mirror of
https://github.com/huggingface/transformers.git
synced 2025-07-31 02:02:21 +06:00
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>
This commit is contained in:
Weizhen
GitHub
parent
8f8f8d99fc
commit
2422cda01b
@ -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).
|
||||
|
||||
@ -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
|
||||
|
||||
|
||||
@ -27,4 +27,4 @@ EncoderDecoderModel
|
||||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||||
|
||||
.. autoclass:: transformers.EncoderDecoderModel
|
||||
:members: forward
|
||||
:members: forward, from_encoder_decoder_pretrained
|
||||
|
||||
83
docs/source/model_doc/prophetnet.rst
Normal file
83
docs/source/model_doc/prophetnet.rst
Normal file
@ -0,0 +1,83 @@
|
||||
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
|
||||
63
docs/source/model_doc/xlmprophetnet.rst
Normal file
63
docs/source/model_doc/xlmprophetnet.rst
Normal file
@ -0,0 +1,63 @@
|
||||
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
|
||||
@ -0,0 +1,29 @@
|
||||
## 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}
|
||||
}
|
||||
```
|
||||
17
model_cards/microsoft/prophetnet-large-uncased/README.md
Normal file
17
model_cards/microsoft/prophetnet-large-uncased/README.md
Normal file
@ -0,0 +1,17 @@
|
||||
## 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}
|
||||
}
|
||||
```
|
||||
@ -0,0 +1,32 @@
|
||||
## 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}
|
||||
}
|
||||
```
|
||||
@ -0,0 +1,31 @@
|
||||
## 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}
|
||||
}
|
||||
```
|
||||
@ -0,0 +1,19 @@
|
||||
## 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}
|
||||
}
|
||||
```
|
||||
@ -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,
|
||||
|
||||
@ -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"),
|
||||
]
|
||||
)
|
||||
|
||||
|
||||
@ -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": "",
|
||||
|
||||
167
src/transformers/configuration_prophetnet.py
Normal file
167
src/transformers/configuration_prophetnet.py
Normal file
@ -0,0 +1,167 @@
|
||||
# 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
|
||||
35
src/transformers/configuration_xlm_prophetnet.py
Normal file
35
src/transformers/configuration_xlm_prophetnet.py
Normal file
@ -0,0 +1,35 @@
|
||||
# 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"
|
||||
@ -0,0 +1,162 @@
|
||||
# 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)
|
||||
@ -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),
|
||||
]
|
||||
)
|
||||
|
||||
|
||||
@ -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"
|
||||
|
||||
2052
src/transformers/modeling_prophetnet.py
Normal file
2052
src/transformers/modeling_prophetnet.py
Normal file
File diff suppressed because it is too large
Load Diff
@ -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"
|
||||
|
||||
@ -490,12 +490,17 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin):
|
||||
self._tie_or_clone_weights(output_embeddings, self.get_input_embeddings())
|
||||
|
||||
if self.config.is_encoder_decoder and self.config.tie_encoder_decoder:
|
||||
if hasattr(self, self.base_model_prefix):
|
||||
self = getattr(self, self.base_model_prefix)
|
||||
self._tie_encoder_decoder_weights(self.encoder, self.decoder, self.base_model_prefix)
|
||||
|
||||
@staticmethod
|
||||
def _tie_encoder_decoder_weights(encoder: nn.Module, decoder: nn.Module, base_model_prefix: str):
|
||||
uninitialized_encoder_weights: List[str] = []
|
||||
assert decoder.__class__ == encoder.__class__, f"{decoder.__class__} and {encoder.__class__} have to be equal."
|
||||
if decoder.__class__ != encoder.__class__:
|
||||
logger.info(
|
||||
f"{decoder.__class__} and {encoder.__class__} are not equal. In this case make sure that all encoder weights are correctly initialized."
|
||||
)
|
||||
|
||||
def tie_encoder_to_decoder_recursively(
|
||||
decoder_pointer: nn.Module,
|
||||
@ -528,7 +533,9 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin):
|
||||
if name.isdigit():
|
||||
encoder_name = str(int(name) + encoder_layer_pos)
|
||||
decoder_name = name
|
||||
if not isinstance(decoder_modules[decoder_name], type(encoder_modules[encoder_name])):
|
||||
if not isinstance(decoder_modules[decoder_name], type(encoder_modules[encoder_name])) and len(
|
||||
encoder_modules
|
||||
) != len(decoder_modules):
|
||||
# this can happen if the name corresponds to the position in a list module list of layers
|
||||
# in this case the decoder has added a cross-attention that the encoder does not have
|
||||
# thus skip this step and substract one layer pos from encoder
|
||||
|
||||
159
src/transformers/modeling_xlm_prophetnet.py
Normal file
159
src/transformers/modeling_xlm_prophetnet.py
Normal file
@ -0,0 +1,159 @@
|
||||
# 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 XLM-ProphetNet model."""
|
||||
|
||||
from .configuration_xlm_prophetnet import XLMProphetNetConfig
|
||||
from .modeling_prophetnet import (
|
||||
ProphetNetDecoder,
|
||||
ProphetNetEncoder,
|
||||
ProphetNetForCausalLM,
|
||||
ProphetNetForConditionalGeneration,
|
||||
ProphetNetModel,
|
||||
)
|
||||
from .utils import logging
|
||||
|
||||
|
||||
logger = logging.get_logger(__name__)
|
||||
|
||||
_TOKENIZER_FOR_DOC = "XLMProphetNetTokenizer"
|
||||
|
||||
XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST = [
|
||||
"microsoft/xprophetnet-large-wiki100-cased",
|
||||
# See all ProphetNet models at https://huggingface.co/models?filter=xprophetnet
|
||||
]
|
||||
|
||||
|
||||
class XLMProphetNetEncoder(ProphetNetEncoder):
|
||||
r"""
|
||||
This class overrides :class:`~transformers.ProphetNetEncoder`. Please check the
|
||||
superclass for the appropriate documentation alongside usage examples.
|
||||
|
||||
Example::
|
||||
|
||||
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetEncoder
|
||||
>>> import torch
|
||||
|
||||
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained('microsoft/xprophetnet-large-wiki100-cased')
|
||||
>>> model = XLMProphetNetEncoder.from_pretrained('patrickvonplaten/xprophetnet-large-uncased-standalone', 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
|
||||
"""
|
||||
|
||||
config_class = XLMProphetNetConfig
|
||||
|
||||
|
||||
class XLMProphetNetDecoder(ProphetNetDecoder):
|
||||
r"""
|
||||
This class overrides :class:`~transformers.ProphetNetDecoder`. Please check the
|
||||
superclass for the appropriate documentation alongside usage examples.
|
||||
|
||||
Example::
|
||||
|
||||
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetDecoder
|
||||
>>> import torch
|
||||
|
||||
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained('microsoft/xprophetnet-large-wiki100-cased')
|
||||
>>> model = XLMProphetNetDecoder.from_pretrained('patrickvonplaten/xprophetnet-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
|
||||
"""
|
||||
|
||||
config_class = XLMProphetNetConfig
|
||||
|
||||
|
||||
class XLMProphetNetModel(ProphetNetModel):
|
||||
r"""
|
||||
This class overrides :class:`~transformers.ProphetNetModel`. Please check the
|
||||
superclass for the appropriate documentation alongside usage examples.
|
||||
|
||||
Example::
|
||||
|
||||
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetModel
|
||||
|
||||
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained('microsoft/xprophetnet-large-wiki100-cased')
|
||||
>>> model = XLMProphetNetModel.from_pretrained('microsoft/xprophetnet-large-wiki100-cased')
|
||||
|
||||
>>> 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
|
||||
"""
|
||||
|
||||
config_class = XLMProphetNetConfig
|
||||
|
||||
|
||||
class XLMProphetNetForConditionalGeneration(ProphetNetForConditionalGeneration):
|
||||
r"""
|
||||
This class overrides :class:`~transformers.ProphetNetForConditionalGeneration`. Please check the
|
||||
superclass for the appropriate documentation alongside usage examples.
|
||||
|
||||
Example::
|
||||
|
||||
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetForConditionalGeneration
|
||||
|
||||
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained('microsoft/xprophetnet-large-wiki100-cased')
|
||||
>>> model = XLMProphetNetForConditionalGeneration.from_pretrained('microsoft/xprophetnet-large-wiki100-cased')
|
||||
|
||||
>>> 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
|
||||
"""
|
||||
|
||||
config_class = XLMProphetNetConfig
|
||||
|
||||
|
||||
class XLMProphetNetForCausalLM(ProphetNetForCausalLM):
|
||||
r"""
|
||||
This class overrides :class:`~transformers.ProphetNetForCausalLM`. Please check the
|
||||
superclass for the appropriate documentation alongside usage examples.
|
||||
|
||||
Example::
|
||||
|
||||
>>> from transformers import XLMProphetNetTokenizer, XLMProphetNetForCausalLM
|
||||
>>> import torch
|
||||
|
||||
>>> tokenizer = XLMProphetNetTokenizer.from_pretrained('microsoft/xprophetnet-large-wiki100-cased')
|
||||
>>> model = XLMProphetNetForCausalLM.from_pretrained('patrickvonplaten/xprophetnet-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)
|
||||
|
||||
>>> 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/xprophetnet-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"])
|
||||
|
||||
>>> loss = outputs.loss
|
||||
"""
|
||||
|
||||
config_class = XLMProphetNetConfig
|
||||
@ -43,6 +43,7 @@ from .configuration_auto import (
|
||||
MobileBertConfig,
|
||||
OpenAIGPTConfig,
|
||||
PegasusConfig,
|
||||
ProphetNetConfig,
|
||||
RagConfig,
|
||||
ReformerConfig,
|
||||
RetriBertConfig,
|
||||
@ -51,6 +52,7 @@ from .configuration_auto import (
|
||||
T5Config,
|
||||
TransfoXLConfig,
|
||||
XLMConfig,
|
||||
XLMProphetNetConfig,
|
||||
XLMRobertaConfig,
|
||||
XLNetConfig,
|
||||
replace_list_option_in_docstrings,
|
||||
@ -77,6 +79,7 @@ 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
|
||||
@ -95,6 +98,7 @@ 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:
|
||||
@ -199,6 +203,8 @@ TOKENIZER_MAPPING = OrderedDict(
|
||||
(BertGenerationConfig, (BertGenerationTokenizer, None)),
|
||||
(DebertaConfig, (DebertaTokenizer, None)),
|
||||
(RagConfig, (RagTokenizer, None)),
|
||||
(XLMProphetNetConfig, (XLMProphetNetTokenizer, None)),
|
||||
(ProphetNetConfig, (ProphetNetTokenizer, None)),
|
||||
]
|
||||
)
|
||||
|
||||
|
||||
@ -14,7 +14,6 @@
|
||||
# limitations under the License.
|
||||
""" Tokenization class for model DeBERTa."""
|
||||
|
||||
import logging
|
||||
import os
|
||||
import pathlib
|
||||
import random
|
||||
@ -28,6 +27,7 @@ import tqdm
|
||||
import requests
|
||||
|
||||
from .tokenization_utils import PreTrainedTokenizer
|
||||
from .utils import logging
|
||||
|
||||
|
||||
try:
|
||||
@ -36,7 +36,7 @@ except ImportError:
|
||||
raise ImportError("Please install regex with: pip install regex")
|
||||
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
logger = logging.get_logger(__name__)
|
||||
|
||||
VOCAB_FILES_NAMES = {"vocab_file": "bpe_encoder.bin"}
|
||||
|
||||
|
||||
@ -16,7 +16,6 @@
|
||||
|
||||
|
||||
import json
|
||||
import logging
|
||||
import os
|
||||
import re
|
||||
import unicodedata
|
||||
@ -27,9 +26,10 @@ import sacremoses as sm
|
||||
from .file_utils import add_start_docstrings
|
||||
from .tokenization_utils import BatchEncoding, PreTrainedTokenizer
|
||||
from .tokenization_utils_base import PREPARE_SEQ2SEQ_BATCH_DOCSTRING
|
||||
from .utils import logging
|
||||
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
logger = logging.get_logger(__name__)
|
||||
|
||||
VOCAB_FILES_NAMES = {
|
||||
"src_vocab_file": "vocab-src.json",
|
||||
|
||||
284
src/transformers/tokenization_prophetnet.py
Normal file
284
src/transformers/tokenization_prophetnet.py
Normal file
@ -0,0 +1,284 @@
|
||||
# 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.
|
||||
|
||||
import collections
|
||||
import os
|
||||
from typing import List, Optional, Tuple
|
||||
|
||||
from .tokenization_bert import BasicTokenizer, WordpieceTokenizer
|
||||
from .tokenization_utils import PreTrainedTokenizer
|
||||
from .utils import logging
|
||||
|
||||
|
||||
logger = logging.get_logger(__name__)
|
||||
|
||||
VOCAB_FILES_NAMES = {"vocab_file": "prophetnet.tokenizer"}
|
||||
|
||||
PRETRAINED_VOCAB_FILES_MAP = {
|
||||
"vocab_file": {
|
||||
"microsoft/prophetnet-large-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/microsoft/prophetnet-large-uncased/prophetnet.tokenizer",
|
||||
}
|
||||
}
|
||||
|
||||
PRETRAINED_INIT_CONFIGURATION = {
|
||||
"microsoft/prophetnet-large-uncased": {"do_lower_case": True},
|
||||
}
|
||||
|
||||
PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
|
||||
"microsoft/prophetnet-large-uncased": 512,
|
||||
}
|
||||
|
||||
|
||||
def load_vocab(vocab_file):
|
||||
"""Loads a vocabulary file into a dictionary."""
|
||||
vocab = collections.OrderedDict()
|
||||
with open(vocab_file, "r", encoding="utf-8") as reader:
|
||||
tokens = reader.readlines()
|
||||
for index, token in enumerate(tokens):
|
||||
token = token.rstrip("\n")
|
||||
vocab[token] = index
|
||||
return vocab
|
||||
|
||||
|
||||
class ProphetNetTokenizer(PreTrainedTokenizer):
|
||||
r"""
|
||||
Construct a ProphetNetTokenizer. Based on WordPiece.
|
||||
|
||||
This tokenizer inherits from :class:`~transformers.PreTrainedTokenizer` which contains most of the main methods.
|
||||
Users should refer to this superclass for more information regarding those methods.
|
||||
|
||||
Args:
|
||||
vocab_file (:obj:`str`):
|
||||
File containing the vocabulary.
|
||||
do_lower_case (:obj:`bool`, `optional`, defaults to :obj:`True`):
|
||||
Whether or not to lowercase the input when tokenizing.
|
||||
do_basic_tokenize (:obj:`bool`, `optional`, defaults to :obj:`True`):
|
||||
Whether or not to do basic tokenization before WordPiece.
|
||||
never_split (:obj:`Iterable`, `optional`):
|
||||
Collection of tokens which will never be split during tokenization. Only has an effect when
|
||||
:obj:`do_basic_tokenize=True`
|
||||
unk_token (:obj:`str`, `optional`, defaults to :obj:`"[UNK]"`):
|
||||
The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
|
||||
token instead.
|
||||
sep_token (:obj:`str`, `optional`, defaults to :obj:`"[SEP]"`):
|
||||
The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences
|
||||
for sequence classification or for a text and a question for question answering.
|
||||
It is also used as the last token of a sequence built with special tokens.
|
||||
x_sep_token (:obj:`str`, `optional`, defaults to :obj:`"[X_SEP]"`):
|
||||
Special second separator token, which can be generated by :class:`~transformers.ProphetNetForConditionalGeneration`.
|
||||
It is used to separate bullet-point like sentences in summarization, *e.g.*.
|
||||
pad_token (:obj:`str`, `optional`, defaults to :obj:`"[PAD]"`):
|
||||
The token used for padding, for example when batching sequences of different lengths.
|
||||
cls_token (:obj:`str`, `optional`, defaults to :obj:`"[CLS]"`):
|
||||
The classifier token which is used when doing sequence classification (classification of the whole
|
||||
sequence instead of per-token classification). It is the first token of the sequence when built with
|
||||
special tokens.
|
||||
mask_token (:obj:`str`, `optional`, defaults to :obj:`"[MASK]"`):
|
||||
The token used for masking values. This is the token used when training this model with masked language
|
||||
modeling. This is the token which the model will try to predict.
|
||||
tokenize_chinese_chars (:obj:`bool`, `optional`, defaults to :obj:`True`):
|
||||
Whether or not to tokenize Chinese characters.
|
||||
|
||||
This should likely be deactivated for Japanese (see this `issue
|
||||
<https://github.com/huggingface/transformers/issues/328>`__).
|
||||
strip_accents: (:obj:`bool`, `optional`):
|
||||
Whether or not to strip all accents. If this option is not specified, then it will be determined by the
|
||||
value for :obj:`lowercase` (as in the original BERT).
|
||||
"""
|
||||
|
||||
vocab_files_names = VOCAB_FILES_NAMES
|
||||
pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
|
||||
pretrained_init_configuration = PRETRAINED_INIT_CONFIGURATION
|
||||
max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
vocab_file,
|
||||
do_lower_case=True,
|
||||
do_basic_tokenize=True,
|
||||
never_split=None,
|
||||
unk_token="[UNK]",
|
||||
sep_token="[SEP]",
|
||||
x_sep_token="[X_SEP]",
|
||||
pad_token="[PAD]",
|
||||
mask_token="[MASK]",
|
||||
tokenize_chinese_chars=True,
|
||||
strip_accents=None,
|
||||
**kwargs
|
||||
):
|
||||
super().__init__(
|
||||
unk_token=unk_token,
|
||||
sep_token=sep_token,
|
||||
pad_token=pad_token,
|
||||
mask_token=mask_token,
|
||||
x_sep_token=x_sep_token,
|
||||
**kwargs,
|
||||
)
|
||||
self.unique_no_split_tokens.append(x_sep_token)
|
||||
|
||||
if not os.path.isfile(vocab_file):
|
||||
raise ValueError(
|
||||
"Can't find a vocabulary file at path '{}'. To load the vocabulary from a Google pretrained "
|
||||
"model use `tokenizer = ProphetNetTokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`".format(vocab_file)
|
||||
)
|
||||
self.vocab = load_vocab(vocab_file)
|
||||
self.ids_to_tokens = collections.OrderedDict([(ids, tok) for tok, ids in self.vocab.items()])
|
||||
self.do_basic_tokenize = do_basic_tokenize
|
||||
if do_basic_tokenize:
|
||||
self.basic_tokenizer = BasicTokenizer(
|
||||
do_lower_case=do_lower_case,
|
||||
never_split=never_split,
|
||||
tokenize_chinese_chars=tokenize_chinese_chars,
|
||||
strip_accents=strip_accents,
|
||||
)
|
||||
self.wordpiece_tokenizer = WordpieceTokenizer(vocab=self.vocab, unk_token=self.unk_token)
|
||||
|
||||
@property
|
||||
def vocab_size(self):
|
||||
return len(self.vocab)
|
||||
|
||||
def get_vocab(self):
|
||||
return dict(self.vocab, **self.added_tokens_encoder)
|
||||
|
||||
def _tokenize(self, text):
|
||||
split_tokens = []
|
||||
if self.do_basic_tokenize:
|
||||
for token in self.basic_tokenizer.tokenize(text, never_split=self.all_special_tokens):
|
||||
|
||||
# If the token is part of the never_split set
|
||||
if token in self.basic_tokenizer.never_split:
|
||||
split_tokens.append(token)
|
||||
else:
|
||||
split_tokens += self.wordpiece_tokenizer.tokenize(token)
|
||||
else:
|
||||
split_tokens = self.wordpiece_tokenizer.tokenize(text)
|
||||
return split_tokens
|
||||
|
||||
def _convert_token_to_id(self, token):
|
||||
""" Converts a token (str) in an id using the vocab. """
|
||||
return self.vocab.get(token, self.vocab.get(self.unk_token))
|
||||
|
||||
def _convert_id_to_token(self, index):
|
||||
"""Converts an index (integer) in a token (str) using the vocab."""
|
||||
return self.ids_to_tokens.get(index, self.unk_token)
|
||||
|
||||
def convert_tokens_to_string(self, tokens):
|
||||
""" Converts a sequence of tokens (string) in a single string. """
|
||||
out_string = " ".join(tokens).replace(" ##", "").strip()
|
||||
return out_string
|
||||
|
||||
def get_special_tokens_mask(
|
||||
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
|
||||
) -> List[int]:
|
||||
"""
|
||||
Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
|
||||
special tokens using the tokenizer ``prepare_for_model`` method.
|
||||
|
||||
Args:
|
||||
token_ids_0 (:obj:`List[int]`):
|
||||
List of IDs.
|
||||
token_ids_1 (:obj:`List[int]`, `optional`):
|
||||
Optional second list of IDs for sequence pairs.
|
||||
already_has_special_tokens (:obj:`bool`, `optional`, defaults to :obj:`False`):
|
||||
Whether or not the token list is already formatted with special tokens for the model.
|
||||
|
||||
Returns:
|
||||
:obj:`List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
|
||||
"""
|
||||
if already_has_special_tokens:
|
||||
if token_ids_1 is not None:
|
||||
raise ValueError(
|
||||
"You should not supply a second sequence if the provided sequence of "
|
||||
"ids is already formated with special tokens for the model."
|
||||
)
|
||||
return list(map(lambda x: 1 if x in [self.sep_token_id, self.cls_token_id] else 0, token_ids_0))
|
||||
|
||||
if token_ids_1 is None:
|
||||
return ([0] * len(token_ids_0)) + [1]
|
||||
return ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
|
||||
|
||||
def create_token_type_ids_from_sequences(
|
||||
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
|
||||
) -> List[int]:
|
||||
"""
|
||||
Create a mask from the two sequences passed to be used in a sequence-pair classification task.
|
||||
A ProphetNet sequence pair mask has the following format:
|
||||
|
||||
::
|
||||
|
||||
0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1
|
||||
| first sequence | second sequence |
|
||||
|
||||
If :obj:`token_ids_1` is :obj:`None`, this method only returns the first portion of the mask (0s).
|
||||
|
||||
Args:
|
||||
token_ids_0 (:obj:`List[int]`):
|
||||
List of IDs.
|
||||
token_ids_1 (:obj:`List[int]`, `optional`):
|
||||
Optional second list of IDs for sequence pairs.
|
||||
|
||||
Returns:
|
||||
:obj:`List[int]`: List of `token type IDs <../glossary.html#token-type-ids>`_ according to the given
|
||||
sequence(s).
|
||||
"""
|
||||
sep = [self.sep_token_id]
|
||||
if token_ids_1 is None:
|
||||
return len(token_ids_0 + sep) * [0]
|
||||
return len(token_ids_0 + sep) * [0] + len(token_ids_1 + sep) * [1]
|
||||
|
||||
def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
|
||||
index = 0
|
||||
if os.path.isdir(save_directory):
|
||||
vocab_file = os.path.join(
|
||||
save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
|
||||
)
|
||||
else:
|
||||
vocab_file = (filename_prefix + "-" if filename_prefix else "") + save_directory
|
||||
with open(vocab_file, "w", encoding="utf-8") as writer:
|
||||
for token, token_index in sorted(self.vocab.items(), key=lambda kv: kv[1]):
|
||||
if index != token_index:
|
||||
logger.warning(
|
||||
"Saving vocabulary to {}: vocabulary indices are not consecutive."
|
||||
" Please check that the vocabulary is not corrupted!".format(vocab_file)
|
||||
)
|
||||
index = token_index
|
||||
writer.write(token + "\n")
|
||||
index += 1
|
||||
return (vocab_file,)
|
||||
|
||||
def build_inputs_with_special_tokens(
|
||||
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
|
||||
) -> List[int]:
|
||||
"""
|
||||
Build model inputs from a sequence or a pair of sequence for sequence classification tasks
|
||||
by concatenating and adding special tokens.
|
||||
A BERT sequence has the following format:
|
||||
|
||||
- single sequence: ``[CLS] X [SEP]``
|
||||
- pair of sequences: ``[CLS] A [SEP] B [SEP]``
|
||||
|
||||
Args:
|
||||
token_ids_0 (:obj:`List[int]`):
|
||||
List of IDs to which the special tokens will be added.
|
||||
token_ids_1 (:obj:`List[int]`, `optional`):
|
||||
Optional second list of IDs for sequence pairs.
|
||||
|
||||
Returns:
|
||||
:obj:`List[int]`: List of `input IDs <../glossary.html#input-ids>`__ with the appropriate special tokens.
|
||||
"""
|
||||
if token_ids_1 is None:
|
||||
return token_ids_0 + [self.sep_token_id]
|
||||
sep = [self.sep_token_id]
|
||||
return token_ids_0 + sep + token_ids_1 + sep
|
||||
307
src/transformers/tokenization_xlm_prophetnet.py
Normal file
307
src/transformers/tokenization_xlm_prophetnet.py
Normal file
@ -0,0 +1,307 @@
|
||||
# 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.
|
||||
|
||||
import collections
|
||||
import os
|
||||
from shutil import copyfile
|
||||
from typing import List, Optional, Tuple
|
||||
|
||||
from .tokenization_utils import PreTrainedTokenizer
|
||||
from .utils import logging
|
||||
|
||||
|
||||
logger = logging.get_logger(__name__)
|
||||
|
||||
SPIECE_UNDERLINE = "▁"
|
||||
|
||||
VOCAB_FILES_NAMES = {"vocab_file": "prophetnet.tokenizer"}
|
||||
|
||||
PRETRAINED_VOCAB_FILES_MAP = {
|
||||
"vocab_file": {
|
||||
"microsoft/xprophetnet-large-wiki100-cased": "https://cdn.huggingface.co/microsoft/xprophetnet-large-wiki100-cased/prophetnet.tokenizer",
|
||||
}
|
||||
}
|
||||
|
||||
PRETRAINED_INIT_CONFIGURATION = {
|
||||
"microsoft/xprophetnet-large-wiki100-cased": {"do_lower_case": False},
|
||||
}
|
||||
|
||||
PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
|
||||
"microsoft/xprophetnet-large-wiki100-cased": 512,
|
||||
}
|
||||
|
||||
|
||||
def load_vocab(vocab_file):
|
||||
"""Loads a vocabulary file into a dictionary."""
|
||||
vocab = collections.OrderedDict()
|
||||
with open(vocab_file, "r", encoding="utf-8") as reader:
|
||||
tokens = reader.readlines()
|
||||
for index, token in enumerate(tokens):
|
||||
token = token.rstrip("\n")
|
||||
vocab[token] = index
|
||||
return vocab
|
||||
|
||||
|
||||
class XLMProphetNetTokenizer(PreTrainedTokenizer):
|
||||
"""
|
||||
Adapted from :class:`~transfomers.RobertaTokenizer` and class:`~transfomers.XLNetTokenizer`. Based on
|
||||
`SentencePiece <https://github.com/google/sentencepiece>`__.
|
||||
|
||||
This tokenizer inherits from :class:`~transformers.PreTrainedTokenizer` which contains most of the main methods.
|
||||
Users should refer to this superclass for more information regarding those methods.
|
||||
|
||||
Args:
|
||||
vocab_file (:obj:`str`):
|
||||
Path to the vocabulary file.
|
||||
bos_token (:obj:`str`, `optional`, defaults to :obj:`"<s>"`):
|
||||
The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token.
|
||||
|
||||
.. note::
|
||||
|
||||
When building a sequence using special tokens, this is not the token that is used for the beginning
|
||||
of sequence. The token used is the :obj:`cls_token`.
|
||||
eos_token (:obj:`str`, `optional`, defaults to :obj:`"</s>"`):
|
||||
The end of sequence token.
|
||||
|
||||
.. note::
|
||||
|
||||
When building a sequence using special tokens, this is not the token that is used for the end
|
||||
of sequence. The token used is the :obj:`sep_token`.
|
||||
sep_token (:obj:`str`, `optional`, defaults to :obj:`"</s>"`):
|
||||
The separator token, which is used when building a sequence from multiple sequences, e.g. two sequences
|
||||
for sequence classification or for a text and a question for question answering.
|
||||
It is also used as the last token of a sequence built with special tokens.
|
||||
cls_token (:obj:`str`, `optional`, defaults to :obj:`"<s>"`):
|
||||
The classifier token which is used when doing sequence classification (classification of the whole
|
||||
sequence instead of per-token classification). It is the first token of the sequence when built with
|
||||
special tokens.
|
||||
unk_token (:obj:`str`, `optional`, defaults to :obj:`"<unk>"`):
|
||||
The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
|
||||
token instead.
|
||||
pad_token (:obj:`str`, `optional`, defaults to :obj:`"<pad>"`):
|
||||
The token used for padding, for example when batching sequences of different lengths.
|
||||
mask_token (:obj:`str`, `optional`, defaults to :obj:`"<mask>"`):
|
||||
The token used for masking values. This is the token used when training this model with masked language
|
||||
modeling. This is the token which the model will try to predict.
|
||||
additional_special_tokens (:obj:`List[str]`, `optional`, defaults to :obj:`["<s>NOTUSED", "</s>NOTUSED"]`):
|
||||
Additional special tokens used by the tokenizer.
|
||||
|
||||
Attributes:
|
||||
sp_model (:obj:`SentencePieceProcessor`):
|
||||
The `SentencePiece` processor that is used for every conversion (string, tokens and IDs).
|
||||
"""
|
||||
|
||||
vocab_files_names = VOCAB_FILES_NAMES
|
||||
pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
|
||||
max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
|
||||
model_input_names = ["attention_mask"]
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
vocab_file,
|
||||
bos_token="[SEP]",
|
||||
eos_token="[SEP]",
|
||||
sep_token="[SEP]",
|
||||
unk_token="[UNK]",
|
||||
pad_token="[PAD]",
|
||||
cls_token="[CLS]",
|
||||
mask_token="[MASK]",
|
||||
**kwargs
|
||||
):
|
||||
super().__init__(
|
||||
bos_token=bos_token,
|
||||
eos_token=eos_token,
|
||||
unk_token=unk_token,
|
||||
sep_token=sep_token,
|
||||
pad_token=pad_token,
|
||||
mask_token=mask_token,
|
||||
**kwargs,
|
||||
)
|
||||
|
||||
try:
|
||||
import sentencepiece as spm
|
||||
except ImportError:
|
||||
logger.warning(
|
||||
"You need to install SentencePiece to use XLMRobertaTokenizer: https://github.com/google/sentencepiece"
|
||||
"pip install sentencepiece"
|
||||
)
|
||||
raise
|
||||
|
||||
self.sp_model = spm.SentencePieceProcessor()
|
||||
self.sp_model.Load(str(vocab_file))
|
||||
self.vocab_file = vocab_file
|
||||
|
||||
# Original fairseq vocab and spm vocab must be "aligned":
|
||||
# Vocab | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
|
||||
# -------- | ------- | ------- | ------ | ------- | --- | --- | --- | ----- | ----- | ----
|
||||
# fairseq | '<s>' | '<pad>' | '</s>' | '<unk>' | ',' | '.' | '▁' | 's' | '▁de' | '-'
|
||||
# spm | '<unk>' | '<s>' | '</s>' | ',' | '.' | '▁' | 's' | '▁de' | '-' | '▁a'
|
||||
|
||||
# put special tokens and [unused] tokens into the vocab
|
||||
self.fairseq_tokens_to_ids = {"[PAD]": 0, "[CLS]": 1, "[SEP]": 2, "[UNK]": 3, "[MASK]": 4}
|
||||
|
||||
for i in range(10):
|
||||
tok = "[unused{}]".format(i)
|
||||
self.fairseq_tokens_to_ids[tok] = 5 + i
|
||||
|
||||
# The first "real" token "," has position 15 in the embedding vocab and position 3 in the spm vocab
|
||||
self.fairseq_offset = 12
|
||||
self.fairseq_ids_to_tokens = {v: k for k, v in self.fairseq_tokens_to_ids.items()}
|
||||
for k in self.fairseq_tokens_to_ids.keys():
|
||||
self.unique_no_split_tokens.append(k)
|
||||
|
||||
def __getstate__(self):
|
||||
state = self.__dict__.copy()
|
||||
state["sp_model"] = None
|
||||
return state
|
||||
|
||||
def __setstate__(self, d):
|
||||
self.__dict__ = d
|
||||
try:
|
||||
import sentencepiece as spm
|
||||
except ImportError:
|
||||
logger.warning(
|
||||
"You need to install SentencePiece to use XLMRobertaTokenizer: https://github.com/google/sentencepiece"
|
||||
"pip install sentencepiece"
|
||||
)
|
||||
raise
|
||||
self.sp_model = spm.SentencePieceProcessor()
|
||||
self.sp_model.Load(self.vocab_file)
|
||||
|
||||
def get_special_tokens_mask(
|
||||
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None, already_has_special_tokens: bool = False
|
||||
) -> List[int]:
|
||||
"""
|
||||
Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding
|
||||
special tokens using the tokenizer ``prepare_for_model`` method.
|
||||
|
||||
Args:
|
||||
token_ids_0 (:obj:`List[int]`):
|
||||
List of IDs.
|
||||
token_ids_1 (:obj:`List[int]`, `optional`):
|
||||
Optional second list of IDs for sequence pairs.
|
||||
already_has_special_tokens (:obj:`bool`, `optional`, defaults to :obj:`False`):
|
||||
Whether or not the token list is already formatted with special tokens for the model.
|
||||
|
||||
Returns:
|
||||
:obj:`List[int]`: A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.
|
||||
"""
|
||||
|
||||
if already_has_special_tokens:
|
||||
if token_ids_1 is not None:
|
||||
raise ValueError(
|
||||
"You should not supply a second sequence if the provided sequence of "
|
||||
"ids is already formated with special tokens for the model."
|
||||
)
|
||||
return list(map(lambda x: 1 if x in [self.sep_token_id, self.cls_token_id] else 0, token_ids_0))
|
||||
|
||||
if token_ids_1 is None:
|
||||
return ([0] * len(token_ids_0)) + [1]
|
||||
return ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + [1]
|
||||
|
||||
def create_token_type_ids_from_sequences(
|
||||
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
|
||||
) -> List[int]:
|
||||
"""
|
||||
Create a mask from the two sequences passed to be used in a sequence-pair classification task.
|
||||
XLMProphetNet does not make use of token type ids, therefore a list of zeros is returned.
|
||||
|
||||
Args:
|
||||
token_ids_0 (:obj:`List[int]`):
|
||||
List of IDs.
|
||||
token_ids_1 (:obj:`List[int]`, `optional`):
|
||||
Optional second list of IDs for sequence pairs.
|
||||
|
||||
Returns:
|
||||
:obj:`List[int]`: List of zeros.
|
||||
|
||||
"""
|
||||
|
||||
sep = [self.sep_token_id]
|
||||
|
||||
if token_ids_1 is None:
|
||||
return len(token_ids_0 + sep) * [0]
|
||||
return len(token_ids_0 + sep + sep + token_ids_1 + sep) * [0]
|
||||
|
||||
@property
|
||||
def vocab_size(self):
|
||||
return len(self.sp_model) + self.fairseq_offset
|
||||
|
||||
def get_vocab(self):
|
||||
vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
|
||||
vocab.update(self.added_tokens_encoder)
|
||||
return vocab
|
||||
|
||||
def _tokenize(self, text):
|
||||
return self.sp_model.EncodeAsPieces(text)
|
||||
|
||||
def _convert_token_to_id(self, token):
|
||||
""" Converts a token (str) in an id using the vocab. """
|
||||
if token in self.fairseq_tokens_to_ids:
|
||||
return self.fairseq_tokens_to_ids[token]
|
||||
spm_id = self.sp_model.PieceToId(token)
|
||||
|
||||
# Need to return unknown token if the SP model returned 0
|
||||
return spm_id + self.fairseq_offset if spm_id else self.unk_token_id
|
||||
|
||||
def _convert_id_to_token(self, index):
|
||||
"""Converts an index (integer) in a token (str) using the vocab."""
|
||||
if index in self.fairseq_ids_to_tokens:
|
||||
return self.fairseq_ids_to_tokens[index]
|
||||
return self.sp_model.IdToPiece(index - self.fairseq_offset)
|
||||
|
||||
def convert_tokens_to_string(self, tokens):
|
||||
"""Converts a sequence of tokens (strings for sub-words) in a single string."""
|
||||
out_string = "".join(tokens).replace(SPIECE_UNDERLINE, " ").strip()
|
||||
return out_string
|
||||
|
||||
def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
|
||||
if not os.path.isdir(save_directory):
|
||||
logger.error("Vocabulary path ({}) should be a directory".format(save_directory))
|
||||
return
|
||||
out_vocab_file = os.path.join(
|
||||
save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
|
||||
)
|
||||
|
||||
if os.path.abspath(self.vocab_file) != os.path.abspath(out_vocab_file):
|
||||
copyfile(self.vocab_file, out_vocab_file)
|
||||
|
||||
return (out_vocab_file,)
|
||||
|
||||
def build_inputs_with_special_tokens(
|
||||
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
|
||||
) -> List[int]:
|
||||
"""
|
||||
Build model inputs from a sequence or a pair of sequence for sequence classification tasks
|
||||
by concatenating and adding special tokens.
|
||||
A XLMProphetNet sequence has the following format:
|
||||
|
||||
- single sequence: ``X [SEP]``
|
||||
- pair of sequences: ``A [SEP] B [SEP]``
|
||||
|
||||
Args:
|
||||
token_ids_0 (:obj:`List[int]`):
|
||||
List of IDs to which the special tokens will be added
|
||||
token_ids_1 (:obj:`List[int]`, `optional`, defaults to :obj:`None`):
|
||||
Optional second list of IDs for sequence pairs.
|
||||
|
||||
Returns:
|
||||
:obj:`List[int]`: list of `input IDs <../glossary.html#input-ids>`__ with the appropriate special tokens.
|
||||
"""
|
||||
|
||||
if token_ids_1 is None:
|
||||
return token_ids_0 + [self.sep_token_id]
|
||||
sep = [self.sep_token_id]
|
||||
return token_ids_0 + sep + token_ids_1 + sep
|
||||
@ -22,12 +22,13 @@ from typing import List, Optional, Tuple
|
||||
import sentencepiece as spm
|
||||
|
||||
from .tokenization_utils import PreTrainedTokenizer
|
||||
from .tokenization_xlnet import SPIECE_UNDERLINE
|
||||
from .utils import logging
|
||||
|
||||
|
||||
logger = logging.get_logger(__name__)
|
||||
|
||||
SPIECE_UNDERLINE = "▁"
|
||||
|
||||
VOCAB_FILES_NAMES = {"vocab_file": "sentencepiece.bpe.model"}
|
||||
|
||||
PRETRAINED_VOCAB_FILES_MAP = {
|
||||
|
||||
@ -1305,6 +1305,51 @@ class PegasusForConditionalGeneration:
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST = None
|
||||
|
||||
|
||||
class ProphetNetDecoder:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class ProphetNetEncoder:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class ProphetNetForCausalLM:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class ProphetNetForConditionalGeneration:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
@classmethod
|
||||
def from_pretrained(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class ProphetNetModel:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
@classmethod
|
||||
def from_pretrained(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class ProphetNetPreTrainedModel:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
@classmethod
|
||||
def from_pretrained(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class RagModel:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
@ -1706,6 +1751,42 @@ class XLMWithLMHeadModel:
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST = None
|
||||
|
||||
|
||||
class XLMProphetNetDecoder:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class XLMProphetNetEncoder:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class XLMProphetNetForCausalLM:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class XLMProphetNetForConditionalGeneration:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
@classmethod
|
||||
def from_pretrained(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
class XLMProphetNetModel:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
@classmethod
|
||||
def from_pretrained(self, *args, **kwargs):
|
||||
requires_pytorch(self)
|
||||
|
||||
|
||||
XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST = None
|
||||
|
||||
|
||||
|
||||
@ -74,6 +74,15 @@ class T5Tokenizer:
|
||||
requires_sentencepiece(self)
|
||||
|
||||
|
||||
class XLMProphetNetTokenizer:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_sentencepiece(self)
|
||||
|
||||
@classmethod
|
||||
def from_pretrained(self, *args, **kwargs):
|
||||
requires_sentencepiece(self)
|
||||
|
||||
|
||||
class XLMRobertaTokenizer:
|
||||
def __init__(self, *args, **kwargs):
|
||||
requires_sentencepiece(self)
|
||||
|
||||
@ -112,6 +112,7 @@ class ModelTesterMixin:
|
||||
model.eval()
|
||||
with torch.no_grad():
|
||||
outputs = model(**self._prepare_for_class(inputs_dict, model_class))
|
||||
|
||||
out_2 = outputs[0].cpu().numpy()
|
||||
out_2[np.isnan(out_2)] = 0
|
||||
|
||||
@ -152,6 +153,7 @@ class ModelTesterMixin:
|
||||
with torch.no_grad():
|
||||
first = model(**self._prepare_for_class(inputs_dict, model_class))[0]
|
||||
second = model(**self._prepare_for_class(inputs_dict, model_class))[0]
|
||||
|
||||
out_1 = first.cpu().numpy()
|
||||
out_2 = second.cpu().numpy()
|
||||
out_1 = out_1[~np.isnan(out_1)]
|
||||
@ -183,10 +185,12 @@ class ModelTesterMixin:
|
||||
|
||||
def test_attention_outputs(self):
|
||||
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
|
||||
config.return_dict = True
|
||||
|
||||
seq_len = getattr(self.model_tester, "seq_length", None)
|
||||
decoder_seq_length = getattr(self.model_tester, "decoder_seq_length", seq_len)
|
||||
encoder_seq_length = getattr(self.model_tester, "encoder_seq_length", seq_len)
|
||||
decoder_key_length = getattr(self.model_tester, "key_length", decoder_seq_length)
|
||||
decoder_key_length = getattr(self.model_tester, "decoder_key_length", decoder_seq_length)
|
||||
encoder_key_length = getattr(self.model_tester, "key_length", encoder_seq_length)
|
||||
chunk_length = getattr(self.model_tester, "chunk_length", None)
|
||||
if chunk_length is not None and hasattr(self.model_tester, "num_hashes"):
|
||||
@ -211,7 +215,7 @@ class ModelTesterMixin:
|
||||
model.eval()
|
||||
with torch.no_grad():
|
||||
outputs = model(**self._prepare_for_class(inputs_dict, model_class), return_dict=True)
|
||||
attentions = outputs[-1]
|
||||
attentions = outputs["attentions"] if "attentions" in outputs.keys() else outputs[-1]
|
||||
self.assertEqual(len(attentions), self.model_tester.num_hidden_layers)
|
||||
|
||||
if chunk_length is not None:
|
||||
@ -227,8 +231,14 @@ class ModelTesterMixin:
|
||||
out_len = len(outputs)
|
||||
|
||||
if self.is_encoder_decoder:
|
||||
correct_outlen = 4
|
||||
decoder_attention_idx = 1
|
||||
correct_outlen = (
|
||||
self.model_tester.base_model_out_len if hasattr(self.model_tester, "base_model_out_len") else 4
|
||||
)
|
||||
decoder_attention_idx = (
|
||||
self.model_tester.decoder_attention_idx
|
||||
if hasattr(self.model_tester, "decoder_attention_idx")
|
||||
else 1
|
||||
)
|
||||
|
||||
# loss is at first position
|
||||
if "labels" in inputs_dict:
|
||||
@ -238,6 +248,7 @@ class ModelTesterMixin:
|
||||
if model_class in MODEL_FOR_QUESTION_ANSWERING_MAPPING.values():
|
||||
correct_outlen += 1 # start_logits and end_logits instead of only 1 output
|
||||
decoder_attention_idx += 1
|
||||
|
||||
self.assertEqual(out_len, correct_outlen)
|
||||
|
||||
decoder_attentions = outputs[decoder_attention_idx]
|
||||
@ -256,9 +267,16 @@ class ModelTesterMixin:
|
||||
model.eval()
|
||||
with torch.no_grad():
|
||||
outputs = model(**self._prepare_for_class(inputs_dict, model_class))
|
||||
self.assertEqual(out_len + (2 if self.is_encoder_decoder else 1), len(outputs))
|
||||
|
||||
self_attentions = outputs[-1]
|
||||
if hasattr(self.model_tester, "num_hidden_states_types"):
|
||||
added_hidden_states = self.model_tester.num_hidden_states_types
|
||||
elif self.is_encoder_decoder:
|
||||
added_hidden_states = 2
|
||||
else:
|
||||
added_hidden_states = 1
|
||||
self.assertEqual(out_len + added_hidden_states, len(outputs))
|
||||
|
||||
self_attentions = outputs["attentions"] if "attentions" in outputs else outputs[-1]
|
||||
self.assertEqual(len(self_attentions), self.model_tester.num_hidden_layers)
|
||||
if chunk_length is not None:
|
||||
self.assertListEqual(
|
||||
@ -571,8 +589,8 @@ class ModelTesterMixin:
|
||||
model.eval()
|
||||
|
||||
with torch.no_grad():
|
||||
outputs = model(**self._prepare_for_class(inputs_dict, model_class))
|
||||
hidden_states = outputs[-1]
|
||||
outputs = model(**self._prepare_for_class(inputs_dict, model_class), return_dict=True)
|
||||
hidden_states = outputs["hidden_states"] if "hidden_states" in outputs else outputs[-1]
|
||||
|
||||
expected_num_layers = getattr(
|
||||
self.model_tester, "expected_num_hidden_layers", self.model_tester.num_hidden_layers + 1
|
||||
@ -822,6 +840,7 @@ class ModelTesterMixin:
|
||||
model.eval()
|
||||
|
||||
inputs = copy.deepcopy(self._prepare_for_class(inputs_dict, model_class))
|
||||
|
||||
if not self.is_encoder_decoder:
|
||||
input_ids = inputs["input_ids"]
|
||||
del inputs["input_ids"]
|
||||
@ -839,7 +858,7 @@ class ModelTesterMixin:
|
||||
inputs["decoder_inputs_embeds"] = wte(decoder_input_ids)
|
||||
|
||||
with torch.no_grad():
|
||||
model(**inputs)
|
||||
model(**inputs)[0]
|
||||
|
||||
def test_lm_head_model_random_no_beam_search_generate(self):
|
||||
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
|
||||
|
||||
@ -24,6 +24,7 @@ from .test_modeling_bert import BertModelTester
|
||||
from .test_modeling_bert_generation import BertGenerationEncoderTester
|
||||
from .test_modeling_common import ids_tensor
|
||||
from .test_modeling_gpt2 import GPT2ModelTester
|
||||
from .test_modeling_prophetnet import ProphetNetStandaloneDecoderModelTester
|
||||
from .test_modeling_roberta import RobertaModelTester
|
||||
|
||||
|
||||
@ -41,9 +42,11 @@ if is_torch_available():
|
||||
EncoderDecoderConfig,
|
||||
EncoderDecoderModel,
|
||||
GPT2LMHeadModel,
|
||||
ProphetNetForCausalLM,
|
||||
RobertaForCausalLM,
|
||||
RobertaModel,
|
||||
)
|
||||
from transformers.modeling_outputs import BaseModelOutput
|
||||
|
||||
|
||||
@require_torch
|
||||
@ -82,10 +85,15 @@ class EncoderDecoderMixin:
|
||||
decoder_input_ids=decoder_input_ids,
|
||||
attention_mask=attention_mask,
|
||||
decoder_attention_mask=decoder_attention_mask,
|
||||
return_dict=True,
|
||||
)
|
||||
|
||||
self.assertEqual(outputs_encoder_decoder[0].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,)))
|
||||
self.assertEqual(outputs_encoder_decoder[1].shape, (input_ids.shape + (config.hidden_size,)))
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["logits"].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,))
|
||||
)
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["encoder_last_hidden_state"].shape, (input_ids.shape + (config.hidden_size,))
|
||||
)
|
||||
|
||||
def check_encoder_decoder_model(
|
||||
self,
|
||||
@ -109,20 +117,30 @@ class EncoderDecoderMixin:
|
||||
decoder_input_ids=decoder_input_ids,
|
||||
attention_mask=attention_mask,
|
||||
decoder_attention_mask=decoder_attention_mask,
|
||||
return_dict=True,
|
||||
)
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["logits"].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,))
|
||||
)
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["encoder_last_hidden_state"].shape, (input_ids.shape + (config.hidden_size,))
|
||||
)
|
||||
|
||||
self.assertEqual(outputs_encoder_decoder[0].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,)))
|
||||
self.assertEqual(outputs_encoder_decoder[1].shape, (input_ids.shape + (config.hidden_size,)))
|
||||
encoder_outputs = (encoder_hidden_states,)
|
||||
encoder_outputs = BaseModelOutput(last_hidden_state=encoder_hidden_states)
|
||||
outputs_encoder_decoder = enc_dec_model(
|
||||
encoder_outputs=encoder_outputs,
|
||||
decoder_input_ids=decoder_input_ids,
|
||||
attention_mask=attention_mask,
|
||||
decoder_attention_mask=decoder_attention_mask,
|
||||
return_dict=True,
|
||||
)
|
||||
|
||||
self.assertEqual(outputs_encoder_decoder[0].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,)))
|
||||
self.assertEqual(outputs_encoder_decoder[1].shape, (input_ids.shape + (config.hidden_size,)))
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["logits"].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,))
|
||||
)
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["encoder_last_hidden_state"].shape, (input_ids.shape + (config.hidden_size,))
|
||||
)
|
||||
|
||||
def check_encoder_decoder_model_from_pretrained(
|
||||
self,
|
||||
@ -145,10 +163,15 @@ class EncoderDecoderMixin:
|
||||
decoder_input_ids=decoder_input_ids,
|
||||
attention_mask=attention_mask,
|
||||
decoder_attention_mask=decoder_attention_mask,
|
||||
return_dict=True,
|
||||
)
|
||||
|
||||
self.assertEqual(outputs_encoder_decoder[0].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,)))
|
||||
self.assertEqual(outputs_encoder_decoder[1].shape, (input_ids.shape + (config.hidden_size,)))
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["logits"].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,))
|
||||
)
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["encoder_last_hidden_state"].shape, (input_ids.shape + (config.hidden_size,))
|
||||
)
|
||||
|
||||
def check_save_and_load(
|
||||
self,
|
||||
@ -255,14 +278,19 @@ class EncoderDecoderMixin:
|
||||
attention_mask=attention_mask,
|
||||
decoder_attention_mask=decoder_attention_mask,
|
||||
labels=labels,
|
||||
return_dict=True,
|
||||
)
|
||||
|
||||
mlm_loss = outputs_encoder_decoder[0]
|
||||
loss = outputs_encoder_decoder["loss"]
|
||||
# check that backprop works
|
||||
mlm_loss.backward()
|
||||
loss.backward()
|
||||
|
||||
self.assertEqual(outputs_encoder_decoder[1].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,)))
|
||||
self.assertEqual(outputs_encoder_decoder[2].shape, (input_ids.shape + (config.hidden_size,)))
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["logits"].shape, (decoder_input_ids.shape + (decoder_config.vocab_size,))
|
||||
)
|
||||
self.assertEqual(
|
||||
outputs_encoder_decoder["encoder_last_hidden_state"].shape, (input_ids.shape + (config.hidden_size,))
|
||||
)
|
||||
|
||||
def check_encoder_decoder_model_generate(self, input_ids, config, decoder_config, **kwargs):
|
||||
encoder_model, decoder_model = self.get_encoder_decoder_model(config, decoder_config)
|
||||
@ -425,6 +453,7 @@ class EncoderDecoderMixin:
|
||||
self.assertLessEqual(max_diff, 1e-5)
|
||||
|
||||
|
||||
@require_torch
|
||||
class BertEncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCase):
|
||||
def get_pretrained_model(self):
|
||||
return EncoderDecoderModel.from_encoder_decoder_pretrained("bert-base-cased", "bert-base-cased")
|
||||
@ -493,6 +522,7 @@ class BertEncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCase):
|
||||
self.assertEqual(summary, EXPECTED_SUMMARY)
|
||||
|
||||
|
||||
@require_torch
|
||||
class BertGenerationEncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCase):
|
||||
def get_pretrained_model(self):
|
||||
return EncoderDecoderModel.from_encoder_decoder_pretrained(
|
||||
@ -554,6 +584,7 @@ class BertGenerationEncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCa
|
||||
self.assertEqual(summary, EXPECTED_SUMMARY)
|
||||
|
||||
|
||||
@require_torch
|
||||
class RoBertaEncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCase):
|
||||
def get_encoder_decoder_model(self, config, decoder_config):
|
||||
encoder_model = RobertaModel(config)
|
||||
@ -606,6 +637,7 @@ class RoBertaEncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCase):
|
||||
return EncoderDecoderModel.from_encoder_decoder_pretrained("roberta-base", "roberta-base")
|
||||
|
||||
|
||||
@require_torch
|
||||
class GPT2EncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCase):
|
||||
def get_encoder_decoder_model(self, config, decoder_config):
|
||||
encoder_model = BertModel(config)
|
||||
@ -663,3 +695,59 @@ class GPT2EncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCase):
|
||||
|
||||
def test_encoder_decoder_model_shared_weights(self):
|
||||
pass
|
||||
|
||||
|
||||
@require_torch
|
||||
class ProphetNetEncoderDecoderModelTest(EncoderDecoderMixin, unittest.TestCase):
|
||||
def get_encoder_decoder_model(self, config, decoder_config):
|
||||
encoder_model = BertModel(config)
|
||||
decoder_model = ProphetNetForCausalLM(decoder_config)
|
||||
return encoder_model, decoder_model
|
||||
|
||||
def prepare_config_and_inputs(self):
|
||||
model_tester_encoder = BertModelTester(self, batch_size=13)
|
||||
model_tester_decoder = ProphetNetStandaloneDecoderModelTester(
|
||||
self, batch_size=13, hidden_size=32, max_position_embeddings=512
|
||||
)
|
||||
encoder_config_and_inputs = model_tester_encoder.prepare_config_and_inputs()
|
||||
decoder_config_and_inputs = model_tester_decoder.prepare_config_and_inputs_for_decoder()
|
||||
(
|
||||
config,
|
||||
input_ids,
|
||||
token_type_ids,
|
||||
input_mask,
|
||||
sequence_labels,
|
||||
token_labels,
|
||||
choice_labels,
|
||||
) = encoder_config_and_inputs
|
||||
(
|
||||
decoder_config,
|
||||
decoder_input_ids,
|
||||
decoder_attention_mask,
|
||||
encoder_hidden_states,
|
||||
encoder_attention_mask,
|
||||
lm_labels,
|
||||
) = decoder_config_and_inputs
|
||||
|
||||
# make sure that cross attention layers are added
|
||||
decoder_config.add_cross_attention = True
|
||||
# disable cache for now
|
||||
decoder_config.use_cache = False
|
||||
return {
|
||||
"config": config,
|
||||
"input_ids": input_ids,
|
||||
"attention_mask": input_mask,
|
||||
"decoder_config": decoder_config,
|
||||
"decoder_input_ids": decoder_input_ids,
|
||||
"decoder_attention_mask": decoder_attention_mask,
|
||||
"encoder_hidden_states": encoder_hidden_states,
|
||||
"labels": lm_labels,
|
||||
}
|
||||
|
||||
def get_pretrained_model(self):
|
||||
return EncoderDecoderModel.from_encoder_decoder_pretrained(
|
||||
"bert-large-uncased", "patrickvonplaten/prophetnet-decoder-clm-large-uncased"
|
||||
)
|
||||
|
||||
def test_encoder_decoder_model_shared_weights(self):
|
||||
pass
|
||||
|
||||
1075
tests/test_modeling_prophetnet.py
Normal file
1075
tests/test_modeling_prophetnet.py
Normal file
File diff suppressed because it is too large
Load Diff
@ -43,7 +43,6 @@ class T5ModelTester:
|
||||
encoder_seq_length=7,
|
||||
decoder_seq_length=9,
|
||||
# For common tests
|
||||
seq_length=7,
|
||||
is_training=True,
|
||||
use_attention_mask=True,
|
||||
use_labels=True,
|
||||
|
||||
141
tests/test_modeling_xlm_prophetnet.py
Normal file
141
tests/test_modeling_xlm_prophetnet.py
Normal file
@ -0,0 +1,141 @@
|
||||
# coding=utf-8
|
||||
# Copyright 2020 The HuggingFace Inc. team, The Microsoft Research 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.
|
||||
|
||||
|
||||
import unittest
|
||||
|
||||
from transformers import is_torch_available
|
||||
from transformers.testing_utils import slow, torch_device
|
||||
|
||||
|
||||
if is_torch_available():
|
||||
import torch
|
||||
|
||||
from transformers import XLMProphetNetForConditionalGeneration, XLMProphetNetTokenizer
|
||||
|
||||
|
||||
class XLMProphetNetModelIntegrationTest(unittest.TestCase):
|
||||
@slow
|
||||
def test_pretrained_checkpoint_hidden_states(self):
|
||||
model = XLMProphetNetForConditionalGeneration.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
|
||||
model.to(torch_device)
|
||||
|
||||
# encoder-decoder outputs
|
||||
encoder_ids = torch.tensor([[17, 96208, 103471, 2]]).to(torch_device)
|
||||
decoder_prev_ids = torch.tensor(
|
||||
[[2, 250, 9953, 34, 69489, 1620, 32, 118424, 624, 210, 105, 2913, 1032, 351]]
|
||||
).to(torch_device)
|
||||
output = model(
|
||||
input_ids=encoder_ids, attention_mask=None, encoder_outputs=None, decoder_input_ids=decoder_prev_ids
|
||||
)
|
||||
output_predited_logis = output[0]
|
||||
expected_shape = torch.Size((1, 14, 250012))
|
||||
self.assertEqual(output_predited_logis.shape, expected_shape)
|
||||
expected_slice = torch.tensor(
|
||||
[[[-6.6042, -8.3838, 12.4717], [-6.4426, -8.1994, 12.4542], [-6.0851, -7.8209, 12.9493]]]
|
||||
).to(torch_device)
|
||||
self.assertTrue(torch.allclose(output_predited_logis[:, :3, :3], expected_slice, atol=1e-4))
|
||||
|
||||
# encoder outputs
|
||||
encoder_outputs = model.prophetnet.encoder(encoder_ids)[0]
|
||||
expected_encoder_outputs_slice = torch.tensor(
|
||||
[[[-1.4260, -0.7628, 0.8453], [-1.4719, -0.1391, 0.7807], [-1.7678, 0.0114, 0.4646]]]
|
||||
).to(torch_device)
|
||||
expected_shape_encoder = torch.Size((1, 4, 1024))
|
||||
self.assertEqual(encoder_outputs.shape, expected_shape_encoder)
|
||||
self.assertTrue(torch.allclose(encoder_outputs[:, :3, :3], expected_encoder_outputs_slice, atol=1e-4))
|
||||
|
||||
# decoder outputs
|
||||
decoder_outputs = model.prophetnet.decoder(
|
||||
decoder_prev_ids,
|
||||
encoder_hidden_states=encoder_outputs,
|
||||
)
|
||||
predicting_streams = decoder_outputs[1].view(1, model.config.ngram, 14, -1)
|
||||
predicting_streams_logits = model.lm_head(predicting_streams)
|
||||
next_first_stream_logits = predicting_streams_logits[:, 0]
|
||||
self.assertTrue(torch.allclose(next_first_stream_logits[:, :3, :3], expected_slice, atol=1e-4))
|
||||
|
||||
@slow
|
||||
def test_ntg_hidden_states(self):
|
||||
model = XLMProphetNetForConditionalGeneration.from_pretrained(
|
||||
"microsoft/xprophetnet-large-wiki100-cased-xglue-ntg"
|
||||
)
|
||||
model.to(torch_device)
|
||||
|
||||
encoder_ids = torch.tensor([[17, 96208, 103471, 2]]).to(torch_device)
|
||||
decoder_prev_ids = torch.tensor(
|
||||
[[2, 250, 9953, 34, 69489, 1620, 32, 118424, 624, 210, 105, 2913, 1032, 351]]
|
||||
).to(torch_device)
|
||||
output = model(
|
||||
input_ids=encoder_ids, attention_mask=None, encoder_outputs=None, decoder_input_ids=decoder_prev_ids
|
||||
)
|
||||
output_predited_logis = output[0]
|
||||
expected_shape = torch.Size((1, 14, 250012))
|
||||
self.assertEqual(output_predited_logis.shape, expected_shape)
|
||||
# compare the actual values for a slice.
|
||||
expected_slice = torch.tensor(
|
||||
[[[-8.8815, -9.2996, -4.4506], [-6.7202, -7.8944, -0.9402], [-8.6890, -7.4528, -1.9437]]]
|
||||
).to(torch_device)
|
||||
|
||||
self.assertTrue(torch.allclose(output_predited_logis[:, :3, :3], expected_slice, atol=1e-4))
|
||||
|
||||
@slow
|
||||
def test_xprophetnet_ntg_inference(self):
|
||||
model = XLMProphetNetForConditionalGeneration.from_pretrained(
|
||||
"microsoft/xprophetnet-large-wiki100-cased-xglue-ntg"
|
||||
)
|
||||
model.to(torch_device)
|
||||
model.config.max_length = 512
|
||||
|
||||
tokenizer = XLMProphetNetTokenizer.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。"
|
||||
)
|
||||
|
||||
input_ids = tokenizer(
|
||||
[EN_SENTENCE, RU_SENTENCE, ZH_SENTENCE], padding=True, max_length=255, return_tensors="pt"
|
||||
).input_ids
|
||||
input_ids = input_ids.to(torch_device)
|
||||
|
||||
summary_ids = model.generate(
|
||||
input_ids, num_beams=10, length_penalty=1.0, no_repeat_ngram_size=3, early_stopping=True
|
||||
)
|
||||
generated_titles = [tokenizer.decode(g, skip_special_tokens=True) for g in summary_ids]
|
||||
EXPECTED_TITLE_EN = "Microsoft to end Windows 7 free support after January 14, 2020"
|
||||
EXPECTED_TITLE_RU = "Microsoft намерена прекратить бесплатную поддержку Windows 7 после 14 января 2020 года"
|
||||
EXPECTED_TITLE_ZH = "微软打算终止对Windows 7操作系统的免费支持"
|
||||
self.assertListEqual(
|
||||
[EXPECTED_TITLE_EN, EXPECTED_TITLE_RU, EXPECTED_TITLE_ZH],
|
||||
generated_titles,
|
||||
)
|
||||
|
||||
summary_ids_beam1 = model.generate(
|
||||
input_ids, num_beams=1, length_penalty=1.0, no_repeat_ngram_size=3, early_stopping=True
|
||||
)
|
||||
generated_titles_beam1_tok = [
|
||||
tokenizer.convert_ids_to_tokens(g, skip_special_tokens=True) for g in summary_ids_beam1
|
||||
]
|
||||
EXPECTED_TITLE_EN_BEAM1_TOK = "▁Microsoft ▁to ▁end ▁free ▁support ▁for ▁Windows ▁7".split(" ")
|
||||
EXPECTED_TITLE_RU_BEAM1_TOK = "▁Microsoft ▁намерен а ▁прекрати ть ▁бес плат ную ▁поддержку ▁Windows ▁7 ▁после ▁14 ▁января ▁2020 ▁года".split(
|
||||
" "
|
||||
)
|
||||
EXPECTED_TITLE_ZH_BEAM1_TOK = "微软 公司 打算 终止 对 Windows ▁7 操作 系统的 免费 支持".split(" ")
|
||||
self.assertListEqual(
|
||||
[EXPECTED_TITLE_EN_BEAM1_TOK, EXPECTED_TITLE_RU_BEAM1_TOK, EXPECTED_TITLE_ZH_BEAM1_TOK],
|
||||
generated_titles_beam1_tok,
|
||||
)
|
||||
191
tests/test_tokenization_prophetnet.py
Normal file
191
tests/test_tokenization_prophetnet.py
Normal file
@ -0,0 +1,191 @@
|
||||
# coding=utf-8
|
||||
# Copyright 2020 The HuggingFace Inc. team, The Microsoft Research 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.
|
||||
|
||||
|
||||
import os
|
||||
import unittest
|
||||
|
||||
from transformers.testing_utils import slow
|
||||
from transformers.tokenization_bert import (
|
||||
BasicTokenizer,
|
||||
WordpieceTokenizer,
|
||||
_is_control,
|
||||
_is_punctuation,
|
||||
_is_whitespace,
|
||||
)
|
||||
from transformers.tokenization_prophetnet import VOCAB_FILES_NAMES, ProphetNetTokenizer
|
||||
|
||||
from .test_tokenization_common import TokenizerTesterMixin
|
||||
|
||||
|
||||
class ProphetNetTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
|
||||
|
||||
tokenizer_class = ProphetNetTokenizer
|
||||
test_rust_tokenizer = False
|
||||
|
||||
def setUp(self):
|
||||
super().setUp()
|
||||
|
||||
vocab_tokens = [
|
||||
"[UNK]",
|
||||
"[CLS]",
|
||||
"[SEP]",
|
||||
"[PAD]",
|
||||
"[MASK]",
|
||||
"want",
|
||||
"##want",
|
||||
"##ed",
|
||||
"wa",
|
||||
"un",
|
||||
"runn",
|
||||
"##ing",
|
||||
",",
|
||||
"low",
|
||||
"lowest",
|
||||
]
|
||||
self.vocab_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["vocab_file"])
|
||||
with open(self.vocab_file, "w", encoding="utf-8") as vocab_writer:
|
||||
vocab_writer.write("".join([x + "\n" for x in vocab_tokens]))
|
||||
|
||||
def get_input_output_texts(self, tokenizer):
|
||||
input_text = "UNwant\u00E9d,running"
|
||||
output_text = "unwanted, running"
|
||||
return input_text, output_text
|
||||
|
||||
def test_full_tokenizer(self):
|
||||
tokenizer = self.tokenizer_class(self.vocab_file)
|
||||
|
||||
tokens = tokenizer.tokenize("UNwant\u00E9d,running")
|
||||
self.assertListEqual(tokens, ["un", "##want", "##ed", ",", "runn", "##ing"])
|
||||
self.assertListEqual(tokenizer.convert_tokens_to_ids(tokens), [9, 6, 7, 12, 10, 11])
|
||||
|
||||
def test_chinese(self):
|
||||
tokenizer = BasicTokenizer()
|
||||
|
||||
self.assertListEqual(tokenizer.tokenize("ah\u535A\u63A8zz"), ["ah", "\u535A", "\u63A8", "zz"])
|
||||
|
||||
def test_basic_tokenizer_lower(self):
|
||||
tokenizer = BasicTokenizer(do_lower_case=True)
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.tokenize(" \tHeLLo!how \n Are yoU? "), ["hello", "!", "how", "are", "you", "?"]
|
||||
)
|
||||
self.assertListEqual(tokenizer.tokenize("H\u00E9llo"), ["hello"])
|
||||
|
||||
def test_basic_tokenizer_lower_strip_accents_false(self):
|
||||
tokenizer = BasicTokenizer(do_lower_case=True, strip_accents=False)
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["hällo", "!", "how", "are", "you", "?"]
|
||||
)
|
||||
self.assertListEqual(tokenizer.tokenize("H\u00E9llo"), ["h\u00E9llo"])
|
||||
|
||||
def test_basic_tokenizer_lower_strip_accents_true(self):
|
||||
tokenizer = BasicTokenizer(do_lower_case=True, strip_accents=True)
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["hallo", "!", "how", "are", "you", "?"]
|
||||
)
|
||||
self.assertListEqual(tokenizer.tokenize("H\u00E9llo"), ["hello"])
|
||||
|
||||
def test_basic_tokenizer_lower_strip_accents_default(self):
|
||||
tokenizer = BasicTokenizer(do_lower_case=True)
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["hallo", "!", "how", "are", "you", "?"]
|
||||
)
|
||||
self.assertListEqual(tokenizer.tokenize("H\u00E9llo"), ["hello"])
|
||||
|
||||
def test_basic_tokenizer_no_lower(self):
|
||||
tokenizer = BasicTokenizer(do_lower_case=False)
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.tokenize(" \tHeLLo!how \n Are yoU? "), ["HeLLo", "!", "how", "Are", "yoU", "?"]
|
||||
)
|
||||
|
||||
def test_basic_tokenizer_no_lower_strip_accents_false(self):
|
||||
tokenizer = BasicTokenizer(do_lower_case=False, strip_accents=False)
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["HäLLo", "!", "how", "Are", "yoU", "?"]
|
||||
)
|
||||
|
||||
def test_basic_tokenizer_no_lower_strip_accents_true(self):
|
||||
tokenizer = BasicTokenizer(do_lower_case=False, strip_accents=True)
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["HaLLo", "!", "how", "Are", "yoU", "?"]
|
||||
)
|
||||
|
||||
def test_basic_tokenizer_respects_never_split_tokens(self):
|
||||
tokenizer = BasicTokenizer(do_lower_case=False, never_split=["[UNK]"])
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.tokenize(" \tHeLLo!how \n Are yoU? [UNK]"), ["HeLLo", "!", "how", "Are", "yoU", "?", "[UNK]"]
|
||||
)
|
||||
|
||||
def test_wordpiece_tokenizer(self):
|
||||
vocab_tokens = ["[UNK]", "[CLS]", "[SEP]", "want", "##want", "##ed", "wa", "un", "runn", "##ing"]
|
||||
|
||||
vocab = {}
|
||||
for (i, token) in enumerate(vocab_tokens):
|
||||
vocab[token] = i
|
||||
tokenizer = WordpieceTokenizer(vocab=vocab, unk_token="[UNK]")
|
||||
|
||||
self.assertListEqual(tokenizer.tokenize(""), [])
|
||||
|
||||
self.assertListEqual(tokenizer.tokenize("unwanted running"), ["un", "##want", "##ed", "runn", "##ing"])
|
||||
|
||||
self.assertListEqual(tokenizer.tokenize("unwantedX running"), ["[UNK]", "runn", "##ing"])
|
||||
|
||||
def test_is_whitespace(self):
|
||||
self.assertTrue(_is_whitespace(" "))
|
||||
self.assertTrue(_is_whitespace("\t"))
|
||||
self.assertTrue(_is_whitespace("\r"))
|
||||
self.assertTrue(_is_whitespace("\n"))
|
||||
self.assertTrue(_is_whitespace("\u00A0"))
|
||||
|
||||
self.assertFalse(_is_whitespace("A"))
|
||||
self.assertFalse(_is_whitespace("-"))
|
||||
|
||||
def test_is_control(self):
|
||||
self.assertTrue(_is_control("\u0005"))
|
||||
|
||||
self.assertFalse(_is_control("A"))
|
||||
self.assertFalse(_is_control(" "))
|
||||
self.assertFalse(_is_control("\t"))
|
||||
self.assertFalse(_is_control("\r"))
|
||||
|
||||
def test_is_punctuation(self):
|
||||
self.assertTrue(_is_punctuation("-"))
|
||||
self.assertTrue(_is_punctuation("$"))
|
||||
self.assertTrue(_is_punctuation("`"))
|
||||
self.assertTrue(_is_punctuation("."))
|
||||
|
||||
self.assertFalse(_is_punctuation("A"))
|
||||
self.assertFalse(_is_punctuation(" "))
|
||||
|
||||
@slow
|
||||
def test_sequence_builders(self):
|
||||
tokenizer = self.tokenizer_class.from_pretrained("microsoft/prophetnet-large-uncased")
|
||||
|
||||
text = tokenizer.encode("sequence builders", add_special_tokens=False)
|
||||
text_2 = tokenizer.encode("multi-sequence build", add_special_tokens=False)
|
||||
|
||||
encoded_sentence = tokenizer.build_inputs_with_special_tokens(text)
|
||||
encoded_pair = tokenizer.build_inputs_with_special_tokens(text, text_2)
|
||||
|
||||
assert encoded_sentence == text + [102]
|
||||
assert encoded_pair == text + [102] + text_2 + [102]
|
||||
125
tests/test_tokenization_xlm_prophetnet.py
Normal file
125
tests/test_tokenization_xlm_prophetnet.py
Normal file
@ -0,0 +1,125 @@
|
||||
# coding=utf-8
|
||||
# Copyright 2020 The HuggingFace Inc. team, The Microsoft Research 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.
|
||||
|
||||
|
||||
import os
|
||||
import unittest
|
||||
|
||||
from transformers.file_utils import cached_property
|
||||
from transformers.testing_utils import slow
|
||||
from transformers.tokenization_xlm_prophetnet import SPIECE_UNDERLINE, XLMProphetNetTokenizer
|
||||
|
||||
from .test_tokenization_common import TokenizerTesterMixin
|
||||
|
||||
|
||||
SAMPLE_VOCAB = os.path.join(os.path.dirname(os.path.abspath(__file__)), "fixtures/test_sentencepiece.model")
|
||||
|
||||
|
||||
class XLMProphetNetTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
|
||||
|
||||
tokenizer_class = XLMProphetNetTokenizer
|
||||
test_rust_tokenizer = False
|
||||
|
||||
def setUp(self):
|
||||
super().setUp()
|
||||
|
||||
# We have a SentencePiece fixture for testing
|
||||
tokenizer = XLMProphetNetTokenizer(SAMPLE_VOCAB, keep_accents=True)
|
||||
tokenizer.save_pretrained(self.tmpdirname)
|
||||
|
||||
def test_full_tokenizer(self):
|
||||
tokenizer = XLMProphetNetTokenizer(SAMPLE_VOCAB, keep_accents=True)
|
||||
|
||||
tokens = tokenizer.tokenize("This is a test")
|
||||
self.assertListEqual(tokens, ["▁This", "▁is", "▁a", "▁t", "est"])
|
||||
|
||||
self.assertListEqual(
|
||||
tokenizer.convert_tokens_to_ids(tokens),
|
||||
[value + tokenizer.fairseq_offset for value in [285, 46, 10, 170, 382]],
|
||||
)
|
||||
|
||||
tokens = tokenizer.tokenize("I was born in 92000, and this is falsé.")
|
||||
self.assertListEqual(
|
||||
tokens,
|
||||
[
|
||||
SPIECE_UNDERLINE + "I",
|
||||
SPIECE_UNDERLINE + "was",
|
||||
SPIECE_UNDERLINE + "b",
|
||||
"or",
|
||||
"n",
|
||||
SPIECE_UNDERLINE + "in",
|
||||
SPIECE_UNDERLINE + "",
|
||||
"9",
|
||||
"2",
|
||||
"0",
|
||||
"0",
|
||||
"0",
|
||||
",",
|
||||
SPIECE_UNDERLINE + "and",
|
||||
SPIECE_UNDERLINE + "this",
|
||||
SPIECE_UNDERLINE + "is",
|
||||
SPIECE_UNDERLINE + "f",
|
||||
"al",
|
||||
"s",
|
||||
"é",
|
||||
".",
|
||||
],
|
||||
)
|
||||
ids = tokenizer.convert_tokens_to_ids(tokens)
|
||||
self.assertListEqual(
|
||||
ids,
|
||||
[
|
||||
value + tokenizer.fairseq_offset
|
||||
for value in [8, 21, 84, 55, 24, 19, 7, -9, 602, 347, 347, 347, 3, 12, 66, 46, 72, 80, 6, -9, 4]
|
||||
],
|
||||
)
|
||||
|
||||
back_tokens = tokenizer.convert_ids_to_tokens(ids)
|
||||
self.assertListEqual(
|
||||
back_tokens,
|
||||
[
|
||||
SPIECE_UNDERLINE + "I",
|
||||
SPIECE_UNDERLINE + "was",
|
||||
SPIECE_UNDERLINE + "b",
|
||||
"or",
|
||||
"n",
|
||||
SPIECE_UNDERLINE + "in",
|
||||
SPIECE_UNDERLINE + "",
|
||||
"[UNK]",
|
||||
"2",
|
||||
"0",
|
||||
"0",
|
||||
"0",
|
||||
",",
|
||||
SPIECE_UNDERLINE + "and",
|
||||
SPIECE_UNDERLINE + "this",
|
||||
SPIECE_UNDERLINE + "is",
|
||||
SPIECE_UNDERLINE + "f",
|
||||
"al",
|
||||
"s",
|
||||
"[UNK]",
|
||||
".",
|
||||
],
|
||||
)
|
||||
|
||||
@cached_property
|
||||
def big_tokenizer(self):
|
||||
return XLMProphetNetTokenizer.from_pretrained("microsoft/xprophetnet-large-wiki100-cased")
|
||||
|
||||
@slow
|
||||
def test_tokenization_base_easy_symbols(self):
|
||||
symbols = "Hello World!"
|
||||
original_tokenizer_encodings = [35389, 6672, 49, 2]
|
||||
self.assertListEqual(original_tokenizer_encodings, self.big_tokenizer.encode(symbols))
|
||||
@ -43,6 +43,7 @@ TEST_FILES_WITH_NO_COMMON_TESTS = [
|
||||
"test_modeling_camembert.py",
|
||||
"test_modeling_tf_camembert.py",
|
||||
"test_modeling_tf_xlm_roberta.py",
|
||||
"test_modeling_xlm_prophetnet.py",
|
||||
"test_modeling_xlm_roberta.py",
|
||||
"test_modeling_pegasus.py",
|
||||
"test_modeling_mbart.py",
|
||||
@ -61,6 +62,7 @@ IGNORE_NON_DOCUMENTED = [
|
||||
MODEL_NAME_TO_DOC_FILE = {
|
||||
"openai": "gpt.rst",
|
||||
"transfo_xl": "transformerxl.rst",
|
||||
"xlm_prophetnet": "xlmprophetnet.rst",
|
||||
"xlm_roberta": "xlmroberta.rst",
|
||||
"bert_generation": "bertgeneration.rst",
|
||||
}
|
||||
@ -243,6 +245,9 @@ def _get_model_name(module):
|
||||
# Secial case for transfo_xl
|
||||
if splits[-1] == "xl":
|
||||
return "_".join(splits[-2:])
|
||||
# Special case for xlm_prophetnet
|
||||
if splits[-1] == "prophetnet" and splits[-2] == "xlm":
|
||||
return "_".join(splits[-2:])
|
||||
# Secial case for xlm_roberta
|
||||
if splits[-1] == "roberta" and splits[-2] == "xlm":
|
||||
return "_".join(splits[-2:])
|
||||
|
||||
Loading…
Reference in New Issue
Block a user