embeddings(inputs_embeds.shape[:2], inputs_embeds.device)
+
+        hidden_states = inputs_embeds + position_embeddings
+        hidden_states = self.embeddings_layer_norm(hidden_states)
+        hidden_states = F.dropout(hidden_states, p=self.config.dropout, training=self.training)
+        hidden_states = hidden_states.transpose(0, 1)  # B x T x C -> T x B x C
+
+        encoder_hidden_states = () if output_hidden_states else None
+        all_attentions = () if output_attentions else None
+
+        for encoder_layer in self.layers:
+            if output_hidden_states:
+                encoder_hidden_states = encoder_hidden_states + (hidden_states.transpose(0, 1),)
+            hidden_states, attn_probs = encoder_layer(hidden_states, attention_mask=extended_attention_mask)
+            if output_attentions:
+                all_attentions = all_attentions + (attn_probs,)
+
+        hidden_states = hidden_states.transpose(0, 1)
+        if output_hidden_states:
+            encoder_hidden_states = encoder_hidden_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(v for v in [hidden_states, encoder_hidden_states, all_attentions] if v is not None)
+        return BaseModelOutput(
+            last_hidden_state=hidden_states, hidden_states=encoder_hidden_states, attentions=all_attentions
+        )
+
+
+@add_start_docstrings(
+    "The standalone decoder part of the ProphetNetModel.",
+    PROPHETNET_START_DOCSTRING,
+)
+class ProphetNetDecoder(ProphetNetPreTrainedModel):
+    r"""
+    word_embeddings  (:obj:`torch.nn.Embeddings` of shape :obj:`(config.vocab_size, config.hidden_size)`, `optional`):
+        The word embedding parameters. This can be used to initialize :class:`~transformers.ProphetNetEncoder` with pre-defined
+        word embeddings instead of randomely initialized word embeddings.
+    """
+
+    def __init__(self, config: ProphetNetConfig, word_embeddings: nn.Embedding = None):
+        super().__init__(config)
+
+        self.ngram = config.ngram
+        self.num_buckets = config.num_buckets
+        self.relative_max_distance = config.relative_max_distance
+        self.dropout = config.dropout
+        self.max_target_positions = config.max_position_embeddings
+
+        self.word_embeddings = (
+            word_embeddings
+            if word_embeddings is not None
+            else nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
+        )
+        self.position_embeddings = ProhpetNetPositionalEmbeddings(config)
+
+        self.ngram_embeddings = nn.Embedding(self.ngram, config.hidden_size, None)
+        self.layers = nn.ModuleList([ProphetNetDecoderLayer(config) for _ in range(config.num_decoder_layers)])
+        self.embeddings_layer_norm = ProphetNetLayerNorm(config.hidden_size)
+
+        self.init_weights()
+
+    def get_input_embeddings(self):
+        return self.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.word_embeddings = value
+
+    @add_start_docstrings_to_callable(PROPHETNET_STANDALONE_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=ProphetNetDecoderModelOutput, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_values=None,
+        inputs_embeds=None,
+        use_cache=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+    ):
+        r"""
+        encoder_hidden_states  (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`):
+            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention
+            if the model is configured as a decoder.
+        encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
+            Mask to avoid performing attention on the padding token indices of the encoder input. This mask
+            is used in the cross-attention if the model is configured as a decoder.
+            Mask values selected in ``[0, 1]``:
+        past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
+            Contains precomputed key and value hidden-states of the attention blocks. Can be used to speed up decoding.
+
+            If :obj:`past_key_values` are used, the user can optionally input only the last
+            ``decoder_input_ids`` (those that don't have their past key value states given to this model) of shape
+            :obj:`(batch_size, 1)` instead of all ``decoder_input_ids`` of shape :obj:`(batch_size, sequence_length)`.
+        use_cache (:obj:`bool`, `optional`):
+            If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up
+            decoding (see :obj:`past_key_values`).
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+        Returns:
+
+        Example::
+
+            >>> from transformers import ProphetNetTokenizer, ProphetNetDecoder
+            >>> import torch
+
+            >>> tokenizer = ProphetNetTokenizer.from_pretrained('microsoft/prophetnet-large-uncased')
+            >>> model = ProphetNetDecoder.from_pretrained('patrickvonplaten/prophetnet-large-uncased-standalone', add_cross_attention=False, return_dict=True)
+            >>> assert model.config.is_decoder, f"{model.__class__} has to be configured as a decoder."
+            >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
+            >>> outputs = model(**inputs)
+
+            >>> last_hidden_states = outputs.last_hidden_state
+        """
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if input_ids is None and inputs_embeds is None:
+            raise ValueError("Either `decoder_input_ids` or `decoder_inputs_embeds` has to be passed.")
+        elif input_ids is not None and inputs_embeds is not None:
+            raise ValueError("Make sure to only pass `decoder_input_ids` or `decoder_inputs_embeds`.")
+        elif input_ids is not None and inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+
+        batch_size, sequence_length = inputs_embeds.shape[:2]
+
+        main_stream_pos_embed, position_ids = self.position_embeddings(
+            (batch_size, sequence_length),
+            device=inputs_embeds.device,
+            past_key_values=past_key_values,
+        )
+
+        if past_key_values is not None:
+            main_relative_position_buckets, predict_relative_position_buckets = None, None
+        else:
+            (
+                main_relative_position_buckets,
+                predict_relative_position_buckets,
+            ) = self.compute_buffered_relative_buckets(position_ids)
+        predicting_stream_pos_embed = self.position_embeddings._forward(position_ids + 1)
+
+        # add position embeddings
+        hidden_states = inputs_embeds + main_stream_pos_embed
+        hidden_states = hidden_states.transpose(0, 1)
+
+        ngram_embeddings = self.ngram_embeddings.weight
+
+        # prepare attention mask
+        if past_key_values is not None:
+            assert (
+                hidden_states.size(0) == 1
+            ), "At the moment `use_cache` is only supported for `decoder_input_ids` of length 1"
+
+            ngram_hidden_states = [
+                (ngram_embeddings[ngram - 1] + predicting_stream_pos_embed).transpose(0, 1).repeat(1, batch_size, 1)
+                for ngram in range(self.ngram)
+            ]
+            extended_attention_mask = None
+            extended_predict_attention_mask = None
+        else:
+            ngram_hidden_states = [
+                (ngram_embeddings[ngram - 1] + predicting_stream_pos_embed).transpose(0, 1)
+                for ngram in range(self.ngram)
+            ]
+            extended_attention_mask = self.prepare_attention_mask(hidden_states, attention_mask)
+            extended_predict_attention_mask = self.prepare_predict_attention_mask(hidden_states, attention_mask)
+
+        # prepare encoder attention mask
+        if encoder_attention_mask is not None:
+            extended_encoder_attention_mask = (
+                1.0 - encoder_attention_mask[:, None, :].repeat(self.config.num_attention_heads, 1, 1)
+            ) * -10000.0
+            extended_encoder_attention_mask = extended_encoder_attention_mask.to(inputs_embeds.dtype)
+        else:
+            extended_encoder_attention_mask = None
+
+        hidden_states = torch.cat([hidden_states] + ngram_hidden_states, 0)
+
+        if self.embeddings_layer_norm:
+            hidden_states = self.embeddings_layer_norm(hidden_states)
+
+        hidden_states = F.dropout(hidden_states, p=self.dropout, training=self.training)
+
+        if encoder_hidden_states is not None:
+            encoder_hidden_states = encoder_hidden_states.transpose(0, 1)
+
+        # init attentions, hidden_states and cache with empty tuples
+        all_main_stream_hidden_states = () if output_hidden_states else None
+        all_ngram_stream_hidden_states = () if output_hidden_states and self.config.ngram > 0 else None
+
+        all_main_stream_attns = () if output_attentions else None
+        all_ngram_stream_attns = () if output_attentions else None
+        all_cross_attns = () if output_attentions and self.config.add_cross_attention else None
+        present_key_values = () if use_cache else None
+
+        for idx, decoder_layer in enumerate(self.layers):
+            if output_hidden_states:
+                all_main_stream_hidden_states += (hidden_states[:sequence_length].transpose(0, 1),)
+                if self.config.ngram > 0:
+                    all_ngram_stream_hidden_states += (hidden_states[sequence_length:].transpose(0, 1),)
+
+            layer_state = past_key_values[idx] if past_key_values is not None else None
+            (
+                hidden_states,
+                layer_self_attn,
+                layer_self_predict_attn_output,
+                layer_cross_attn,
+                layer_past,
+            ) = decoder_layer(
+                hidden_states,
+                encoder_hidden_states=encoder_hidden_states,
+                encoder_attn_mask=extended_encoder_attention_mask,
+                layer_state=layer_state,
+                attention_mask=extended_attention_mask,
+                extended_predict_attention_mask=extended_predict_attention_mask,
+                main_relative_position_buckets=main_relative_position_buckets,
+                predict_relative_position_buckets=predict_relative_position_buckets,
+                position_ids=position_ids,
+            )
+            if use_cache:
+                present_key_values += (layer_past,)
+
+            if output_attentions:
+                all_main_stream_attns += (layer_self_attn,)
+                all_ngram_stream_attns += (layer_self_predict_attn_output,)
+
+                if self.config.add_cross_attention:
+                    all_cross_attns += (layer_cross_attn,)
+
+        if output_hidden_states:
+            all_main_stream_hidden_states += (hidden_states[:sequence_length].transpose(0, 1),)
+            if self.config.ngram > 0:
+                all_ngram_stream_hidden_states += (hidden_states[sequence_length:].transpose(0, 1),)
+
+        # split last_hidden_state for return
+        last_hidden_state = hidden_states[:sequence_length].transpose(0, 1)
+        last_hidden_state_ngram = hidden_states[sequence_length:].transpose(0, 1) if self.config.ngram > 0 else None
+        encoder_hidden_states = encoder_hidden_states.transpose(0, 1) if encoder_hidden_states is not None else None
+
+        if not return_dict:
+            return tuple(
+                v
+                for v in [
+                    last_hidden_state,
+                    last_hidden_state_ngram,
+                    present_key_values,
+                    all_main_stream_hidden_states,
+                    all_ngram_stream_hidden_states,
+                    all_main_stream_attns,
+                    all_ngram_stream_attns,
+                    all_cross_attns,
+                ]
+                if v is not None
+            )
+        return ProphetNetDecoderModelOutput(
+            last_hidden_state=last_hidden_state,
+            last_hidden_state_ngram=last_hidden_state_ngram,
+            past_key_values=present_key_values,
+            hidden_states=all_main_stream_hidden_states,
+            hidden_states_ngram=all_ngram_stream_hidden_states,
+            attentions=all_main_stream_attns,
+            ngram_attentions=all_ngram_stream_attns,
+            cross_attentions=all_cross_attns,
+        )
+
+    def compute_buffered_relative_buckets(self, position_ids):
+        batch_size, sequence_length = position_ids.shape
+
+        position_ids = torch.arange(1, self.max_target_positions).to(position_ids.device).repeat(1, 1)
+        main_relative_buckets, predict_relative_buckets = compute_all_stream_relative_buckets(
+            self.num_buckets, self.relative_max_distance, position_ids
+        )
+
+        # buffer relative buckets
+        main_relative_buckets = main_relative_buckets[:, :sequence_length, :sequence_length].repeat(batch_size, 1, 1)
+        predict_relative_buckets = torch.cat(
+            [
+                predict_relative_buckets[:, :sequence_length, :sequence_length],
+                predict_relative_buckets[
+                    :, :sequence_length, self.max_target_positions : self.max_target_positions + sequence_length
+                ],
+            ],
+            2,
+        ).repeat(batch_size, 1, 1)
+
+        return main_relative_buckets, predict_relative_buckets
+
+    def prepare_attention_mask(self, hidden_states, attention_mask):
+        seq_length, batch_size = hidden_states.shape[:2]
+
+        # get causal mask
+        causal_mask = hidden_states.new(seq_length, seq_length).float().fill_(-float("inf"))
+        causal_mask = torch.triu(causal_mask, 1)
+        extended_causal_mask = causal_mask[:seq_length, :seq_length][None, :, :].expand(
+            (batch_size,) + causal_mask.shape
+        )
+
+        # add usual attention mask
+        if attention_mask is not None:
+            extended_attention_mask = (1.0 - attention_mask[:, None, :]) * -10000.0
+            extended_attention_mask = extended_causal_mask + extended_attention_mask
+        else:
+            extended_attention_mask = extended_causal_mask
+        return extended_attention_mask.repeat(self.config.num_decoder_attention_heads, 1, 1).to(hidden_states.dtype)
+
+    def prepare_predict_attention_mask(self, hidden_states, attention_mask):
+        seq_length, batch_size = hidden_states.shape[:2]
+
+        # get causal mask
+        predict_causal_mask = ngram_attention_bias(
+            self.max_target_positions, self.ngram, hidden_states.device, hidden_states.dtype
+        )
+        predict_causal_mask = torch.cat(
+            [
+                predict_causal_mask[:, :seq_length, :seq_length],
+                predict_causal_mask[
+                    :, :seq_length, self.max_target_positions : self.max_target_positions + seq_length
+                ],
+            ],
+            dim=-1,
+        )
+        extended_predict_causal_mask = predict_causal_mask[:, None, :, :].expand(
+            predict_causal_mask.shape[:1] + (batch_size,) + predict_causal_mask.shape[1:]
+        )
+
+        # add usual attention mask
+        if attention_mask is not None:
+            extended_attention_mask = (1.0 - attention_mask[None, :, None, :]) * -10000.0
+            extended_attention_mask = extended_attention_mask.expand((self.ngram, batch_size, seq_length, seq_length))
+            # predicted stream attention_mask should always be 0
+            extended_attention_mask = torch.cat(
+                [extended_attention_mask, torch.zeros_like(extended_attention_mask)], dim=-1
+            )
+            extended_predict_attention_mask = extended_predict_causal_mask + extended_attention_mask
+        else:
+            extended_predict_attention_mask = extended_predict_causal_mask
+        return extended_predict_attention_mask.repeat(1, self.config.num_decoder_attention_heads, 1, 1).to(
+            hidden_states.dtype
+        )
+
+
+@add_start_docstrings(
+    "The bare ProphetNet Model outputting raw hidden-states without any specific head on top.",
+    PROPHETNET_START_DOCSTRING,
+)
+class ProphetNetModel(ProphetNetPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
+
+        encoder_config = copy.deepcopy(config)
+        encoder_config.is_encoder_decoder = False
+        encoder_config.use_cache = False
+        self.encoder = ProphetNetEncoder(encoder_config, self.word_embeddings)
+
+        decoder_config = copy.deepcopy(config)
+        decoder_config.is_decoder = True
+        decoder_config.is_encoder_decoder = False
+        self.decoder = ProphetNetDecoder(decoder_config, self.word_embeddings)
+
+        self.init_weights()
+
+    def get_input_embeddings(self):
+        return self.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.word_embeddings = value
+        self.encoder.word_embeddings = self.word_embeddings
+        self.decoder.word_embeddings = self.word_embeddings
+
+    def get_encoder(self):
+        return self.encoder
+
+    def get_decoder(self):
+        return self.decoder
+
+    @add_start_docstrings_to_callable(PROPHETNET_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=ProphetNetSeq2SeqModelOutput, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        decoder_input_ids=None,
+        decoder_attention_mask=None,
+        encoder_outputs: Optional[Tuple] = None,
+        past_key_values=None,
+        inputs_embeds=None,
+        decoder_inputs_embeds=None,
+        use_cache=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+    ):
+        r"""
+        Returns:
+
+        Example::
+
+            >>> from transformers import ProphetNetTokenizer, ProphetNetModel
+
+            >>> tokenizer = ProphetNetTokenizer.from_pretrained('microsoft/prophetnet-large-uncased')
+            >>> model = ProphetNetModel.from_pretrained('microsoft/prophetnet-large-uncased')
+
+            >>> input_ids = tokenizer("Studies have been shown that owning a dog is good for you", return_tensors="pt").input_ids  # Batch size 1
+            >>> decoder_input_ids = tokenizer("Studies show that", return_tensors="pt").input_ids  # Batch size 1
+            >>> outputs = model(input_ids=input_ids, decoder_input_ids=decoder_input_ids, return_dict=True)
+
+            >>> last_hidden_states = outputs.last_hidden_state  # main stream hidden states
+            >>> last_hidden_states_ngram = outputs.last_hidden_state_ngram  # predict hidden states
+        """
+
+        use_cache == use_cache if use_cache is not None else self.config.use_cache
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if encoder_outputs is None:
+            encoder_outputs = self.encoder(
+                input_ids=input_ids,
+                attention_mask=attention_mask,
+                inputs_embeds=inputs_embeds,
+                output_attentions=output_attentions,
+                output_hidden_states=output_hidden_states,
+                return_dict=return_dict,
+            )
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        decoder_outputs = self.decoder(
+            input_ids=decoder_input_ids,
+            attention_mask=decoder_attention_mask,
+            encoder_hidden_states=encoder_outputs[0],
+            encoder_attention_mask=attention_mask,
+            past_key_values=past_key_values,
+            inputs_embeds=decoder_inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            use_cache=use_cache,
+            return_dict=return_dict,
+        )
+
+        if not return_dict:
+            return decoder_outputs + encoder_outputs
+        return ProphetNetSeq2SeqModelOutput(
+            last_hidden_state=decoder_outputs.last_hidden_state,
+            last_hidden_state_ngram=decoder_outputs.last_hidden_state_ngram,
+            past_key_values=decoder_outputs.past_key_values,
+            decoder_hidden_states=decoder_outputs.hidden_states,
+            decoder_ngram_hidden_states=decoder_outputs.hidden_states_ngram,
+            decoder_attentions=decoder_outputs.attentions,
+            decoder_ngram_attentions=decoder_outputs.ngram_attentions,
+            decoder_cross_attentions=decoder_outputs.cross_attentions,
+            encoder_last_hidden_state=encoder_outputs.last_hidden_state,
+            encoder_hidden_states=encoder_outputs.hidden_states,
+            encoder_attentions=encoder_outputs.attentions,
+        )
+
+
+@add_start_docstrings(
+    "The ProphetNet Model with a language modeling head. Can be used for sequence generation tasks.",
+    PROPHETNET_START_DOCSTRING,
+)
+class ProphetNetForConditionalGeneration(ProphetNetPreTrainedModel):
+    def __init__(self, config: ProphetNetConfig):
+        super().__init__(config)
+        self.prophetnet = ProphetNetModel(config)
+        self.padding_idx = config.pad_token_id
+        self.disable_ngram_loss = config.disable_ngram_loss
+
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        self.init_weights()
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def get_input_embeddings(self):
+        return self.prophetnet.word_embeddings
+
+    @add_start_docstrings_to_callable(PROPHETNET_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=ProphetNetSeq2SeqLMOutput, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        decoder_input_ids=None,
+        decoder_attention_mask=None,
+        encoder_outputs=None,
+        past_key_values=None,
+        inputs_embeds=None,
+        decoder_inputs_embeds=None,
+        labels=None,
+        use_cache=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+    ):
+        r"""
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`):
+            Labels for computing the sequence classification/regression loss.
+            Indices should be in :obj:`[-100, 0, ..., config.vocab_size - 1]`.
+            All labels set to ``-100`` are ignored (masked), the loss is only
+            computed for labels in ``[0, ..., config.vocab_size]``
+
+        Returns:
+
+        Example::
+
+            >>> from transformers import ProphetNetTokenizer, ProphetNetForConditionalGeneration
+
+            >>> tokenizer = ProphetNetTokenizer.from_pretrained('microsoft/prophetnet-large-uncased')
+            >>> model = ProphetNetForConditionalGeneration.from_pretrained('microsoft/prophetnet-large-uncased')
+
+            >>> input_ids = tokenizer("Studies have been shown that owning a dog is good for you", return_tensors="pt").input_ids  # Batch size 1
+            >>> decoder_input_ids = tokenizer("Studies show that", return_tensors="pt").input_ids  # Batch size 1
+            >>> outputs = model(input_ids=input_ids, decoder_input_ids=decoder_input_ids, return_dict=True)
+
+            >>> logits_next_token = outputs.logits  # logits to predict next token as usual
+            >>> logits_ngram_next_tokens = outputs.logits_ngram  # logits to predict 2nd, 3rd, ... next tokens
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if labels is not None and decoder_input_ids is None and decoder_inputs_embeds is None:
+            # get decoder inputs from shifting lm labels to the right
+            decoder_input_ids = self._shift_right(labels)
+
+        outputs = self.prophetnet(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            decoder_input_ids=decoder_input_ids,
+            decoder_attention_mask=decoder_attention_mask,
+            encoder_outputs=encoder_outputs,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            decoder_inputs_embeds=decoder_inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        batch_size, sequence_length = (
+            decoder_input_ids.shape if decoder_input_ids is not None else decoder_inputs_embeds.shape[:2]
+        )
+
+        predicting_streams = outputs[1].view(batch_size, self.config.ngram, sequence_length, -1)
+        predict_logits = self.lm_head(predicting_streams)
+
+        logits = predict_logits[:, 0]
+        logits_ngram = predict_logits[:, 1:] if self.config.ngram > 1 else None
+
+        loss = None
+        if labels is not None:
+            loss = self._compute_loss(predict_logits, labels)
+
+        if not return_dict:
+            all_logits = tuple(v for v in [logits, logits_ngram] if v is not None)
+            return (loss,) + all_logits + outputs[2:] if loss is not None else all_logits + outputs[2:]
+        else:
+            return ProphetNetSeq2SeqLMOutput(
+                loss=loss,
+                logits=logits,
+                logits_ngram=logits_ngram,
+                past_key_values=outputs.past_key_values,
+                decoder_hidden_states=outputs.decoder_hidden_states,
+                decoder_ngram_hidden_states=outputs.decoder_ngram_hidden_states,
+                decoder_attentions=outputs.decoder_attentions,
+                decoder_ngram_attentions=outputs.decoder_ngram_attentions,
+                decoder_cross_attentions=outputs.decoder_cross_attentions,
+                encoder_last_hidden_state=outputs.encoder_last_hidden_state,
+                encoder_hidden_states=outputs.encoder_hidden_states,
+                encoder_attentions=outputs.encoder_attentions,
+            )
+
+    def _compute_loss(self, logits, labels):
+        expend_targets = labels.new_zeros(self.config.ngram, labels.size(0), labels.size(1)).fill_(self.padding_idx)
+
+        for i in range(self.config.ngram):
+            if i > 0 and self.disable_ngram_loss:
+                break
+            expend_targets[i, :, :] = labels
+
+        lprobs = F.log_softmax(
+            logits.view(-1, logits.size(-1)),
+            dim=-1,
+            dtype=torch.float32,
+        )
+
+        loss = F.nll_loss(lprobs, expend_targets.view(-1), reduction="sum")
+
+        if self.config.eps > 0.0:
+            smooth_loss = -lprobs.sum(dim=-1, keepdim=True)
+            non_pad_mask = expend_targets.ne(self.padding_idx).view(-1)
+            smooth_loss = smooth_loss[non_pad_mask]
+            smooth_loss = smooth_loss.sum()
+
+            eps_i = self.config.eps / lprobs.size(-1)
+            loss = (1.0 - self.config.eps) * loss + eps_i * smooth_loss
+
+        return loss
+
+    def prepare_inputs_for_generation(
+        self, decoder_input_ids, past, attention_mask, use_cache, encoder_outputs, **kwargs
+    ):
+        assert encoder_outputs is not None, "`encoder_outputs` have to be passed for generation."
+
+        if past:
+            decoder_input_ids = decoder_input_ids[:, -1:]
+        # first step, decoder_cached_states are empty
+        return {
+            "input_ids": None,  # encoder_outputs is defined. input_ids not needed
+            "encoder_outputs": encoder_outputs,
+            "past_key_values": past,
+            "decoder_input_ids": decoder_input_ids,
+            "attention_mask": attention_mask,
+            "use_cache": use_cache,
+        }
+
+    @staticmethod
+    def _reorder_cache(past, beam_idx):
+        # this function reorders the cache for beam search
+        def _reorder_cache(cache_dict, beam_idx):
+            for k, key_value_states in cache_dict.items():
+                if key_value_states is not None:
+                    cache_dict[k] = key_value_states.index_select(0, beam_idx)
+            return cache_dict
+
+        reordered_past = []
+        for layer_past in past:
+            # get the correct batch idx from decoder layer's batch dim for cross and self-attn
+            layer_past_new = {
+                attn_key: _reorder_cache(attn_cache, beam_idx) for attn_key, attn_cache in layer_past.items()
+            }
+            reordered_past.append(layer_past_new)
+        return reordered_past
+
+    def get_encoder(self):
+        return self.prophetnet.encoder
+
+    def get_decoder(self):
+        return self.prophetnet.decoder
+
+
+@add_start_docstrings(
+    "The standalone decoder part of the ProphetNetModel with a lm head on top. The model can be used for causal language modeling.",
+    PROPHETNET_START_DOCSTRING,
+)
+class ProphetNetForCausalLM(ProphetNetPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        # set config for CLM
+        config = copy.deepcopy(config)
+        config.is_decoder = True
+        config.is_encoder_decoder = False
+        self.decoder = ProphetNetDecoder(config)
+
+        self.padding_idx = config.pad_token_id
+        self.disable_ngram_loss = config.disable_ngram_loss
+
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        self.init_weights()
+
+    def get_input_embeddings(self):
+        return self.decoder.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.decoder.word_embeddings = value
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    @add_start_docstrings_to_callable(PROPHETNET_STANDALONE_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=ProphetNetDecoderLMOutput, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_values=None,
+        inputs_embeds=None,
+        labels=None,
+        use_cache=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+    ):
+        r"""
+        encoder_hidden_states (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`):
+            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention
+            if the model is configured as a decoder.
+        encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
+            Mask to avoid performing attention on the padding token indices of the encoder input. This mask
+            is used in the cross-attention if the model is configured as a decoder.
+            Mask values selected in ``[0, 1]``:
+        past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
+            Contains precomputed key and value hidden-states of the attention blocks. Can be used to speed up decoding.
+
+            If :obj:`past_key_values` are used, the user can optionally input only the last
+            ``decoder_input_ids`` (those that don't have their past key value states given to this model) of shape
+            :obj:`(batch_size, 1)` instead of all ``decoder_input_ids`` of shape :obj:`(batch_size, sequence_length)`.
+        use_cache (:obj:`bool`, `optional`):
+            If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up
+            decoding (see :obj:`past_key_values`).
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
+            Labels for computing the left-to-right language modeling loss (next word prediction).
+            Indices should be in ``[-100, 0, ..., config.vocab_size]`` (see ``input_ids`` docstring)
+            Tokens with indices set to ``-100`` are ignored (masked), the loss is only computed for the tokens with labels
+            n ``[0, ..., config.vocab_size]``
+
+        Returns:
+
+        Example::
+
+            >>> from transformers import ProphetNetTokenizer, ProphetNetForCausalLM
+            >>> import torch
+
+            >>> tokenizer = ProphetNetTokenizer.from_pretrained('microsoft/prophetnet-large-uncased')
+            >>> model = ProphetNetForCausalLM.from_pretrained('patrickvonplaten/prophetnet-decoder-clm-large-uncased', return_dict=True)
+            >>> assert model.config.is_decoder, f"{model.__class__} has to be configured as a decoder."
+            >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
+            >>> outputs = model(**inputs, return_dict=True)
+
+            >>> logits = outputs.logits
+
+            >>> # Model can also be used with EncoderDecoder framework
+            >>> from transformers import BertTokenizer, EncoderDecoderModel
+            >>> import torch
+
+            >>> tokenizer = BertTokenizer.from_pretrained('bert-uncased-large')
+            >>> model = EncoderDecoderModel.from_encoder_decoder_pretrained("bert-uncased-large", "patrickvonplaten/prophetnet-decoder-clm-large-uncased")
+
+            >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
+            >>> outputs = model(input_ids=inputs["input_ids"], labels=inputs["input_ids"], return_dict=True)
+
+            >>> loss = outputs.loss
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.decoder(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        batch_size, sequence_length = input_ids.shape if input_ids is not None else inputs_embeds.shape[:2]
+
+        predicting_streams = outputs[1].view(batch_size, self.config.ngram, sequence_length, -1)
+        predict_logits = self.lm_head(predicting_streams)
+
+        logits = predict_logits[:, 0]
+        logits_ngram = predict_logits[:, 1:] if self.config.ngram > 1 else None
+
+        loss = None
+        if labels is not None:
+            loss = self._compute_loss(predict_logits, labels)
+
+        if not return_dict:
+            all_logits = tuple(v for v in [logits, logits_ngram] if v is not None)
+            return (loss,) + all_logits + outputs[2:] if loss is not None else all_logits + outputs[2:]
+        else:
+            return ProphetNetDecoderLMOutput(
+                loss=loss,
+                logits=logits,
+                logits_ngram=logits_ngram,
+                past_key_values=outputs.past_key_values,
+                hidden_states=outputs.hidden_states,
+                hidden_states_ngram=outputs.hidden_states_ngram,
+                attentions=outputs.attentions,
+                ngram_attentions=outputs.ngram_attentions,
+            )
+
+    def _compute_loss(self, logits, labels):
+        expend_targets = labels.new_zeros(self.config.ngram, labels.size(0), labels.size(1)).fill_(self.padding_idx)
+
+        for i in range(self.config.ngram):
+            if i > 0 and self.disable_ngram_loss:
+                break
+            expend_targets[i, :, :] = labels
+
+        lprobs = F.log_softmax(
+            logits.view(-1, logits.size(-1)),
+            dim=-1,
+            dtype=torch.float32,
+        )
+
+        loss = F.nll_loss(lprobs, expend_targets.view(-1), reduction="sum")
+
+        if self.config.eps > 0.0:
+            smooth_loss = -lprobs.sum(dim=-1, keepdim=True)
+            non_pad_mask = expend_targets.ne(self.padding_idx).view(-1)
+            smooth_loss = smooth_loss[non_pad_mask]
+            smooth_loss = smooth_loss.sum()
+
+            eps_i = self.config.eps / lprobs.size(-1)
+            loss = (1.0 - self.config.eps) * loss + eps_i * smooth_loss
+
+        return loss
+
+    def prepare_inputs_for_generation(self, input_ids, past=None, attention_mask=None, use_cache=None, **kwargs):
+        # if model is used as a decoder in encoder-decoder model, the decoder attention mask is created on the fly
+        if attention_mask is None:
+            attention_mask = input_ids.new_ones(input_ids.shape)
+
+        if past:
+            input_ids = input_ids[:, -1:]
+        # first step, decoder_cached_states are empty
+        return {
+            "input_ids": input_ids,  # encoder_outputs is defined. input_ids not needed
+            "attention_mask": attention_mask,
+            "past_key_values": past,
+            "use_cache": use_cache,
+        }
+
+    @staticmethod
+    def _reorder_cache(past, beam_idx):
+        # this function reorders the cache for beam search
+        def _reorder_cache(cache_dict, beam_idx):
+            for k, key_value_states in cache_dict.items():
+                if key_value_states is not None:
+                    cache_dict[k] = key_value_states.index_select(0, beam_idx)
+            return cache_dict
+
+        reordered_past = []
+        for layer_past in past:
+            # get the correct batch idx from decoder layer's batch dim for cross and self-attn
+            layer_past_new = {
+                attn_key: _reorder_cache(attn_cache, beam_idx) for attn_key, attn_cache in layer_past.items()
+            }
+            reordered_past.append(layer_past_new)
+        return reordered_past
+
+    def set_decoder(self, decoder):
+        self.decoder = decoder
+
+    def get_decoder(self):
+        return self.decoder

src/transformers/modeling_tf_lxmert.py

@@ -17,7 +17,6 @@
 """ TF 2.0 LXMERT model. """
 
 
-import logging
 from dataclasses import dataclass
 from typing import Dict, Optional, Tuple
 
@@ -34,9 +33,10 @@ from .file_utils import (
 )
 from .modeling_tf_utils import TFPreTrainedModel, get_initializer, keras_serializable, shape_list
 from .tokenization_utils_base import BatchEncoding
+from .utils import logging
 
 
-logger = logging.getLogger(__name__)
+logger = logging.get_logger(__name__)
 
 
 _CONFIG_FOR_DOC = "LxmertConfig"