Skip to content

GitLab

"vscode:/vscode.git/clone" did not exist on "47a98fc4cb6a561576309a57b315b042977d194c"
Unverified Commit f617490e authored by abhishek thakur's avatar abhishek thakur Committed by GitHub
Browse files

ConvBERT Model (#9717) 

* finalize convbert

* finalize convbert

* fix

* fix

* fix

* push

* fix

* tf image patches

* fix torch model

* tf tests

* conversion

* everything aligned

* remove print

* tf tests

* fix tf

* make tf tests pass

* everything works

* fix init

* fix

* special treatment for sepconv1d

* style

* 🙏🏽



* add doc and cleanup

* add electra test again

* fix doc

* fix doc again

* fix doc again

* Update src/transformers/modeling_tf_pytorch_utils.py
Co-authored-by: default avatarLysandre Debut <lysandre@huggingface.co>

* Update src/transformers/models/conv_bert/configuration_conv_bert.py
Co-authored-by: default avatarLysandre Debut <lysandre@huggingface.co>

* Update docs/source/model_doc/conv_bert.rst
Co-authored-by: default avatarSylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update src/transformers/models/auto/configuration_auto.py
Co-authored-by: default avatarSylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update src/transformers/models/conv_bert/configuration_conv_bert.py
Co-authored-by: default avatarSylvain Gugger <35901082+sgugger@users.noreply.github.com>

* conv_bert -> convbert

* more fixes from review

* add conversion script

* dont use pretrained embed

* unused config

* suggestions from julien

* some more fixes

* p -> param

* fix copyright

* fix doc

* Update src/transformers/models/convbert/configuration_convbert.py
Co-authored-by: default avatarPatrick von Platen <patrick.v.platen@gmail.com>

* comments from reviews

* fix-copies

* fix style

* revert shape_list
Co-authored-by: default avatarLysandre Debut <lysandre@huggingface.co>
Co-authored-by: default avatarSylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: default avatarPatrick von Platen <patrick.v.platen@gmail.com>
parent e575e062
Expand all Show whitespace changes
Inline Side-by-side
Showing with 3629 additions and 33 deletions
README.md
View file @ f617490e
......@@ -197,6 +197,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
1. **[Blenderbot](https://huggingface.co/transformers/model_doc/blenderbot.html)** (from Facebook) released with the paper [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) by Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston.
1. **[BlenderbotSmall](https://huggingface.co/transformers/model_doc/blenderbot_small.html)** (from Facebook) released with the paper [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) by Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston.
1. **[CamemBERT](https://huggingface.co/transformers/model_doc/camembert.html)** (from Inria/Facebook/Sorbonne) released with the paper [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
1. **[ConvBERT](https://huggingface.co/transformers/model_doc/convbert.html)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
1. **[CTRL](https://huggingface.co/transformers/model_doc/ctrl.html)** (from Salesforce) released with the paper [CTRL: A Conditional Transformer Language Model for Controllable Generation](https://arxiv.org/abs/1909.05858) by Nitish Shirish Keskar*, Bryan McCann*, Lav R. Varshney, Caiming Xiong and Richard Socher.
1. **[DeBERTa](https://huggingface.co/transformers/model_doc/deberta.html)** (from Microsoft Research) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
1. **[DialoGPT](https://huggingface.co/transformers/model_doc/dialogpt.html)** (from Microsoft Research) released with the paper [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://arxiv.org/abs/1911.00536) by Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan.
......
docs/source/index.rst
View file @ f617490e
......@@ -106,97 +106,100 @@ and conversion utilities for the following models:
8. :doc:`CamemBERT <model_doc/camembert>` (from Inria/Facebook/Sorbonne) released with the paper `CamemBERT: a Tasty
French Language Model <https://arxiv.org/abs/1911.03894>`__ by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz
Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
9. :doc:`CTRL <model_doc/ctrl>` (from Salesforce) released with the paper `CTRL: A Conditional Transformer Language
9. :doc:`ConvBERT <model_doc/convbert>` (from YituTech) released with the paper `ConvBERT: Improving BERT with
Span-based Dynamic Convolution <https://arxiv.org/abs/2008.02496>`__ by Zihang Jiang, Weihao Yu, Daquan Zhou,
Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
10. :doc:`CTRL <model_doc/ctrl>` (from Salesforce) released with the paper `CTRL: A Conditional Transformer Language
Model for Controllable Generation <https://arxiv.org/abs/1909.05858>`__ by Nitish Shirish Keskar*, Bryan McCann*,
Lav R. Varshney, Caiming Xiong and Richard Socher.
10. :doc:`DeBERTa <model_doc/deberta>` (from Microsoft Research) released with the paper `DeBERTa: Decoding-enhanced
11. :doc:`DeBERTa <model_doc/deberta>` (from Microsoft Research) released with the paper `DeBERTa: Decoding-enhanced
BERT with Disentangled Attention <https://arxiv.org/abs/2006.03654>`__ by Pengcheng He, Xiaodong Liu, Jianfeng Gao,
Weizhu Chen.
11. :doc:`DialoGPT <model_doc/dialogpt>` (from Microsoft Research) released with the paper `DialoGPT: Large-Scale
12. :doc:`DialoGPT <model_doc/dialogpt>` (from Microsoft Research) released with the paper `DialoGPT: Large-Scale
Generative Pre-training for Conversational Response Generation <https://arxiv.org/abs/1911.00536>`__ by Yizhe
Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan.
12. :doc:`DistilBERT <model_doc/distilbert>` (from HuggingFace), released together with the paper `DistilBERT, a
13. :doc:`DistilBERT <model_doc/distilbert>` (from HuggingFace), released together with the paper `DistilBERT, a
distilled version of BERT: smaller, faster, cheaper and lighter <https://arxiv.org/abs/1910.01108>`__ by Victor
Sanh, Lysandre Debut and Thomas Wolf. The same method has been applied to compress GPT2 into `DistilGPT2
<https://github.com/huggingface/transformers/tree/master/examples/distillation>`__, RoBERTa into `DistilRoBERTa
<https://github.com/huggingface/transformers/tree/master/examples/distillation>`__, Multilingual BERT into
`DistilmBERT <https://github.com/huggingface/transformers/tree/master/examples/distillation>`__ and a German
version of DistilBERT.
13. :doc:`DPR <model_doc/dpr>` (from Facebook) released with the paper `Dense Passage Retrieval for Open-Domain
14. :doc:`DPR <model_doc/dpr>` (from Facebook) released with the paper `Dense Passage Retrieval for Open-Domain
Question Answering <https://arxiv.org/abs/2004.04906>`__ by Vladimir Karpukhin, Barlas Oğuz, Sewon Min, Patrick
Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
14. :doc:`ELECTRA <model_doc/electra>` (from Google Research/Stanford University) released with the paper `ELECTRA:
15. :doc:`ELECTRA <model_doc/electra>` (from Google Research/Stanford University) released with the paper `ELECTRA:
Pre-training text encoders as discriminators rather than generators <https://arxiv.org/abs/2003.10555>`__ by Kevin
Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning.
15. :doc:`FlauBERT <model_doc/flaubert>` (from CNRS) released with the paper `FlauBERT: Unsupervised Language Model
16. :doc:`FlauBERT <model_doc/flaubert>` (from CNRS) released with the paper `FlauBERT: Unsupervised Language Model
Pre-training for French <https://arxiv.org/abs/1912.05372>`__ by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne,
Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
16. :doc:`Funnel Transformer <model_doc/funnel>` (from CMU/Google Brain) released with the paper `Funnel-Transformer:
17. :doc:`Funnel Transformer <model_doc/funnel>` (from CMU/Google Brain) released with the paper `Funnel-Transformer:
Filtering out Sequential Redundancy for Efficient Language Processing <https://arxiv.org/abs/2006.03236>`__ by
Zihang Dai, Guokun Lai, Yiming Yang, Quoc V. Le.
17. :doc:`GPT <model_doc/gpt>` (from OpenAI) released with the paper `Improving Language Understanding by Generative
18. :doc:`GPT <model_doc/gpt>` (from OpenAI) released with the paper `Improving Language Understanding by Generative
Pre-Training <https://blog.openai.com/language-unsupervised/>`__ by Alec Radford, Karthik Narasimhan, Tim Salimans
and Ilya Sutskever.
18. :doc:`GPT-2 <model_doc/gpt2>` (from OpenAI) released with the paper `Language Models are Unsupervised Multitask
19. :doc:`GPT-2 <model_doc/gpt2>` (from OpenAI) released with the paper `Language Models are Unsupervised Multitask
Learners <https://blog.openai.com/better-language-models/>`__ by Alec Radford*, Jeffrey Wu*, Rewon Child, David
Luan, Dario Amodei** and Ilya Sutskever**.
19. :doc:`LayoutLM <model_doc/layoutlm>` (from Microsoft Research Asia) released with the paper `LayoutLM: Pre-training
20. :doc:`LayoutLM <model_doc/layoutlm>` (from Microsoft Research Asia) released with the paper `LayoutLM: Pre-training
of Text and Layout for Document Image Understanding <https://arxiv.org/abs/1912.13318>`__ by Yiheng Xu, Minghao Li,
Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou.
20. :doc:`LED <model_doc/led>` (from AllenAI) released with the paper `Longformer: The Long-Document Transformer
21. :doc:`LED <model_doc/led>` (from AllenAI) released with the paper `Longformer: The Long-Document Transformer
<https://arxiv.org/abs/2004.05150>`__ by Iz Beltagy, Matthew E. Peters, Arman Cohan.
21. :doc:`Longformer <model_doc/longformer>` (from AllenAI) released with the paper `Longformer: The Long-Document
22. :doc:`Longformer <model_doc/longformer>` (from AllenAI) released with the paper `Longformer: The Long-Document
Transformer <https://arxiv.org/abs/2004.05150>`__ by Iz Beltagy, Matthew E. Peters, Arman Cohan.
22. :doc:`LXMERT <model_doc/lxmert>` (from UNC Chapel Hill) released with the paper `LXMERT: Learning Cross-Modality
23. :doc:`LXMERT <model_doc/lxmert>` (from UNC Chapel Hill) released with the paper `LXMERT: Learning Cross-Modality
Encoder Representations from Transformers for Open-Domain Question Answering <https://arxiv.org/abs/1908.07490>`__
by Hao Tan and Mohit Bansal.
23. :doc:`MarianMT <model_doc/marian>` Machine translation models trained using `OPUS <http://opus.nlpl.eu/>`__ data by
24. :doc:`MarianMT <model_doc/marian>` 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.
24. :doc:`MBart <model_doc/mbart>` (from Facebook) released with the paper `Multilingual Denoising Pre-training for
25. :doc:`MBart <model_doc/mbart>` (from Facebook) released with the paper `Multilingual Denoising Pre-training for
Neural Machine Translation <https://arxiv.org/abs/2001.08210>`__ by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li,
Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer.
25. :doc:`MPNet <model_doc/mpnet>` (from Microsoft Research) released with the paper `MPNet: Masked and Permuted
26. :doc:`MPNet <model_doc/mpnet>` (from Microsoft Research) released with the paper `MPNet: Masked and Permuted
Pre-training for Language Understanding <https://arxiv.org/abs/2004.09297>`__ by Kaitao Song, Xu Tan, Tao Qin,
Jianfeng Lu, Tie-Yan Liu.
26. :doc:`MT5 <model_doc/mt5>` (from Google AI) released with the paper `mT5: A massively multilingual pre-trained
27. :doc:`MT5 <model_doc/mt5>` (from Google AI) released with the paper `mT5: A massively multilingual pre-trained
text-to-text transformer <https://arxiv.org/abs/2010.11934>`__ by Linting Xue, Noah Constant, Adam Roberts, Mihir
Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel.
27. :doc:`Pegasus <model_doc/pegasus>` (from Google) released with the paper `PEGASUS: Pre-training with Extracted
28. :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.
28. :doc:`ProphetNet <model_doc/prophetnet>` (from Microsoft Research) released with the paper `ProphetNet: Predicting
29. :doc:`ProphetNet <model_doc/prophetnet>` (from Microsoft Research) released with the paper `ProphetNet: Predicting
Future N-gram for Sequence-to-Sequence Pre-training <https://arxiv.org/abs/2001.04063>`__ by Yu Yan, Weizhen Qi,
Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
29. :doc:`Reformer <model_doc/reformer>` (from Google Research) released with the paper `Reformer: The Efficient
30. :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.
30. :doc:`RoBERTa <model_doc/roberta>` (from Facebook), released together with the paper a `Robustly Optimized BERT
31. :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.
31. :doc:`SqueezeBert <model_doc/squeezebert>` released with the paper `SqueezeBERT: What can computer vision teach NLP
32. :doc:`SqueezeBert <model_doc/squeezebert>` released with the paper `SqueezeBERT: What can computer vision teach NLP
about efficient neural networks? <https://arxiv.org/abs/2006.11316>`__ by Forrest N. Iandola, Albert E. Shaw, Ravi
Krishna, and Kurt W. Keutzer.
32. :doc:`T5 <model_doc/t5>` (from Google AI) released with the paper `Exploring the Limits of Transfer Learning with a
33. :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.
33. :doc:`TAPAS <model_doc/tapas>` (from Google AI) released with the paper `TAPAS: Weakly Supervised Table Parsing via
34. :doc:`TAPAS <model_doc/tapas>` (from Google AI) released with the paper `TAPAS: Weakly Supervised Table Parsing via
Pre-training <https://arxiv.org/abs/2004.02349>`__ by Jonathan Herzig, Paweł Krzysztof Nowak, Thomas Müller,
Francesco Piccinno and Julian Martin Eisenschlos.
34. :doc:`Transformer-XL <model_doc/transformerxl>` (from Google/CMU) released with the paper `Transformer-XL:
35. :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.
35. :doc:`XLM <model_doc/xlm>` (from Facebook) released together with the paper `Cross-lingual Language Model
36. :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.
36. :doc:`XLM-ProphetNet <model_doc/xlmprophetnet>` (from Microsoft Research) released with the paper `ProphetNet:
37. :doc:`XLM-ProphetNet <model_doc/xlmprophetnet>` (from Microsoft Research) released with the paper `ProphetNet:
Predicting Future N-gram for Sequence-to-Sequence Pre-training <https://arxiv.org/abs/2001.04063>`__ by Yu Yan,
Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
37. :doc:`XLM-RoBERTa <model_doc/xlmroberta>` (from Facebook AI), released together with the paper `Unsupervised
38. :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.
38. :doc:`XLNet <model_doc/xlnet>` (from Google/CMU) released with the paper `​XLNet: Generalized Autoregressive
39. :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.
......@@ -231,6 +234,8 @@ TensorFlow and/or Flax.
+-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
| CamemBERT | ✅ | ✅ | ✅ | ✅ | ❌ |
+-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
| ConvBERT | ✅ | ✅ | ✅ | ✅ | ❌ |
+-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
| DPR | ✅ | ✅ | ✅ | ✅ | ❌ |
+-----------------------------+----------------+----------------+-----------------+--------------------+--------------+
| DeBERTa | ✅ | ❌ | ✅ | ❌ | ❌ |
......@@ -371,6 +376,7 @@ TensorFlow and/or Flax.
model_doc/blenderbot
model_doc/blenderbot_small
model_doc/camembert
model_doc/convbert
model_doc/ctrl
model_doc/deberta
model_doc/dialogpt
......
docs/source/model_doc/convbert.rst 0 → 100644
View file @ f617490e
..
Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
ConvBERT
-----------------------------------------------------------------------------------------------------------------------
Overview
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The ConvBERT model was proposed in `ConvBERT: Improving BERT with Span-based Dynamic Convolution
<https://arxiv.org/abs/2008.02496>`__ by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng
Yan.
The abstract from the paper is the following:
*Pre-trained language models like BERT and its variants have recently achieved impressive performance in various
natural language understanding tasks. However, BERT heavily relies on the global self-attention block and thus suffers
large memory footprint and computation cost. Although all its attention heads query on the whole input sequence for
generating the attention map from a global perspective, we observe some heads only need to learn local dependencies,
which means the existence of computation redundancy. We therefore propose a novel span-based dynamic convolution to
replace these self-attention heads to directly model local dependencies. The novel convolution heads, together with the
rest self-attention heads, form a new mixed attention block that is more efficient at both global and local context
learning. We equip BERT with this mixed attention design and build a ConvBERT model. Experiments have shown that
ConvBERT significantly outperforms BERT and its variants in various downstream tasks, with lower training cost and
fewer model parameters. Remarkably, ConvBERTbase model achieves 86.4 GLUE score, 0.7 higher than ELECTRAbase, while
using less than 1/4 training cost. Code and pre-trained models will be released.*
ConvBERT training tips are similar to those of BERT. The original implementation can be found here:
https://github.com/yitu-opensource/ConvBert
ConvBertConfig
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertConfig
:members:
ConvBertTokenizer
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertTokenizer
:members: build_inputs_with_special_tokens, get_special_tokens_mask,
create_token_type_ids_from_sequences, save_vocabulary
ConvBertTokenizerFast
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertTokenizerFast
:members: build_inputs_with_special_tokens, get_special_tokens_mask,
create_token_type_ids_from_sequences, save_vocabulary
ConvBertModel
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertModel
:members: forward
ConvBertForMaskedLM
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertForMaskedLM
:members: forward
ConvBertForSequenceClassification
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertForSequenceClassification
:members: forward
ConvBertForMultipleChoice
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertForMultipleChoice
:members: forward
ConvBertForTokenClassification
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertForTokenClassification
:members: forward
ConvBertForQuestionAnswering
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.ConvBertForQuestionAnswering
:members: forward
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.TFConvBertModel
:members: call
TFConvBertForMaskedLM
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.TFConvBertForMaskedLM
:members: call
TFConvBertForSequenceClassification
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.TFConvBertForSequenceClassification
:members: call
TFConvBertForMultipleChoice
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.TFConvBertForMultipleChoice
:members: call
TFConvBertForTokenClassification
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.TFConvBertForTokenClassification
:members: call
TFConvBertForQuestionAnswering
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. autoclass:: transformers.TFConvBertForQuestionAnswering
:members: call
docs/source/model_summary.rst
View file @ f617490e
......@@ -330,6 +330,36 @@ the same probabilities as the larger model. The actual objective is a combinatio
The library provides a version of the model for masked language modeling, token classification, sentence classification
and question answering.
ConvBERT
-----------------------------------------------------------------------------------------------------------------------
.. raw:: html
<a href="https://huggingface.co/models?filter=convbert">
<img alt="Models" src="https://img.shields.io/badge/All_model_pages-convbert-blueviolet">
</a>
<a href="model_doc/convbert.html">
<img alt="Doc" src="https://img.shields.io/badge/Model_documentation-convbert-blueviolet">
</a>
`ConvBERT: Improving BERT with Span-based Dynamic Convolution <https://arxiv.org/abs/1910.01108>`_, Zihang Jiang,
Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
Pre-trained language models like BERT and its variants have recently achieved impressive performance in various natural
language understanding tasks. However, BERT heavily relies on the global self-attention block and thus suffers large
memory footprint and computation cost. Although all its attention heads query on the whole input sequence for
generating the attention map from a global perspective, we observe some heads only need to learn local dependencies,
which means the existence of computation redundancy. We therefore propose a novel span-based dynamic convolution to
replace these self-attention heads to directly model local dependencies. The novel convolution heads, together with the
rest self-attention heads, form a new mixed attention block that is more efficient at both global and local context
learning. We equip BERT with this mixed attention design and build a ConvBERT model. Experiments have shown that
ConvBERT significantly outperforms BERT and its variants in various downstream tasks, with lower training cost and
fewer model parameters. Remarkably, ConvBERTbase model achieves 86.4 GLUE score, 0.7 higher than ELECTRAbase, while
using less than 1/4 training cost.
The library provides a version of the model for masked language modeling, token classification, sentence classification
and question answering.
XLM
-----------------------------------------------------------------------------------------------------------------------
......
src/transformers/__init__.py
View file @ f617490e
......@@ -125,6 +125,7 @@ _import_structure = {
],
"models": [],
# Models
"models.convbert": ["CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "ConvBertConfig", "ConvBertTokenizer"],
"models.albert": ["ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "AlbertConfig"],
"models.auto": [
"ALL_PRETRAINED_CONFIG_ARCHIVE_MAP",
......@@ -275,6 +276,7 @@ else:
# tokenziers-backed objects
if is_tokenizers_available():
# Fast tokenizers
_import_structure["models.convbert"].append("ConvBertTokenizerFast")
_import_structure["models.albert"].append("AlbertTokenizerFast")
_import_structure["models.bart"].append("BartTokenizerFast")
_import_structure["models.barthez"].append("BarthezTokenizerFast")
......@@ -360,6 +362,21 @@ if is_torch_available():
_import_structure["generation_utils"] = ["top_k_top_p_filtering"]
_import_structure["modeling_utils"] = ["Conv1D", "PreTrainedModel", "apply_chunking_to_forward", "prune_layer"]
# PyTorch models structure
_import_structure["models.convbert"].extend(
[
"CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"ConvBertForMaskedLM",
"ConvBertForMultipleChoice",
"ConvBertForQuestionAnswering",
"ConvBertForSequenceClassification",
"ConvBertForTokenClassification",
"ConvBertLayer",
"ConvBertModel",
"ConvBertPreTrainedModel",
"load_tf_weights_in_convbert",
]
)
_import_structure["models.albert"].extend(
[
"ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST",
......@@ -816,6 +833,20 @@ if is_tf_available():
"shape_list",
]
# TensorFlow models structure
_import_structure["models.convbert"].extend(
[
"TF_CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"TFConvBertForMaskedLM",
"TFConvBertForMultipleChoice",
"TFConvBertForQuestionAnswering",
"TFConvBertForSequenceClassification",
"TFConvBertForTokenClassification",
"TFConvBertLayer",
"TFConvBertModel",
"TFConvBertPreTrainedModel",
]
)
_import_structure["models.albert"].extend(
[
"TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST",
......@@ -1235,6 +1266,7 @@ if TYPE_CHECKING:
BlenderbotSmallTokenizer,
)
from .models.camembert import CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, CamembertConfig
from .models.convbert import CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, ConvBertConfig, ConvBertTokenizer
from .models.ctrl import CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP, CTRLConfig, CTRLTokenizer
from .models.deberta import DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, DebertaConfig, DebertaTokenizer
from .models.distilbert import DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, DistilBertConfig, DistilBertTokenizer
......@@ -1359,6 +1391,7 @@ if TYPE_CHECKING:
from .models.barthez import BarthezTokenizerFast
from .models.bert import BertTokenizerFast
from .models.camembert import CamembertTokenizerFast
from .models.convbert import ConvBertTokenizerFast
from .models.distilbert import DistilBertTokenizerFast
from .models.dpr import DPRContextEncoderTokenizerFast, DPRQuestionEncoderTokenizerFast, DPRReaderTokenizerFast
from .models.electra import ElectraTokenizerFast
......@@ -1521,6 +1554,18 @@ if TYPE_CHECKING:
CamembertForTokenClassification,
CamembertModel,
)
from .models.convbert import (
CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
ConvBertForMaskedLM,
ConvBertForMultipleChoice,
ConvBertForQuestionAnswering,
ConvBertForSequenceClassification,
ConvBertForTokenClassification,
ConvBertLayer,
ConvBertModel,
ConvBertPreTrainedModel,
load_tf_weights_in_convbert,
)
from .models.ctrl import (
CTRL_PRETRAINED_MODEL_ARCHIVE_LIST,
CTRLForSequenceClassification,
......@@ -1879,6 +1924,17 @@ if TYPE_CHECKING:
TFCamembertForTokenClassification,
TFCamembertModel,
)
from .models.convbert import (
TF_CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
TFConvBertForMaskedLM,
TFConvBertForMultipleChoice,
TFConvBertForQuestionAnswering,
TFConvBertForSequenceClassification,
TFConvBertForTokenClassification,
TFConvBertLayer,
TFConvBertModel,
TFConvBertPreTrainedModel,
)
from .models.ctrl import (
TF_CTRL_PRETRAINED_MODEL_ARCHIVE_LIST,
TFCTRLForSequenceClassification,
......
src/transformers/convert_slow_tokenizer.py
View file @ f617490e
......@@ -605,6 +605,7 @@ SLOW_TO_FAST_CONVERTERS = {
"BarthezTokenizer": BarthezConverter,
"BertTokenizer": BertConverter,
"CamembertTokenizer": CamembertConverter,
"ConvBertTokenizer": BertConverter,
"DistilBertTokenizer": BertConverter,
"DPRReaderTokenizer": BertConverter,
"DPRQuestionEncoderTokenizer": BertConverter,
......
src/transformers/modeling_tf_pytorch_utils.py
View file @ f617490e
......@@ -64,6 +64,10 @@ def convert_tf_weight_name_to_pt_weight_name(tf_name, start_prefix_to_remove="")
if tf_name[-1] == "beta":
tf_name[-1] = "bias"
# The SeparableConv1D TF layer contains two weights that are translated to PyTorch Conv1D here
if tf_name[-1] == "pointwise_kernel" or tf_name[-1] == "depthwise_kernel":
tf_name[-1] = tf_name[-1].replace("_kernel", ".weight")
# Remove prefix if needed
tf_name = ".".join(tf_name)
if start_prefix_to_remove:
......@@ -127,7 +131,6 @@ def load_pytorch_weights_in_tf2_model(tf_model, pt_state_dict, tf_inputs=None, a
if tf_inputs is not None:
tf_model(tf_inputs, training=False) # Make sure model is built
# Adapt state dict - TODO remove this and update the AWS weights files instead
# Convert old format to new format if needed from a PyTorch state_dict
old_keys = []
......
src/transformers/models/__init__.py
View file @ f617490e
......@@ -28,6 +28,7 @@ from . import (
blenderbot,
blenderbot_small,
camembert,
convbert,
ctrl,
deberta,
dialogpt,
......
src/transformers/models/auto/configuration_auto.py
View file @ f617490e
......@@ -28,6 +28,7 @@ from ..blenderbot_small.configuration_blenderbot_small import (
BlenderbotSmallConfig,
)
from ..camembert.configuration_camembert import CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, CamembertConfig
from ..convbert.configuration_convbert import CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, ConvBertConfig
from ..ctrl.configuration_ctrl import CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP, CTRLConfig
from ..deberta.configuration_deberta import DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP, DebertaConfig
from ..distilbert.configuration_distilbert import DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, DistilBertConfig
......@@ -71,6 +72,7 @@ ALL_PRETRAINED_CONFIG_ARCHIVE_MAP = dict(
(key, value)
for pretrained_map in [
# Add archive maps here
CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
LED_PRETRAINED_CONFIG_ARCHIVE_MAP,
BLENDERBOT_SMALL_PRETRAINED_CONFIG_ARCHIVE_MAP,
BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
......@@ -112,6 +114,7 @@ ALL_PRETRAINED_CONFIG_ARCHIVE_MAP = dict(
CONFIG_MAPPING = OrderedDict(
[
# Add configs here
("convbert", ConvBertConfig),
("led", LEDConfig),
("blenderbot-small", BlenderbotSmallConfig),
("retribert", RetriBertConfig),
......@@ -159,6 +162,7 @@ CONFIG_MAPPING = OrderedDict(
MODEL_NAMES_MAPPING = OrderedDict(
[
# Add full (and cased) model names here
("convbert", "ConvBERT"),
("led", "LED"),
("blenderbot-small", "BlenderbotSmall"),
("retribert", "RetriBERT"),
......
src/transformers/models/auto/modeling_auto.py
View file @ f617490e
......@@ -61,6 +61,16 @@ from ..camembert.modeling_camembert import (
CamembertForTokenClassification,
CamembertModel,
)
# Add modeling imports here
from ..convbert.modeling_convbert import (
ConvBertForMaskedLM,
ConvBertForMultipleChoice,
ConvBertForQuestionAnswering,
ConvBertForSequenceClassification,
ConvBertForTokenClassification,
ConvBertModel,
)
from ..ctrl.modeling_ctrl import CTRLForSequenceClassification, CTRLLMHeadModel, CTRLModel
from ..deberta.modeling_deberta import (
DebertaForMaskedLM,
......@@ -234,6 +244,7 @@ from .configuration_auto import (
BlenderbotConfig,
BlenderbotSmallConfig,
CamembertConfig,
ConvBertConfig,
CTRLConfig,
DebertaConfig,
DistilBertConfig,
......@@ -277,6 +288,7 @@ logger = logging.get_logger(__name__)
MODEL_MAPPING = OrderedDict(
[
# Base model mapping
(ConvBertConfig, ConvBertModel),
(LEDConfig, LEDModel),
(BlenderbotSmallConfig, BlenderbotSmallModel),
(RetriBertConfig, RetriBertModel),
......@@ -355,6 +367,7 @@ MODEL_FOR_PRETRAINING_MAPPING = OrderedDict(
MODEL_WITH_LM_HEAD_MAPPING = OrderedDict(
[
# Model with LM heads mapping
(ConvBertConfig, ConvBertForMaskedLM),
(LEDConfig, LEDForConditionalGeneration),
(BlenderbotSmallConfig, BlenderbotSmallForConditionalGeneration),
(LayoutLMConfig, LayoutLMForMaskedLM),
......@@ -414,6 +427,7 @@ MODEL_FOR_CAUSAL_LM_MAPPING = OrderedDict(
MODEL_FOR_MASKED_LM_MAPPING = OrderedDict(
[
# Model for Masked LM mapping
(ConvBertConfig, ConvBertForMaskedLM),
(LayoutLMConfig, LayoutLMForMaskedLM),
(DistilBertConfig, DistilBertForMaskedLM),
(AlbertConfig, AlbertForMaskedLM),
......@@ -459,6 +473,7 @@ MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING = OrderedDict(
MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING = OrderedDict(
[
# Model for Sequence Classification mapping
(ConvBertConfig, ConvBertForSequenceClassification),
(LEDConfig, LEDForSequenceClassification),
(DistilBertConfig, DistilBertForSequenceClassification),
(AlbertConfig, AlbertForSequenceClassification),
......@@ -491,6 +506,7 @@ MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING = OrderedDict(
MODEL_FOR_QUESTION_ANSWERING_MAPPING = OrderedDict(
[
# Model for Question Answering mapping
(ConvBertConfig, ConvBertForQuestionAnswering),
(LEDConfig, LEDForQuestionAnswering),
(DistilBertConfig, DistilBertForQuestionAnswering),
(AlbertConfig, AlbertForQuestionAnswering),
......@@ -525,6 +541,7 @@ MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPING = OrderedDict(
MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = OrderedDict(
[
# Model for Token Classification mapping
(ConvBertConfig, ConvBertForTokenClassification),
(LayoutLMConfig, LayoutLMForTokenClassification),
(DistilBertConfig, DistilBertForTokenClassification),
(CamembertConfig, CamembertForTokenClassification),
......@@ -549,6 +566,7 @@ MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = OrderedDict(
MODEL_FOR_MULTIPLE_CHOICE_MAPPING = OrderedDict(
[
# Model for Multiple Choice mapping
(ConvBertConfig, ConvBertForMultipleChoice),
(CamembertConfig, CamembertForMultipleChoice),
(ElectraConfig, ElectraForMultipleChoice),
(XLMRobertaConfig, XLMRobertaForMultipleChoice),
......
src/transformers/models/auto/modeling_tf_auto.py
View file @ f617490e
......@@ -57,6 +57,14 @@ from ..camembert.modeling_tf_camembert import (
TFCamembertForTokenClassification,
TFCamembertModel,
)
from ..convbert.modeling_tf_convbert import (
TFConvBertForMaskedLM,
TFConvBertForMultipleChoice,
TFConvBertForQuestionAnswering,
TFConvBertForSequenceClassification,
TFConvBertForTokenClassification,
TFConvBertModel,
)
from ..ctrl.modeling_tf_ctrl import TFCTRLForSequenceClassification, TFCTRLLMHeadModel, TFCTRLModel
from ..distilbert.modeling_tf_distilbert import (
TFDistilBertForMaskedLM,
......@@ -173,6 +181,7 @@ from .configuration_auto import (
BlenderbotConfig,
BlenderbotSmallConfig,
CamembertConfig,
ConvBertConfig,
CTRLConfig,
DistilBertConfig,
DPRConfig,
......@@ -206,6 +215,7 @@ logger = logging.get_logger(__name__)
TF_MODEL_MAPPING = OrderedDict(
[
# Base model mapping
(ConvBertConfig, TFConvBertModel),
(LEDConfig, TFLEDModel),
(LxmertConfig, TFLxmertModel),
(MT5Config, TFMT5Model),
......@@ -268,6 +278,7 @@ TF_MODEL_FOR_PRETRAINING_MAPPING = OrderedDict(
TF_MODEL_WITH_LM_HEAD_MAPPING = OrderedDict(
[
# Model with LM heads mapping
(ConvBertConfig, TFConvBertForMaskedLM),
(LEDConfig, TFLEDForConditionalGeneration),
(T5Config, TFT5ForConditionalGeneration),
(DistilBertConfig, TFDistilBertForMaskedLM),
......@@ -312,6 +323,7 @@ TF_MODEL_FOR_CAUSAL_LM_MAPPING = OrderedDict(
TF_MODEL_FOR_MASKED_LM_MAPPING = OrderedDict(
[
# Model for Masked LM mapping
(ConvBertConfig, TFConvBertForMaskedLM),
(DistilBertConfig, TFDistilBertForMaskedLM),
(AlbertConfig, TFAlbertForMaskedLM),
(CamembertConfig, TFCamembertForMaskedLM),
......@@ -347,6 +359,7 @@ TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING = OrderedDict(
TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING = OrderedDict(
[
# Model for Sequence Classification mapping
(ConvBertConfig, TFConvBertForSequenceClassification),
(DistilBertConfig, TFDistilBertForSequenceClassification),
(AlbertConfig, TFAlbertForSequenceClassification),
(CamembertConfig, TFCamembertForSequenceClassification),
......@@ -371,6 +384,7 @@ TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING = OrderedDict(
TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING = OrderedDict(
[
# Model for Question Answering mapping
(ConvBertConfig, TFConvBertForQuestionAnswering),
(DistilBertConfig, TFDistilBertForQuestionAnswering),
(AlbertConfig, TFAlbertForQuestionAnswering),
(CamembertConfig, TFCamembertForQuestionAnswering),
......@@ -391,6 +405,7 @@ TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING = OrderedDict(
TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = OrderedDict(
[
# Model for Token Classification mapping
(ConvBertConfig, TFConvBertForTokenClassification),
(DistilBertConfig, TFDistilBertForTokenClassification),
(AlbertConfig, TFAlbertForTokenClassification),
(CamembertConfig, TFCamembertForTokenClassification),
......@@ -411,6 +426,7 @@ TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = OrderedDict(
TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING = OrderedDict(
[
# Model for Multiple Choice mapping
(ConvBertConfig, TFConvBertForMultipleChoice),
(CamembertConfig, TFCamembertForMultipleChoice),
(XLMConfig, TFXLMForMultipleChoice),
(XLMRobertaConfig, TFXLMRobertaForMultipleChoice),
......
src/transformers/models/auto/tokenization_auto.py
View file @ f617490e
......@@ -26,6 +26,7 @@ from ..bert_japanese.tokenization_bert_japanese import BertJapaneseTokenizer
from ..bertweet.tokenization_bertweet import BertweetTokenizer
from ..blenderbot.tokenization_blenderbot import BlenderbotTokenizer
from ..blenderbot_small.tokenization_blenderbot_small import BlenderbotSmallTokenizer
from ..convbert.tokenization_convbert import ConvBertTokenizer
from ..ctrl.tokenization_ctrl import CTRLTokenizer
from ..deberta.tokenization_deberta import DebertaTokenizer
from ..distilbert.tokenization_distilbert import DistilBertTokenizer
......@@ -61,6 +62,7 @@ from .configuration_auto import (
BlenderbotConfig,
BlenderbotSmallConfig,
CamembertConfig,
ConvBertConfig,
CTRLConfig,
DebertaConfig,
DistilBertConfig,
......@@ -134,6 +136,7 @@ if is_tokenizers_available():
from ..barthez.tokenization_barthez_fast import BarthezTokenizerFast
from ..bert.tokenization_bert_fast import BertTokenizerFast
from ..camembert.tokenization_camembert_fast import CamembertTokenizerFast
from ..convbert.tokenization_convbert_fast import ConvBertTokenizerFast
from ..distilbert.tokenization_distilbert_fast import DistilBertTokenizerFast
from ..dpr.tokenization_dpr_fast import DPRQuestionEncoderTokenizerFast
from ..electra.tokenization_electra_fast import ElectraTokenizerFast
......@@ -163,6 +166,7 @@ else:
BarthezTokenizerFast = None
BertTokenizerFast = None
CamembertTokenizerFast = None
ConvBertTokenizerFast = None
DistilBertTokenizerFast = None
DPRQuestionEncoderTokenizerFast = None
ElectraTokenizerFast = None
......@@ -233,6 +237,7 @@ TOKENIZER_MAPPING = OrderedDict(
(MPNetConfig, (MPNetTokenizer, MPNetTokenizerFast)),
(TapasConfig, (TapasTokenizer, None)),
(LEDConfig, (LEDTokenizer, LEDTokenizerFast)),
(ConvBertConfig, (ConvBertTokenizer, ConvBertTokenizerFast)),
]
)
......
src/transformers/models/convbert/__init__.py 0 → 100644
View file @ f617490e
# flake8: noqa
# There's no way to ignore "F401 '...' imported but unused" warnings in this
# module, but to preserve other warnings. So, don't check this module at all.
# Copyright 2020 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import TYPE_CHECKING
from ...file_utils import _BaseLazyModule, is_tf_available, is_tokenizers_available, is_torch_available
_import_structure = {
"configuration_convbert": ["CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "ConvBertConfig"],
"tokenization_convbert": ["ConvBertTokenizer"],
}
if is_tokenizers_available():
_import_structure["tokenization_convbert_fast"] = ["ConvBertTokenizerFast"]
if is_torch_available():
_import_structure["modeling_convbert"] = [
"CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"ConvBertForMaskedLM",
"ConvBertForMultipleChoice",
"ConvBertForQuestionAnswering",
"ConvBertForSequenceClassification",
"ConvBertForTokenClassification",
"ConvBertLayer",
"ConvBertModel",
"ConvBertPreTrainedModel",
"load_tf_weights_in_convbert",
]
if is_tf_available():
_import_structure["modeling_tf_convbert"] = [
"TF_CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST",
"TFConvBertForMaskedLM",
"TFConvBertForMultipleChoice",
"TFConvBertForQuestionAnswering",
"TFConvBertForSequenceClassification",
"TFConvBertForTokenClassification",
"TFConvBertLayer",
"TFConvBertModel",
"TFConvBertPreTrainedModel",
]
if TYPE_CHECKING:
from .configuration_convbert import CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, ConvBertConfig
from .tokenization_convbert import ConvBertTokenizer
if is_tokenizers_available():
from .tokenization_convbert_fast import ConvBertTokenizerFast
if is_torch_available():
from .modeling_convbert import (
CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
ConvBertForMaskedLM,
ConvBertForMultipleChoice,
ConvBertForQuestionAnswering,
ConvBertForSequenceClassification,
ConvBertForTokenClassification,
ConvBertLayer,
ConvBertModel,
ConvBertPreTrainedModel,
load_tf_weights_in_convbert,
)
if is_tf_available():
from .modeling_tf_convbert import (
TF_CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST,
TFConvBertForMaskedLM,
TFConvBertForMultipleChoice,
TFConvBertForQuestionAnswering,
TFConvBertForSequenceClassification,
TFConvBertForTokenClassification,
TFConvBertLayer,
TFConvBertModel,
TFConvBertPreTrainedModel,
)
else:
import importlib
import os
import sys
class _LazyModule(_BaseLazyModule):
"""
Module class that surfaces all objects but only performs associated imports when the objects are requested.
"""
__file__ = globals()["__file__"]
__path__ = [os.path.dirname(__file__)]
def _get_module(self, module_name: str):
return importlib.import_module("." + module_name, self.__name__)
sys.modules[__name__] = _LazyModule(__name__, _import_structure)
src/transformers/models/convbert/configuration_convbert.py 0 → 100644
View file @ f617490e
# coding=utf-8
# Copyright The HuggingFace team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
""" ConvBERT model configuration """
from ...configuration_utils import PretrainedConfig
from ...utils import logging
logger = logging.get_logger(__name__)
CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP = {
"YituTech/conv-bert-base": "https://huggingface.co/YituTech/conv-bert-base/resolve/main/config.json",
"YituTech/conv-bert-medium-small": "https://huggingface.co/YituTech/conv-bert-medium-small/resolve/main/config.json",
"YituTech/conv-bert-small": "https://huggingface.co/YituTech/conv-bert-small/resolve/main/config.json",
# See all ConvBERT models at https://huggingface.co/models?filter=convbert
}
class ConvBertConfig(PretrainedConfig):
r"""
This is the configuration class to store the configuration of a :class:`~transformers.ConvBertModel`. It is used to
instantiate an ConvBERT model according to the specified arguments, defining the model architecture. Instantiating
a configuration with the defaults will yield a similar configuration to that of the ConvBERT `conv-bert-base
<https://huggingface.co/YituTech/conv-bert-base>`__ architecture. Configuration objects inherit from
:class:`~transformers.PretrainedConfig` and can be used to control the model outputs. Read the documentation from
:class:`~transformers.PretrainedConfig` for more information.
Args:
vocab_size (:obj:`int`, `optional`, defaults to 30522):
Vocabulary size of the ConvBERT model. Defines the number of different tokens that can be represented by
the :obj:`inputs_ids` passed when calling :class:`~transformers.ConvBertModel` or
:class:`~transformers.TFConvBertModel`.
hidden_size (:obj:`int`, `optional`, defaults to 768):
Dimensionality of the encoder layers and the pooler layer.
num_hidden_layers (:obj:`int`, `optional`, defaults to 12):
Number of hidden layers in the Transformer encoder.
num_attention_heads (:obj:`int`, `optional`, defaults to 12):
Number of attention heads for each attention layer in the Transformer encoder.
intermediate_size (:obj:`int`, `optional`, defaults to 3072):
Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
hidden_act (: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:`"selu"` and :obj:`"gelu_new"` are supported.
hidden_dropout_prob (:obj:`float`, `optional`, defaults to 0.1):
The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
attention_probs_dropout_prob (:obj:`float`, `optional`, defaults to 0.1):
The dropout ratio for the attention probabilities.
max_position_embeddings (:obj:`int`, `optional`, defaults to 512):
The maximum sequence length that this model might ever be used with. Typically set this to something large
just in case (e.g., 512 or 1024 or 2048).
type_vocab_size (:obj:`int`, `optional`, defaults to 2):
The vocabulary size of the :obj:`token_type_ids` passed when calling :class:`~transformers.ConvBertModel`
or :class:`~transformers.TFConvBertModel`.
initializer_range (:obj:`float`, `optional`, defaults to 0.02):
The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
layer_norm_eps (:obj:`float`, `optional`, defaults to 1e-12):
The epsilon used by the layer normalization layers.
head_ratio (:obj:`int`, `optional`, defaults to 2):
Ratio gamma to reduce the number of attention heads.
num_groups (:obj:`int`, `optional`, defaults to 1):
The number of groups for grouped linear layers for ConvBert model
conv_kernel_size (:obj:`int`, `optional`, defaults to 9):
The size of the convolutional kernel.
Example::
>>> from transformers import ConvBertModel, ConvBertConfig
>>> # Initializing a ConvBERT convbert-base-uncased style configuration
>>> configuration = ConvBertConfig()
>>> # Initializing a model from the convbert-base-uncased style configuration
>>> model = ConvBertModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
"""
model_type = "convbert"
def __init__(
self,
vocab_size=30522,
hidden_size=768,
is_encoder_decoder=False,
num_hidden_layers=12,
num_attention_heads=12,
intermediate_size=3072,
hidden_act="gelu",
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
max_position_embeddings=512,
type_vocab_size=2,
initializer_range=0.02,
layer_norm_eps=1e-12,
pad_token_id=1,
bos_token_id=0,
eos_token_id=2,
embedding_size=768,
head_ratio=2,
conv_kernel_size=9,
num_groups=1,
**kwargs,
):
super().__init__(
pad_token_id=pad_token_id,
is_encoder_decoder=is_encoder_decoder,
bos_token_id=bos_token_id,
eos_token_id=eos_token_id,
**kwargs,
)
self.vocab_size = vocab_size
self.hidden_size = hidden_size
self.num_hidden_layers = num_hidden_layers
self.num_attention_heads = num_attention_heads
self.intermediate_size = intermediate_size
self.hidden_act = hidden_act
self.hidden_dropout_prob = hidden_dropout_prob
self.attention_probs_dropout_prob = attention_probs_dropout_prob
self.max_position_embeddings = max_position_embeddings
self.type_vocab_size = type_vocab_size
self.initializer_range = initializer_range
self.layer_norm_eps = layer_norm_eps
self.embedding_size = embedding_size
self.head_ratio = head_ratio
self.conv_kernel_size = conv_kernel_size
self.num_groups = num_groups
src/transformers/models/convbert/convert_convbert_original_tf1_checkpoint_to_pytorch.py 0 → 100644
View file @ f617490e
# 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 ConvBERT checkpoint."""
import argparse
from ...utils import logging
from .modeling_convbert import ConvBertConfig, ConvBertModel, load_tf_weights_in_convbert
logging.set_verbosity_info()
def convert_orig_tf1_checkpoint_to_pytorch(tf_checkpoint_path, convbert_config_file, pytorch_dump_path):
conf = ConvBertConfig.from_json_file(convbert_config_file)
model = ConvBertModel(conf)
model = load_tf_weights_in_convbert(model, conf, tf_checkpoint_path)
model.save_pretrained(pytorch_dump_path)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
# Required parameters
parser.add_argument(
"--tf_checkpoint_path", default=None, type=str, required=True, help="Path to the TensorFlow checkpoint path."
)
parser.add_argument(
"--convbert_config_file",
default=None,
type=str,
required=True,
help="The config json file corresponding to the pre-trained ConvBERT model. \n"
"This specifies the model architecture.",
)
parser.add_argument(
"--pytorch_dump_path", default=None, type=str, required=True, help="Path to the output PyTorch model."
)
args = parser.parse_args()
convert_orig_tf1_checkpoint_to_pytorch(args.tf_checkpoint_path, args.conv_bert_config_file, args.pytorch_dump_path)
src/transformers/models/convbert/modeling_convbert.py 0 → 100755
View file @ f617490e
This diff is collapsed.
src/transformers/models/convbert/modeling_tf_convbert.py 0 → 100644
View file @ f617490e
This diff is collapsed.
src/transformers/models/convbert/tokenization_convbert.py 0 → 100644
View file @ f617490e
# coding=utf-8
# Copyright 2018 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.
"""Tokenization classes for ConvBERT."""
from ...utils import logging
from ..bert.tokenization_bert import BertTokenizer
logger = logging.get_logger(__name__)
VOCAB_FILES_NAMES = {"vocab_file": "vocab.txt"}
PRETRAINED_VOCAB_FILES_MAP = {
"vocab_file": {
"YituTech/conv-bert-base": "https://huggingface.co/YituTech/conv-bert-base/resolve/main/vocab.txt",
"YituTech/conv-bert-medium-small": "https://huggingface.co/YituTech/conv-bert-medium-small/resolve/main/vocab.txt",
"YituTech/conv-bert-small": "https://huggingface.co/YituTech/conv-bert-small/resolve/main/vocab.txt",
}
}
PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
"YituTech/conv-bert-base": 512,
"YituTech/conv-bert-medium-small": 512,
"YituTech/conv-bert-small": 512,
}
PRETRAINED_INIT_CONFIGURATION = {
"YituTech/conv-bert-base": {"do_lower_case": True},
"YituTech/conv-bert-medium-small": {"do_lower_case": True},
"YituTech/conv-bert-small": {"do_lower_case": True},
}
class ConvBertTokenizer(BertTokenizer):
r"""
Construct a ConvBERT tokenizer. :class:`~transformers.ConvBertTokenizer` is identical to
:class:`~transformers.BertTokenizer` and runs end-to-end tokenization: punctuation splitting and wordpiece. Refer
to superclass :class:`~transformers.BertTokenizer` for usage examples and documentation concerning parameters.
"""
vocab_files_names = VOCAB_FILES_NAMES
pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
pretrained_init_configuration = PRETRAINED_INIT_CONFIGURATION
src/transformers/models/convbert/tokenization_convbert_fast.py 0 → 100644
View file @ f617490e
# coding=utf-8
# Copyright The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Tokenization classes for ConvBERT."""
from ...utils import logging
from ..bert.tokenization_bert_fast import BertTokenizerFast
from .tokenization_convbert import ConvBertTokenizer
logger = logging.get_logger(__name__)
VOCAB_FILES_NAMES = {"vocab_file": "vocab.txt"}
PRETRAINED_VOCAB_FILES_MAP = {
"vocab_file": {
"YituTech/conv-bert-base": "https://huggingface.co/YituTech/conv-bert-base/resolve/main/vocab.txt",
"YituTech/conv-bert-medium-small": "https://huggingface.co/YituTech/conv-bert-medium-small/resolve/main/vocab.txt",
"YituTech/conv-bert-small": "https://huggingface.co/YituTech/conv-bert-small/resolve/main/vocab.txt",
}
}
PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
"YituTech/conv-bert-base": 512,
"YituTech/conv-bert-medium-small": 512,
"YituTech/conv-bert-small": 512,
}
PRETRAINED_INIT_CONFIGURATION = {
"YituTech/conv-bert-base": {"do_lower_case": True},
"YituTech/conv-bert-medium-small": {"do_lower_case": True},
"YituTech/conv-bert-small": {"do_lower_case": True},
}
class ConvBertTokenizerFast(BertTokenizerFast):
r"""
Construct a "fast" ConvBERT tokenizer (backed by HuggingFace's `tokenizers` library).
:class:`~transformers.ConvBertTokenizerFast` is identical to :class:`~transformers.BertTokenizerFast` and runs
end-to-end tokenization: punctuation splitting and wordpiece.
Refer to superclass :class:`~transformers.BertTokenizerFast` for usage examples and documentation concerning
parameters.
"""
vocab_files_names = VOCAB_FILES_NAMES
pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP
max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
pretrained_init_configuration = PRETRAINED_INIT_CONFIGURATION
slow_tokenizer_class = ConvBertTokenizer
src/transformers/utils/dummy_pt_objects.py
View file @ f617490e
......@@ -697,6 +697,81 @@ class CamembertModel:
requires_pytorch(self)
CONVBERT_PRETRAINED_MODEL_ARCHIVE_LIST = None
class ConvBertForMaskedLM:
def __init__(self, *args, **kwargs):
requires_pytorch(self)
@classmethod
def from_pretrained(self, *args, **kwargs):
requires_pytorch(self)
class ConvBertForMultipleChoice:
def __init__(self, *args, **kwargs):
requires_pytorch(self)
@classmethod
def from_pretrained(self, *args, **kwargs):
requires_pytorch(self)
class ConvBertForQuestionAnswering:
def __init__(self, *args, **kwargs):
requires_pytorch(self)
@classmethod
def from_pretrained(self, *args, **kwargs):
requires_pytorch(self)
class ConvBertForSequenceClassification:
def __init__(self, *args, **kwargs):
requires_pytorch(self)
@classmethod
def from_pretrained(self, *args, **kwargs):
requires_pytorch(self)
class ConvBertForTokenClassification:
def __init__(self, *args, **kwargs):
requires_pytorch(self)
@classmethod
def from_pretrained(self, *args, **kwargs):
requires_pytorch(self)
class ConvBertLayer:
def __init__(self, *args, **kwargs):
requires_pytorch(self)
class ConvBertModel:
def __init__(self, *args, **kwargs):
requires_pytorch(self)
@classmethod
def from_pretrained(self, *args, **kwargs):
requires_pytorch(self)
class ConvBertPreTrainedModel:
def __init__(self, *args, **kwargs):
requires_pytorch(self)
@classmethod
def from_pretrained(self, *args, **kwargs):
requires_pytorch(self)
def load_tf_weights_in_convbert(*args, **kwargs):
requires_pytorch(load_tf_weights_in_convbert)
CTRL_PRETRAINED_MODEL_ARCHIVE_LIST = None
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment