From 13e736685ac3482c7a1454800947912835f297eb Mon Sep 17 00:00:00 2001 From: Kamal Raj Kanakarajan Date: Mon, 5 Dec 2022 20:42:03 +0530 Subject: [PATCH] Add BioGPT (#20420) * biogpt initial commit * updated init * fix faster decoding with use_cache * 1. fix input_ids and input_embeds with correct device 2. added _keys_to_ignore_on_load_missing 3. updated prepare_inputs_for_generation * add activation_dropout and scale_embedding * replace fsmt attention with bart attention * added test * run make fix-copies * doc init and fix build * updated README with proper information * 1. added tips to docs 2. updated BioGptTokenizer func * 1. added tokenizer test 2. refactor tokenizer * make fixup * add biogpt fairseq to hf converter * updated layer names more similar to original checkpoints * config update doc string and set defaults * added "#copied" from bart model and updated doc strings * enable model_input_names in tokenizer * 1. positionalembedding depending on attention_mask 2. added attention mask to prepare for generation * added test to verify past and generation * BioGptLMHeadModel -> BioGptForCausalLM * fix typo * tokenization and test Copyright and updated assertion * updated Copyright and one func at time in line * Copyright updates and minor doc fix * replace assertion with ValueError * rm extra space * added code syntax * revert cmnt position change * add tokenizer to auto * updated doc string * tokenizer doc string update * biogpt hub model update to microsoft/biogpt * make fixup * rm cmnt to fix flake8 5.0.4 vs 6 error --- README.md | 1 + README_es.md | 1 + README_ja.md | 1 + README_ko.md | 1 + README_zh-hans.md | 1 + README_zh-hant.md | 1 + docs/source/en/_toctree.yml | 2 + docs/source/en/index.mdx | 2 + docs/source/en/model_doc/biogpt.mdx | 52 ++ src/transformers/__init__.py | 16 + src/transformers/models/__init__.py | 1 + .../models/auto/configuration_auto.py | 3 + src/transformers/models/auto/modeling_auto.py | 2 + .../models/auto/tokenization_auto.py | 1 + src/transformers/models/biogpt/__init__.py | 64 ++ .../models/biogpt/configuration_biogpt.py | 138 ++++ ..._original_pytorch_checkpoint_to_pytorch.py | 293 +++++++ .../models/biogpt/modeling_biogpt.py | 720 ++++++++++++++++++ .../models/biogpt/tokenization_biogpt.py | 370 +++++++++ src/transformers/utils/dummy_pt_objects.py | 24 + tests/models/biogpt/__init__.py | 0 tests/models/biogpt/test_modeling_biogpt.py | 398 ++++++++++ .../models/biogpt/test_tokenization_biogpt.py | 97 +++ 23 files changed, 2189 insertions(+) create mode 100644 docs/source/en/model_doc/biogpt.mdx create mode 100644 src/transformers/models/biogpt/__init__.py create mode 100644 src/transformers/models/biogpt/configuration_biogpt.py create mode 100755 src/transformers/models/biogpt/convert_biogpt_original_pytorch_checkpoint_to_pytorch.py create mode 100755 src/transformers/models/biogpt/modeling_biogpt.py create mode 100644 src/transformers/models/biogpt/tokenization_biogpt.py create mode 100644 tests/models/biogpt/__init__.py create mode 100644 tests/models/biogpt/test_modeling_biogpt.py create mode 100644 tests/models/biogpt/test_tokenization_biogpt.py diff --git a/README.md b/README.md index 22056f7395..8f010ff629 100644 --- a/README.md +++ b/README.md @@ -272,6 +272,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h 1. **[BERTweet](https://huggingface.co/docs/transformers/model_doc/bertweet)** (from VinAI Research) released with the paper [BERTweet: A pre-trained language model for English Tweets](https://aclanthology.org/2020.emnlp-demos.2/) by Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen. 1. **[BigBird-Pegasus](https://huggingface.co/docs/transformers/model_doc/bigbird_pegasus)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. 1. **[BigBird-RoBERTa](https://huggingface.co/docs/transformers/model_doc/big_bird)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. +1. **[BioGpt](https://huggingface.co/docs/transformers/main/model_doc/biogpt)** (from Microsoft Research AI4Science) released with the paper [BioGPT: generative pre-trained transformer for biomedical text generation and mining](https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbac409/6713511?guestAccessKey=a66d9b5d-4f83-4017-bb52-405815c907b9) by Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu. 1. **[Blenderbot](https://huggingface.co/docs/transformers/model_doc/blenderbot)** (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/docs/transformers/model_doc/blenderbot-small)** (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. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/). diff --git a/README_es.md b/README_es.md index f132aad1e4..f59a90b834 100644 --- a/README_es.md +++ b/README_es.md @@ -272,6 +272,7 @@ Número actual de puntos de control: ![](https://img.shields.io/endpoint?url=htt 1. **[BERTweet](https://huggingface.co/docs/transformers/model_doc/bertweet)** (from VinAI Research) released with the paper [BERTweet: A pre-trained language model for English Tweets](https://aclanthology.org/2020.emnlp-demos.2/) by Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen. 1. **[BigBird-Pegasus](https://huggingface.co/docs/transformers/model_doc/bigbird_pegasus)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. 1. **[BigBird-RoBERTa](https://huggingface.co/docs/transformers/model_doc/big_bird)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. +1. **[BioGpt](https://huggingface.co/docs/transformers/main/model_doc/biogpt)** (from Microsoft Research AI4Science) released with the paper [BioGPT: generative pre-trained transformer for biomedical text generation and mining](https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbac409/6713511?guestAccessKey=a66d9b5d-4f83-4017-bb52-405815c907b9) by Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu. 1. **[Blenderbot](https://huggingface.co/docs/transformers/model_doc/blenderbot)** (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/docs/transformers/model_doc/blenderbot-small)** (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. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/). diff --git a/README_ja.md b/README_ja.md index 6062c458ac..44c9f9fc12 100644 --- a/README_ja.md +++ b/README_ja.md @@ -307,6 +307,7 @@ Flax、PyTorch、TensorFlowをcondaでインストールする方法は、それ 1. **[BERTweet](https://huggingface.co/docs/transformers/model_doc/bertweet)** (from VinAI Research) released with the paper [BERTweet: A pre-trained language model for English Tweets](https://aclanthology.org/2020.emnlp-demos.2/) by Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen. 1. **[BigBird-Pegasus](https://huggingface.co/docs/transformers/model_doc/bigbird_pegasus)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. 1. **[BigBird-RoBERTa](https://huggingface.co/docs/transformers/model_doc/big_bird)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. +1. **[BioGpt](https://huggingface.co/docs/transformers/main/model_doc/biogpt)** (from Microsoft Research AI4Science) released with the paper [BioGPT: generative pre-trained transformer for biomedical text generation and mining](https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbac409/6713511?guestAccessKey=a66d9b5d-4f83-4017-bb52-405815c907b9) by Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu. 1. **[Blenderbot](https://huggingface.co/docs/transformers/model_doc/blenderbot)** (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/docs/transformers/model_doc/blenderbot-small)** (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. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/). diff --git a/README_ko.md b/README_ko.md index d63ebc3592..c00d039cbe 100644 --- a/README_ko.md +++ b/README_ko.md @@ -222,6 +222,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는 1. **[BERTweet](https://huggingface.co/docs/transformers/model_doc/bertweet)** (from VinAI Research) released with the paper [BERTweet: A pre-trained language model for English Tweets](https://aclanthology.org/2020.emnlp-demos.2/) by Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen. 1. **[BigBird-Pegasus](https://huggingface.co/docs/transformers/model_doc/bigbird_pegasus)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. 1. **[BigBird-RoBERTa](https://huggingface.co/docs/transformers/model_doc/big_bird)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. +1. **[BioGpt](https://huggingface.co/docs/transformers/main/model_doc/biogpt)** (from Microsoft Research AI4Science) released with the paper [BioGPT: generative pre-trained transformer for biomedical text generation and mining](https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbac409/6713511?guestAccessKey=a66d9b5d-4f83-4017-bb52-405815c907b9) by Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu. 1. **[Blenderbot](https://huggingface.co/docs/transformers/model_doc/blenderbot)** (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/docs/transformers/model_doc/blenderbot-small)** (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. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/). diff --git a/README_zh-hans.md b/README_zh-hans.md index e226ddf2fa..8bd1c4e5ee 100644 --- a/README_zh-hans.md +++ b/README_zh-hans.md @@ -246,6 +246,7 @@ conda install -c huggingface transformers 1. **[BERTweet](https://huggingface.co/docs/transformers/model_doc/bertweet)** (来自 VinAI Research) 伴随论文 [BERTweet: A pre-trained language model for English Tweets](https://aclanthology.org/2020.emnlp-demos.2/) 由 Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen 发布。 1. **[BigBird-Pegasus](https://huggingface.co/docs/transformers/model_doc/bigbird_pegasus)** (来自 Google Research) 伴随论文 [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) 由 Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed 发布。 1. **[BigBird-RoBERTa](https://huggingface.co/docs/transformers/model_doc/big_bird)** (来自 Google Research) 伴随论文 [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) 由 Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed 发布。 +1. **[BioGpt](https://huggingface.co/docs/transformers/main/model_doc/biogpt)** (来自 Microsoft Research AI4Science) 伴随论文 [BioGPT: generative pre-trained transformer for biomedical text generation and mining](https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbac409/6713511?guestAccessKey=a66d9b5d-4f83-4017-bb52-405815c907b9) 由 Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu 发布。 1. **[Blenderbot](https://huggingface.co/docs/transformers/model_doc/blenderbot)** (来自 Facebook) 伴随论文 [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) 由 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 发布。 1. **[BlenderbotSmall](https://huggingface.co/docs/transformers/model_doc/blenderbot-small)** (来自 Facebook) 伴随论文 [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) 由 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 发布。 1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/). diff --git a/README_zh-hant.md b/README_zh-hant.md index 237afce47a..c0bf3a1119 100644 --- a/README_zh-hant.md +++ b/README_zh-hant.md @@ -258,6 +258,7 @@ conda install -c huggingface transformers 1. **[BERTweet](https://huggingface.co/docs/transformers/model_doc/bertweet)** (from VinAI Research) released with the paper [BERTweet: A pre-trained language model for English Tweets](https://aclanthology.org/2020.emnlp-demos.2/) by Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen. 1. **[BigBird-Pegasus](https://huggingface.co/docs/transformers/model_doc/bigbird_pegasus)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. 1. **[BigBird-RoBERTa](https://huggingface.co/docs/transformers/model_doc/big_bird)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. +1. **[BioGpt](https://huggingface.co/docs/transformers/main/model_doc/biogpt)** (from Microsoft Research AI4Science) released with the paper [BioGPT: generative pre-trained transformer for biomedical text generation and mining](https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbac409/6713511?guestAccessKey=a66d9b5d-4f83-4017-bb52-405815c907b9) by Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu. 1. **[Blenderbot](https://huggingface.co/docs/transformers/model_doc/blenderbot)** (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/docs/transformers/model_doc/blenderbot-small)** (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. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/). diff --git a/docs/source/en/_toctree.yml b/docs/source/en/_toctree.yml index 62cbe467d8..1669a3f0c0 100644 --- a/docs/source/en/_toctree.yml +++ b/docs/source/en/_toctree.yml @@ -209,6 +209,8 @@ title: BigBird - local: model_doc/bigbird_pegasus title: BigBirdPegasus + - local: model_doc/biogpt + title: BioGpt - local: model_doc/blenderbot title: Blenderbot - local: model_doc/blenderbot-small diff --git a/docs/source/en/index.mdx b/docs/source/en/index.mdx index fe3a2d20d9..2dc04dadd6 100644 --- a/docs/source/en/index.mdx +++ b/docs/source/en/index.mdx @@ -60,6 +60,7 @@ The documentation is organized into five sections: 1. **[BERTweet](model_doc/bertweet)** (from VinAI Research) released with the paper [BERTweet: A pre-trained language model for English Tweets](https://aclanthology.org/2020.emnlp-demos.2/) by Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen. 1. **[BigBird-Pegasus](model_doc/bigbird_pegasus)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. 1. **[BigBird-RoBERTa](model_doc/big_bird)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed. +1. **[BioGpt](model_doc/biogpt)** (from Microsoft Research AI4Science) released with the paper [BioGPT: generative pre-trained transformer for biomedical text generation and mining](https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbac409/6713511?guestAccessKey=a66d9b5d-4f83-4017-bb52-405815c907b9) by Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu. 1. **[Blenderbot](model_doc/blenderbot)** (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](model_doc/blenderbot-small)** (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. **[BLOOM](model_doc/bloom)** (from BigScience workshop) released by the [BigScience Workshop](https://bigscience.huggingface.co/). @@ -229,6 +230,7 @@ Flax), PyTorch, and/or TensorFlow. | Bert Generation | ✅ | ❌ | ✅ | ❌ | ❌ | | BigBird | ✅ | ✅ | ✅ | ❌ | ✅ | | BigBird-Pegasus | ❌ | ❌ | ✅ | ❌ | ❌ | +| BioGpt | ✅ | ❌ | ✅ | ❌ | ❌ | | Blenderbot | ✅ | ✅ | ✅ | ✅ | ✅ | | BlenderbotSmall | ✅ | ✅ | ✅ | ✅ | ✅ | | BLOOM | ❌ | ✅ | ✅ | ❌ | ❌ | diff --git a/docs/source/en/model_doc/biogpt.mdx b/docs/source/en/model_doc/biogpt.mdx new file mode 100644 index 0000000000..84bd96d768 --- /dev/null +++ b/docs/source/en/model_doc/biogpt.mdx @@ -0,0 +1,52 @@ + + +# BioGPT + +## Overview + +The BioGPT model was proposed in [BioGPT: generative pre-trained transformer for biomedical text generation and mining +](https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbac409/6713511?guestAccessKey=a66d9b5d-4f83-4017-bb52-405815c907b9) by Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon and Tie-Yan Liu. BioGPT is a domain-specific generative pre-trained Transformer language model for biomedical text generation and mining. BioGPT follows the Transformer language model backbone, and is pre-trained on 15M PubMed abstracts from scratch. + +The abstract from the paper is the following: + +*Pre-trained language models have attracted increasing attention in the biomedical domain, inspired by their great success in the general natural language domain. Among the two main branches of pre-trained language models in the general language domain, i.e. BERT (and its variants) and GPT (and its variants), the first one has been extensively studied in the biomedical domain, such as BioBERT and PubMedBERT. While they have achieved great success on a variety of discriminative downstream biomedical tasks, the lack of generation ability constrains their application scope. In this paper, we propose BioGPT, a domain-specific generative Transformer language model pre-trained on large-scale biomedical literature. We evaluate BioGPT on six biomedical natural language processing tasks and demonstrate that our model outperforms previous models on most tasks. Especially, we get 44.98%, 38.42% and 40.76% F1 score on BC5CDR, KD-DTI and DDI end-to-end relation extraction tasks, respectively, and 78.2% accuracy on PubMedQA, creating a new record. Our case study on text generation further demonstrates the advantage of BioGPT on biomedical literature to generate fluent descriptions for biomedical terms.* + +Tips: + +- BioGPT is a model with absolute position embeddings so it’s usually advised to pad the inputs on the right rather than the left. +- BioGPT was trained with a causal language modeling (CLM) objective and is therefore powerful at predicting the next token in a sequence. Leveraging this feature allows BioGPT to generate syntactically coherent text as it can be observed in the run_generation.py example script. +- The model can take the `past_key_values` (for PyTorch) as input, which is the previously computed key/value attention pairs. Using this (past_key_values or past) value prevents the model from re-computing pre-computed values in the context of text generation. For PyTorch, see past_key_values argument of the BioGptForCausalLM.forward() method for more information on its usage. + +This model was contributed by [kamalkraj](https://huggingface.co/kamalkraj). The original code can be found [here](https://github.com/microsoft/BioGPT). + +## BioGptConfig + +[[autodoc]] BioGptConfig + + +## BioGptTokenizer + +[[autodoc]] BioGptTokenizer + - save_vocabulary + + +## BioGptModel + +[[autodoc]] BioGptModel + - forward + + +## BioGptForCausalLM + +[[autodoc]] BioGptForCausalLM + - forward \ No newline at end of file diff --git a/src/transformers/__init__.py b/src/transformers/__init__.py index bf630e8a9c..2e477e7a09 100644 --- a/src/transformers/__init__.py +++ b/src/transformers/__init__.py @@ -160,6 +160,7 @@ _import_structure = { "BIGBIRD_PEGASUS_PRETRAINED_CONFIG_ARCHIVE_MAP", "BigBirdPegasusConfig", ], + "models.biogpt": ["BIOGPT_PRETRAINED_CONFIG_ARCHIVE_MAP", "BioGptConfig", "BioGptTokenizer"], "models.blenderbot": ["BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP", "BlenderbotConfig", "BlenderbotTokenizer"], "models.blenderbot_small": [ "BLENDERBOT_SMALL_PRETRAINED_CONFIG_ARCHIVE_MAP", @@ -1046,6 +1047,14 @@ else: "BigBirdPegasusPreTrainedModel", ] ) + _import_structure["models.biogpt"].extend( + [ + "BIOGPT_PRETRAINED_MODEL_ARCHIVE_LIST", + "BioGptForCausalLM", + "BioGptModel", + "BioGptPreTrainedModel", + ] + ) _import_structure["models.blenderbot"].extend( [ "BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST", @@ -3392,6 +3401,7 @@ if TYPE_CHECKING: from .models.bertweet import BertweetTokenizer from .models.big_bird import BIG_BIRD_PRETRAINED_CONFIG_ARCHIVE_MAP, BigBirdConfig from .models.bigbird_pegasus import BIGBIRD_PEGASUS_PRETRAINED_CONFIG_ARCHIVE_MAP, BigBirdPegasusConfig + from .models.biogpt import BIOGPT_PRETRAINED_CONFIG_ARCHIVE_MAP, BioGptConfig, BioGptTokenizer from .models.blenderbot import BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP, BlenderbotConfig, BlenderbotTokenizer from .models.blenderbot_small import ( BLENDERBOT_SMALL_PRETRAINED_CONFIG_ARCHIVE_MAP, @@ -4166,6 +4176,12 @@ if TYPE_CHECKING: BigBirdPegasusModel, BigBirdPegasusPreTrainedModel, ) + from .models.biogpt import ( + BIOGPT_PRETRAINED_MODEL_ARCHIVE_LIST, + BioGptForCausalLM, + BioGptModel, + BioGptPreTrainedModel, + ) from .models.blenderbot import ( BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST, BlenderbotForCausalLM, diff --git a/src/transformers/models/__init__.py b/src/transformers/models/__init__.py index 774ac3eb1e..11498fd91e 100644 --- a/src/transformers/models/__init__.py +++ b/src/transformers/models/__init__.py @@ -30,6 +30,7 @@ from . import ( bertweet, big_bird, bigbird_pegasus, + biogpt, blenderbot, blenderbot_small, bloom, diff --git a/src/transformers/models/auto/configuration_auto.py b/src/transformers/models/auto/configuration_auto.py index d2c322c5b1..3f783caebb 100644 --- a/src/transformers/models/auto/configuration_auto.py +++ b/src/transformers/models/auto/configuration_auto.py @@ -37,6 +37,7 @@ CONFIG_MAPPING_NAMES = OrderedDict( ("bert-generation", "BertGenerationConfig"), ("big_bird", "BigBirdConfig"), ("bigbird_pegasus", "BigBirdPegasusConfig"), + ("biogpt", "BioGptConfig"), ("blenderbot", "BlenderbotConfig"), ("blenderbot-small", "BlenderbotSmallConfig"), ("bloom", "BloomConfig"), @@ -189,6 +190,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict( ("bert", "BERT_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("big_bird", "BIG_BIRD_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("bigbird_pegasus", "BIGBIRD_PEGASUS_PRETRAINED_CONFIG_ARCHIVE_MAP"), + ("biogpt", "BIOGPT_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("blenderbot", "BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("blenderbot-small", "BLENDERBOT_SMALL_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("bloom", "BLOOM_PRETRAINED_CONFIG_ARCHIVE_MAP"), @@ -331,6 +333,7 @@ MODEL_NAMES_MAPPING = OrderedDict( ("bertweet", "BERTweet"), ("big_bird", "BigBird"), ("bigbird_pegasus", "BigBird-Pegasus"), + ("biogpt", "BioGpt"), ("blenderbot", "Blenderbot"), ("blenderbot-small", "BlenderbotSmall"), ("bloom", "BLOOM"), diff --git a/src/transformers/models/auto/modeling_auto.py b/src/transformers/models/auto/modeling_auto.py index bb67dbfc5d..52329c49c2 100644 --- a/src/transformers/models/auto/modeling_auto.py +++ b/src/transformers/models/auto/modeling_auto.py @@ -36,6 +36,7 @@ MODEL_MAPPING_NAMES = OrderedDict( ("bert-generation", "BertGenerationEncoder"), ("big_bird", "BigBirdModel"), ("bigbird_pegasus", "BigBirdPegasusModel"), + ("biogpt", "BioGptModel"), ("blenderbot", "BlenderbotModel"), ("blenderbot-small", "BlenderbotSmallModel"), ("bloom", "BloomModel"), @@ -308,6 +309,7 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict( ("bert-generation", "BertGenerationDecoder"), ("big_bird", "BigBirdForCausalLM"), ("bigbird_pegasus", "BigBirdPegasusForCausalLM"), + ("biogpt", "BioGptForCausalLM"), ("blenderbot", "BlenderbotForCausalLM"), ("blenderbot-small", "BlenderbotSmallForCausalLM"), ("bloom", "BloomForCausalLM"), diff --git a/src/transformers/models/auto/tokenization_auto.py b/src/transformers/models/auto/tokenization_auto.py index 6cffb8e53f..e2fa1ddd7c 100644 --- a/src/transformers/models/auto/tokenization_auto.py +++ b/src/transformers/models/auto/tokenization_auto.py @@ -74,6 +74,7 @@ else: ), ), ("bigbird_pegasus", ("PegasusTokenizer", "PegasusTokenizerFast" if is_tokenizers_available() else None)), + ("biogpt", ("BioGptTokenizer", None)), ("blenderbot", ("BlenderbotTokenizer", "BlenderbotTokenizerFast")), ("blenderbot-small", ("BlenderbotSmallTokenizer", None)), ("bloom", (None, "BloomTokenizerFast" if is_tokenizers_available() else None)), diff --git a/src/transformers/models/biogpt/__init__.py b/src/transformers/models/biogpt/__init__.py new file mode 100644 index 0000000000..90d1f4b40b --- /dev/null +++ b/src/transformers/models/biogpt/__init__.py @@ -0,0 +1,64 @@ +# 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 2022 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 + +# rely on isort to merge the imports +from ...utils import OptionalDependencyNotAvailable, _LazyModule, is_tokenizers_available, is_torch_available + + +_import_structure = { + "configuration_biogpt": ["BIOGPT_PRETRAINED_CONFIG_ARCHIVE_MAP", "BioGptConfig"], + "tokenization_biogpt": ["BioGptTokenizer"], +} + +try: + if not is_torch_available(): + raise OptionalDependencyNotAvailable() +except OptionalDependencyNotAvailable: + pass +else: + _import_structure["modeling_biogpt"] = [ + "BIOGPT_PRETRAINED_MODEL_ARCHIVE_LIST", + "BioGptForCausalLM", + "BioGptModel", + "BioGptPreTrainedModel", + ] + + +if TYPE_CHECKING: + from .configuration_biogpt import BIOGPT_PRETRAINED_CONFIG_ARCHIVE_MAP, BioGptConfig + from .tokenization_biogpt import BioGptTokenizer + + try: + if not is_torch_available(): + raise OptionalDependencyNotAvailable() + except OptionalDependencyNotAvailable: + pass + else: + from .modeling_biogpt import ( + BIOGPT_PRETRAINED_MODEL_ARCHIVE_LIST, + BioGptForCausalLM, + BioGptModel, + BioGptPreTrainedModel, + ) + + +else: + import sys + + sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__) diff --git a/src/transformers/models/biogpt/configuration_biogpt.py b/src/transformers/models/biogpt/configuration_biogpt.py new file mode 100644 index 0000000000..4803b9dc12 --- /dev/null +++ b/src/transformers/models/biogpt/configuration_biogpt.py @@ -0,0 +1,138 @@ +# coding=utf-8 +# Copyright 2022 The HuggingFace Team and Microsoft Research AI4Science 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. +""" BioGPT model configuration""" + +from ...configuration_utils import PretrainedConfig +from ...utils import logging + + +logger = logging.get_logger(__name__) + +BIOGPT_PRETRAINED_CONFIG_ARCHIVE_MAP = { + "microsoft/biogpt": "https://huggingface.co/microsoft/biogpt/resolve/main/config.json", + # See all BioGPT models at https://huggingface.co/models?filter=biogpt +} + + +class BioGptConfig(PretrainedConfig): + r""" + This is the configuration class to store the configuration of a [`BioGptModel`]. It is used to instantiate an + BioGPT 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 BioGPT + [microsoft/biogpt](https://huggingface.co/microsoft/biogpt) architecture. + + Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the + documentation from [`PretrainedConfig`] for more information. + + + Args: + vocab_size (`int`, *optional*, defaults to 42384): + Vocabulary size of the BioGPT model. Defines the number of different tokens that can be represented by the + `inputs_ids` passed when calling [`BioGptModel`]. + hidden_size (`int`, *optional*, defaults to 1024): + Dimension of the encoder layers and the pooler layer. + num_hidden_layers (`int`, *optional*, defaults to 24): + Number of hidden layers in the Transformer encoder. + num_attention_heads (`int`, *optional*, defaults to 16): + Number of attention heads for each attention layer in the Transformer encoder. + intermediate_size (`int`, *optional*, defaults to 4096): + Dimension of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder. + hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`): + The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`, + `"relu"`, `"selu"` and `"gelu_new"` are supported. + hidden_dropout_prob (`float`, *optional*, defaults to 0.1): + The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler. + attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1): + The dropout ratio for the attention probabilities. + max_position_embeddings (`int`, *optional*, defaults to 1024): + The maximum sequence length that this model might ever be used with. Typically set this to something large + just in case (e.g., 512 or 1024 or 2048). + initializer_range (`float`, *optional*, defaults to 0.02): + The standard deviation of the truncated_normal_initializer for initializing all weight matrices. + layer_norm_eps (`float`, *optional*, defaults to 1e-12): + The epsilon used by the layer normalization layers. + scale_embedding (`bool`, *optional*, defaults to `True`): + Scale embeddings by diving by sqrt(d_model). + use_cache (`bool`, *optional*, defaults to `True`): + Whether or not the model should return the last key/values attentions (not used by all models). Only + relevant if `config.is_decoder=True`. + is_encoder_decoder (`bool`, *optional*, defaults to `False`): + Whether this is an encoder/decoder model. + layerdrop (`float`, *optional*, defaults to 0.0): + Please refer to the paper about LayerDrop: https://arxiv.org/abs/1909.11556 for further details + activation_dropout (`float`, *optional*, defaults to 0.0): + The dropout ratio for activations inside the fully connected layer. + pad_token_id (`int`, *optional*, defaults to 1) + Padding token id. + bos_token_id (`int`, *optional*, defaults to 0) + Beginning of stream token id. + eos_token_id (`int`, *optional*, defaults to 2) + End of stream token id. + Example: + + ```python + >>> from transformers import BioGptModel, BioGptConfig + + >>> # Initializing a BioGPT microsoft/biogpt style configuration + >>> configuration = BioGptConfig() + + >>> # Initializing a model from the microsoft/biogpt style configuration + >>> model = BioGptModel(configuration) + + >>> # Accessing the model configuration + >>> configuration = model.config + ```""" + model_type = "biogpt" + + def __init__( + self, + vocab_size=42384, + hidden_size=1024, + num_hidden_layers=24, + num_attention_heads=16, + intermediate_size=4096, + hidden_act="gelu", + hidden_dropout_prob=0.1, + attention_probs_dropout_prob=0.1, + max_position_embeddings=1024, + initializer_range=0.02, + layer_norm_eps=1e-12, + scale_embedding=True, + use_cache=True, + is_encoder_decoder=False, + layerdrop=0.0, + activation_dropout=0.0, + pad_token_id=1, + bos_token_id=0, + eos_token_id=2, + **kwargs + ): + self.vocab_size = vocab_size + self.max_position_embeddings = max_position_embeddings + 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.initializer_range = initializer_range + self.layer_norm_eps = layer_norm_eps + self.scale_embedding = scale_embedding + self.use_cache = use_cache + self.is_encoder_decoder = is_encoder_decoder + self.layerdrop = layerdrop + self.activation_dropout = activation_dropout + super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs) diff --git a/src/transformers/models/biogpt/convert_biogpt_original_pytorch_checkpoint_to_pytorch.py b/src/transformers/models/biogpt/convert_biogpt_original_pytorch_checkpoint_to_pytorch.py new file mode 100755 index 0000000000..bcbda452a3 --- /dev/null +++ b/src/transformers/models/biogpt/convert_biogpt_original_pytorch_checkpoint_to_pytorch.py @@ -0,0 +1,293 @@ +# coding=utf-8 +# Copyright 2022 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 argparse +import json +import os +import re +import shutil + +import torch + +from transformers import BioGptConfig, BioGptForCausalLM +from transformers.models.biogpt.tokenization_biogpt import VOCAB_FILES_NAMES +from transformers.tokenization_utils_base import TOKENIZER_CONFIG_FILE +from transformers.utils import WEIGHTS_NAME, logging + + +logging.set_verbosity_warning() + +json_indent = 2 + + +# modified from https://github.com/facebookresearch/fairseq/blob/dd74992d0d143155998e9ed4076826bcea80fb06/fairseq/data/dictionary.py#L18 +class Dictionary: + """A mapping from symbols to consecutive integers""" + + def __init__( + self, + *, # begin keyword-only arguments + bos="", + pad="", + eos="", + unk="", + extra_special_symbols=None, + ): + self.bos_word, self.unk_word, self.pad_word, self.eos_word = bos, unk, pad, eos + self.symbols = [] + self.count = [] + self.indices = {} + self.bos_index = self.add_symbol(bos) + self.pad_index = self.add_symbol(pad) + self.eos_index = self.add_symbol(eos) + self.unk_index = self.add_symbol(unk) + if extra_special_symbols: + for s in extra_special_symbols: + self.add_symbol(s) + self.nspecial = len(self.symbols) + + def __eq__(self, other): + return self.indices == other.indices + + def __getitem__(self, idx): + if idx < len(self.symbols): + return self.symbols[idx] + return self.unk_word + + def __len__(self): + """Returns the number of symbols in the dictionary""" + return len(self.symbols) + + def __contains__(self, sym): + return sym in self.indices + + @classmethod + def load(cls, f): + """Loads the dictionary from a text file with the format: + + ``` + + + ... + ``` + """ + d = cls() + d.add_from_file(f) + return d + + def add_symbol(self, word, n=1, overwrite=False): + """Adds a word to the dictionary""" + if word in self.indices and not overwrite: + idx = self.indices[word] + self.count[idx] = self.count[idx] + n + return idx + else: + idx = len(self.symbols) + self.indices[word] = idx + self.symbols.append(word) + self.count.append(n) + return idx + + def _load_meta(self, lines): + return 0 + + def add_from_file(self, f): + """ + Loads a pre-existing dictionary from a text file and adds its symbols to this instance. + """ + if isinstance(f, str): + try: + with open(f, "r", encoding="utf-8") as fd: + self.add_from_file(fd) + except FileNotFoundError as fnfe: + raise fnfe + except UnicodeError: + raise Exception("Incorrect encoding detected in {}, please rebuild the dataset".format(f)) + return + + lines = f.readlines() + indices_start_line = self._load_meta(lines) + + for line in lines[indices_start_line:]: + try: + line, field = line.rstrip().rsplit(" ", 1) + if field == "#fairseq:overwrite": + overwrite = True + line, field = line.rsplit(" ", 1) + else: + overwrite = False + count = int(field) + word = line + if word in self and not overwrite: + raise RuntimeError( + "Duplicate word found when loading Dictionary: '{}'. " + "Duplicate words can overwrite earlier ones by adding the " + "#fairseq:overwrite flag at the end of the corresponding row " + "in the dictionary file. If using the Camembert model, please " + "download an updated copy of the model file.".format(word) + ) + self.add_symbol(word, n=count, overwrite=overwrite) + except ValueError: + raise ValueError("Incorrect dictionary format, expected ' [flags]'") + + +def rewrite_dict_keys(d): + # (1) remove word breaking symbol, (2) add word ending symbol where the word is not broken up, + # e.g.: d = {'le@@': 5, 'tt@@': 6, 'er': 7} => {'le': 5, 'tt': 6, 'er': 7} + d2 = dict((re.sub(r"@@$", "", k), v) if k.endswith("@@") else (re.sub(r"$", "", k), v) for k, v in d.items()) + keep_keys = " ".split() + # restore the special tokens + for k in keep_keys: + del d2[f"{k}"] + d2[k] = d[k] # restore + return d2 + + +def convert_biogpt_checkpoint_to_pytorch(biogpt_checkpoint_path, pytorch_dump_folder_path): + + # prep + if not os.path.exists(biogpt_checkpoint_path): + raise ValueError(f"path {biogpt_checkpoint_path} does not exist!") + os.makedirs(pytorch_dump_folder_path, exist_ok=True) + print(f"Writing results to {pytorch_dump_folder_path}") + + # handle various types of models + + checkpoint_file = os.path.join(biogpt_checkpoint_path, "checkpoint.pt") + if not os.path.isfile(checkpoint_file): + raise ValueError(f"path to the file {checkpoint_file} does not exist!") + chkpt = torch.load(checkpoint_file, map_location="cpu") + + args = chkpt["cfg"]["model"] + + # dicts + dict_file = os.path.join(biogpt_checkpoint_path, "dict.txt") + if not os.path.isfile(dict_file): + raise ValueError(f"path to the file {dict_file} does not exist!") + src_dict = Dictionary.load(dict_file) + src_vocab = rewrite_dict_keys(src_dict.indices) + src_vocab_size = len(src_vocab) + src_vocab_file = os.path.join(pytorch_dump_folder_path, VOCAB_FILES_NAMES["vocab_file"]) + print(f"Generating {src_vocab_file} of {src_vocab_size} records") + with open(src_vocab_file, "w", encoding="utf-8") as f: + f.write(json.dumps(src_vocab, ensure_ascii=False, indent=json_indent)) + + # merges_file (bpecodes) + bpecodes_file = os.path.join(biogpt_checkpoint_path, "bpecodes") + if not os.path.isfile(bpecodes_file): + raise ValueError(f"path to the file {bpecodes_file} does not exist!") + + merges_file = os.path.join(pytorch_dump_folder_path, VOCAB_FILES_NAMES["merges_file"]) + shutil.copyfile(bpecodes_file, merges_file) + + # model config + biogpt_model_config_file = os.path.join(pytorch_dump_folder_path, "config.json") + + model_conf = { + "activation_dropout": args["activation_dropout"], + "architectures": ["BioGptForCausalLM"], + "attention_probs_dropout_prob": args["attention_dropout"], + "bos_token_id": 0, + "eos_token_id": 2, + "hidden_act": args["activation_fn"], + "hidden_dropout_prob": args["dropout"], + "hidden_size": args["decoder_embed_dim"], + "initializer_range": 0.02, + "intermediate_size": args["decoder_ffn_embed_dim"], + "layer_norm_eps": 1e-12, + "layerdrop": args["decoder_layerdrop"], + "max_position_embeddings": args["max_target_positions"], + "model_type": "biogpt", + "num_attention_heads": args["decoder_attention_heads"], + "num_hidden_layers": args["decoder_layers"], + "pad_token_id": 1, + "scale_embedding": not args["no_scale_embedding"], + "tie_word_embeddings": args["share_decoder_input_output_embed"], + "vocab_size": src_vocab_size, + } + + # good hparam defaults to start with + + print(f"Generating {biogpt_model_config_file}") + with open(biogpt_model_config_file, "w", encoding="utf-8") as f: + f.write(json.dumps(model_conf, ensure_ascii=False, indent=json_indent)) + + # tokenizer config + biogpt_tokenizer_config_file = os.path.join(pytorch_dump_folder_path, TOKENIZER_CONFIG_FILE) + + tokenizer_conf = { + "bos_token": "", + "eos_token": "", + "model_max_length": 1024, + "pad_token": "", + "special_tokens_map_file": None, + "tokenizer_class": "BioGptTokenizer", + "unk_token": "", + } + + print(f"Generating {biogpt_tokenizer_config_file}") + with open(biogpt_tokenizer_config_file, "w", encoding="utf-8") as f: + f.write(json.dumps(tokenizer_conf, ensure_ascii=False, indent=json_indent)) + + # model + model_state_dict = chkpt["model"] + + # remove unneeded keys + ignore_keys = [ + "decoder.version", + ] + for k in ignore_keys: + model_state_dict.pop(k, None) + + layer_names = list(model_state_dict.keys()) + for layer_name in layer_names: + if layer_name.endswith("output_projection.weight"): + model_state_dict[layer_name.replace("decoder.", "")] = model_state_dict.pop(layer_name) + else: + model_state_dict[layer_name.replace("decoder", "biogpt")] = model_state_dict.pop(layer_name) + + config = BioGptConfig.from_pretrained(pytorch_dump_folder_path) + model_new = BioGptForCausalLM(config) + + # check that it loads ok + model_new.load_state_dict(model_state_dict) + + # save + pytorch_weights_dump_path = os.path.join(pytorch_dump_folder_path, WEIGHTS_NAME) + print(f"Generating {pytorch_weights_dump_path}") + torch.save(model_state_dict, pytorch_weights_dump_path) + + print("Conversion is done!") + + +if __name__ == "__main__": + parser = argparse.ArgumentParser() + # Required parameters + parser.add_argument( + "--biogpt_checkpoint_path", + default=None, + type=str, + required=True, + help=( + "Path to the official PyTorch checkpoint file which is expected to reside in the dump dir with dicts," + " bpecodes, etc." + ), + ) + 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_biogpt_checkpoint_to_pytorch(args.biogpt_checkpoint_path, args.pytorch_dump_folder_path) diff --git a/src/transformers/models/biogpt/modeling_biogpt.py b/src/transformers/models/biogpt/modeling_biogpt.py new file mode 100755 index 0000000000..39add24ecb --- /dev/null +++ b/src/transformers/models/biogpt/modeling_biogpt.py @@ -0,0 +1,720 @@ +# coding=utf-8 +# Copyright 2022 The HuggingFace Team and Microsoft Research AI4Science 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. +""" PyTorch BioGPT model.""" + + +import math +import random +from typing import Optional, Tuple, Union + +import torch +import torch.utils.checkpoint +from torch import nn +from torch.nn import CrossEntropyLoss + +from ...activations import ACT2FN +from ...modeling_outputs import BaseModelOutputWithPastAndCrossAttentions, CausalLMOutputWithCrossAttentions +from ...modeling_utils import PreTrainedModel +from ...utils import add_code_sample_docstrings, add_start_docstrings, add_start_docstrings_to_model_forward, logging +from .configuration_biogpt import BioGptConfig + + +logger = logging.get_logger(__name__) + +_CHECKPOINT_FOR_DOC = "microsoft/biogpt" +_CONFIG_FOR_DOC = "BioGptConfig" +_TOKENIZER_FOR_DOC = "BioGptTokenizer" + +BIOGPT_PRETRAINED_MODEL_ARCHIVE_LIST = [ + "microsoft/biogpt", + # See all BioGPT models at https://huggingface.co/models?filter=biogpt +] + + +# Copied from transformers.models.bart.modeling_bart._make_causal_mask +def _make_causal_mask(input_ids_shape: torch.Size, dtype: torch.dtype, past_key_values_length: int = 0): + """ + Make causal mask used for bi-directional self-attention. + """ + bsz, tgt_len = input_ids_shape + mask = torch.full((tgt_len, tgt_len), torch.tensor(torch.finfo(dtype).min)) + mask_cond = torch.arange(mask.size(-1)) + mask.masked_fill_(mask_cond < (mask_cond + 1).view(mask.size(-1), 1), 0) + mask = mask.to(dtype) + + if past_key_values_length > 0: + mask = torch.cat([torch.zeros(tgt_len, past_key_values_length, dtype=dtype), mask], dim=-1) + return mask[None, None, :, :].expand(bsz, 1, tgt_len, tgt_len + past_key_values_length) + + +# Copied from transformers.models.bart.modeling_bart._expand_mask +def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None): + """ + Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`. + """ + bsz, src_len = mask.size() + tgt_len = tgt_len if tgt_len is not None else src_len + + expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype) + + inverted_mask = 1.0 - expanded_mask + + return inverted_mask.masked_fill(inverted_mask.to(torch.bool), torch.finfo(dtype).min) + + +# Copied from transformers.models.opt.modeling_opt.OPTLearnedPositionalEmbedding with OPT->BioGpt +class BioGptLearnedPositionalEmbedding(nn.Embedding): + """ + This module learns positional embeddings up to a fixed maximum size. + """ + + def __init__(self, num_embeddings: int, embedding_dim: int): + # BioGpt is set up so that if padding_idx is specified then offset the embedding ids by 2 + # and adjust num_embeddings appropriately. Other models don't have this hack + self.offset = 2 + super().__init__(num_embeddings + self.offset, embedding_dim) + + def forward(self, attention_mask: torch.LongTensor, past_key_values_length: int = 0): + """`input_ids_shape` is expected to be [bsz x seqlen].""" + attention_mask = attention_mask.long() + + # create positions depending on attention_mask + positions = (torch.cumsum(attention_mask, dim=1).type_as(attention_mask) * attention_mask).long() - 1 + + # cut positions if `past_key_values_length` is > 0 + positions = positions[:, past_key_values_length:] + + return super().forward(positions + self.offset) + + +# Copied from transformers.models.bart.modeling_bart.BartAttention with Bart->BioGpt +class BioGptAttention(nn.Module): + """Multi-headed attention from 'Attention Is All You Need' paper""" + + def __init__( + self, + embed_dim: int, + num_heads: int, + dropout: float = 0.0, + is_decoder: bool = False, + bias: bool = True, + ): + super().__init__() + self.embed_dim = embed_dim + self.num_heads = num_heads + self.dropout = dropout + self.head_dim = embed_dim // num_heads + + if (self.head_dim * num_heads) != self.embed_dim: + raise ValueError( + f"embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim}" + f" and `num_heads`: {num_heads})." + ) + self.scaling = self.head_dim**-0.5 + self.is_decoder = is_decoder + + self.k_proj = nn.Linear(embed_dim, embed_dim, bias=bias) + self.v_proj = nn.Linear(embed_dim, embed_dim, bias=bias) + self.q_proj = nn.Linear(embed_dim, embed_dim, bias=bias) + self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias) + + def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int): + return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous() + + def forward( + self, + hidden_states: torch.Tensor, + key_value_states: Optional[torch.Tensor] = None, + past_key_value: Optional[Tuple[torch.Tensor]] = None, + attention_mask: Optional[torch.Tensor] = None, + layer_head_mask: Optional[torch.Tensor] = None, + output_attentions: bool = False, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: + """Input shape: Batch x Time x Channel""" + + # if key_value_states are provided this layer is used as a cross-attention layer + # for the decoder + is_cross_attention = key_value_states is not None + + bsz, tgt_len, _ = hidden_states.size() + + # get query proj + query_states = self.q_proj(hidden_states) * self.scaling + # get key, value proj + if is_cross_attention and past_key_value is not None: + # reuse k,v, cross_attentions + key_states = past_key_value[0] + value_states = past_key_value[1] + elif is_cross_attention: + # cross_attentions + key_states = self._shape(self.k_proj(key_value_states), -1, bsz) + value_states = self._shape(self.v_proj(key_value_states), -1, bsz) + elif past_key_value is not None: + # reuse k, v, self_attention + key_states = self._shape(self.k_proj(hidden_states), -1, bsz) + value_states = self._shape(self.v_proj(hidden_states), -1, bsz) + key_states = torch.cat([past_key_value[0], key_states], dim=2) + value_states = torch.cat([past_key_value[1], value_states], dim=2) + else: + # self_attention + key_states = self._shape(self.k_proj(hidden_states), -1, bsz) + value_states = self._shape(self.v_proj(hidden_states), -1, bsz) + + if self.is_decoder: + # if cross_attention save Tuple(torch.Tensor, torch.Tensor) of all cross attention key/value_states. + # Further calls to cross_attention layer can then reuse all cross-attention + # key/value_states (first "if" case) + # if uni-directional self-attention (decoder) save Tuple(torch.Tensor, torch.Tensor) of + # all previous decoder key/value_states. Further calls to uni-directional self-attention + # can concat previous decoder key/value_states to current projected key/value_states (third "elif" case) + # if encoder bi-directional self-attention `past_key_value` is always `None` + past_key_value = (key_states, value_states) + + proj_shape = (bsz * self.num_heads, -1, self.head_dim) + query_states = self._shape(query_states, tgt_len, bsz).view(*proj_shape) + key_states = key_states.view(*proj_shape) + value_states = value_states.view(*proj_shape) + + src_len = key_states.size(1) + attn_weights = torch.bmm(query_states, key_states.transpose(1, 2)) + + if attn_weights.size() != (bsz * self.num_heads, tgt_len, src_len): + raise ValueError( + f"Attention weights should be of size {(bsz * self.num_heads, tgt_len, src_len)}, but is" + f" {attn_weights.size()}" + ) + + if attention_mask is not None: + if attention_mask.size() != (bsz, 1, tgt_len, src_len): + raise ValueError( + f"Attention mask should be of size {(bsz, 1, tgt_len, src_len)}, but is {attention_mask.size()}" + ) + attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len) + attention_mask + attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len) + + attn_weights = nn.functional.softmax(attn_weights, dim=-1) + + if layer_head_mask is not None: + if layer_head_mask.size() != (self.num_heads,): + raise ValueError( + f"Head mask for a single layer should be of size {(self.num_heads,)}, but is" + f" {layer_head_mask.size()}" + ) + attn_weights = layer_head_mask.view(1, -1, 1, 1) * attn_weights.view(bsz, self.num_heads, tgt_len, src_len) + attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len) + + if output_attentions: + # this operation is a bit awkward, but it's required to + # make sure that attn_weights keeps its gradient. + # In order to do so, attn_weights have to be reshaped + # twice and have to be reused in the following + attn_weights_reshaped = attn_weights.view(bsz, self.num_heads, tgt_len, src_len) + attn_weights = attn_weights_reshaped.view(bsz * self.num_heads, tgt_len, src_len) + else: + attn_weights_reshaped = None + + attn_probs = nn.functional.dropout(attn_weights, p=self.dropout, training=self.training) + + attn_output = torch.bmm(attn_probs, value_states) + + if attn_output.size() != (bsz * self.num_heads, tgt_len, self.head_dim): + raise ValueError( + f"`attn_output` should be of size {(bsz, self.num_heads, tgt_len, self.head_dim)}, but is" + f" {attn_output.size()}" + ) + + attn_output = attn_output.view(bsz, self.num_heads, tgt_len, self.head_dim) + attn_output = attn_output.transpose(1, 2) + + # Use the `embed_dim` from the config (stored in the class) rather than `hidden_state` because `attn_output` can be + # partitioned aross GPUs when using tensor-parallelism. + attn_output = attn_output.reshape(bsz, tgt_len, self.embed_dim) + + attn_output = self.out_proj(attn_output) + + return attn_output, attn_weights_reshaped, past_key_value + + +class BioGptDecoderLayer(nn.Module): + def __init__(self, config: BioGptConfig): + super().__init__() + self.embed_dim = config.hidden_size + + self.self_attn = BioGptAttention( + embed_dim=self.embed_dim, + num_heads=config.num_attention_heads, + dropout=config.attention_probs_dropout_prob, + is_decoder=True, + ) + self.dropout = config.hidden_dropout_prob + self.activation_fn = ACT2FN[config.hidden_act] + self.activation_dropout = config.activation_dropout + + self.self_attn_layer_norm = nn.LayerNorm(self.embed_dim) + + self.fc1 = nn.Linear(self.embed_dim, config.intermediate_size) + self.fc2 = nn.Linear(config.intermediate_size, self.embed_dim) + self.final_layer_norm = nn.LayerNorm(self.embed_dim) + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.Tensor] = None, + layer_head_mask: Optional[torch.Tensor] = None, + past_key_value: Optional[Tuple[torch.Tensor]] = None, + output_attentions: Optional[bool] = False, + use_cache: Optional[bool] = True, + ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]: + """ + Args: + hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` + attention_mask (`torch.FloatTensor`): attention mask of size + `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values. + layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size + `(encoder_attention_heads,)`. + past_key_value (`Tuple(torch.FloatTensor)`): cached past key and value projection states + output_attentions (`bool`, *optional*): + Whether or not to return the attentions tensors of all attention layers. See `attentions` under + returned tensors for more detail. + use_cache (`bool`, *optional*): + If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding + (see `past_key_values`). + """ + residual = hidden_states + + hidden_states = self.self_attn_layer_norm(hidden_states) + + # Self Attention + # decoder uni-directional self-attention cached key/values tuple is at positions 1,2 + self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None + # add present self-attn cache to positions 1,2 of present_key_value tuple + hidden_states, self_attn_weights, present_key_value = self.self_attn( + hidden_states=hidden_states, + past_key_value=self_attn_past_key_value, + attention_mask=attention_mask, + layer_head_mask=layer_head_mask, + output_attentions=output_attentions, + ) + hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training) + hidden_states = residual + hidden_states + + # Fully Connected + residual = hidden_states + hidden_states = self.final_layer_norm(hidden_states) + hidden_states = self.fc1(hidden_states) + hidden_states = self.activation_fn(hidden_states) + hidden_states = nn.functional.dropout(hidden_states, p=self.activation_dropout, training=self.training) + hidden_states = self.fc2(hidden_states) + hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training) + hidden_states = residual + hidden_states + + outputs = (hidden_states,) + + if output_attentions: + outputs += (self_attn_weights,) + + if use_cache: + outputs += (present_key_value,) + + return outputs + + +class BioGptPreTrainedModel(PreTrainedModel): + """ + An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained + models. + """ + + config_class = BioGptConfig + base_model_prefix = "biogpt" + supports_gradient_checkpointing = True + + def _init_weights(self, module): + """Initialize the weights""" + if isinstance(module, nn.Linear): + # Slightly different from the TF version which uses truncated_normal for initialization + # cf https://github.com/pytorch/pytorch/pull/5617 + module.weight.data.normal_(mean=0.0, std=self.config.initializer_range) + if module.bias is not None: + module.bias.data.zero_() + elif isinstance(module, nn.Embedding): + module.weight.data.normal_(mean=0.0, std=self.config.initializer_range) + if module.padding_idx is not None: + module.weight.data[module.padding_idx].zero_() + elif isinstance(module, nn.LayerNorm): + module.bias.data.zero_() + module.weight.data.fill_(1.0) + + def _set_gradient_checkpointing(self, module, value=False): + if isinstance(module, BioGptModel): + module.gradient_checkpointing = value + + +BIOGPT_START_DOCSTRING = r""" + This model is a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) sub-class. Use + it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and + behavior. + + Parameters: + config ([`~BioGptConfig`]): Model configuration class with all the parameters of the model. + Initializing with a config file does not load the weights associated with the model, only the + configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights. +""" + +BIOGPT_INPUTS_DOCSTRING = r""" + Args: + input_ids (`torch.LongTensor` of shape `({0})`): + Indices of input sequence tokens in the vocabulary. + + Indices can be obtained using [`BioGptTokenizer`]. See [`PreTrainedTokenizer.encode`] and + [`PreTrainedTokenizer.__call__`] for details. + + [What are input IDs?](../glossary#input-ids) + attention_mask (`torch.FloatTensor` of shape `({0})`, *optional*): + Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`: + + - 1 for tokens that are **not masked**, + - 0 for tokens that are **masked**. + + [What are attention masks?](../glossary#attention-mask) + head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*): + Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`: + + - 1 indicates the head is **not masked**, + - 0 indicates the head is **masked**. + + inputs_embeds (`torch.FloatTensor` of shape `({0}, hidden_size)`, *optional*): + Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This + is useful if you want more control over how to convert *input_ids* indices into associated vectors than the + model's internal embedding lookup matrix. + past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`): + Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape + `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape + `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`. + + Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention + blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. + + If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that + don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all + `decoder_input_ids` of shape `(batch_size, sequence_length)`. inputs_embeds (`torch.FloatTensor` of shape + `(batch_size, sequence_length, hidden_size)`, *optional*): Optionally, instead of passing `input_ids` you + can choose to directly pass an embedded representation. This is useful if you want more control over how to + convert `input_ids` indices into associated vectors than the model's internal embedding lookup matrix. + use_cache (`bool`, *optional*): + If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see + `past_key_values`). + output_attentions (`bool`, *optional*): + Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned + tensors for more detail. + output_hidden_states (`bool`, *optional*): + Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for + more detail. + return_dict (`bool`, *optional*): + Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. +""" + + +@add_start_docstrings( + "The bare BioGPT Model transformer outputting raw hidden-states without any specific head on top.", + BIOGPT_START_DOCSTRING, +) +class BioGptModel(BioGptPreTrainedModel): + def __init__(self, config: BioGptConfig): + super().__init__(config) + self.config = config + self.layerdrop = config.layerdrop + self.dropout = config.hidden_dropout_prob + self.embed_dim = config.hidden_size + self.padding_idx = config.pad_token_id + self.embed_scale = math.sqrt(config.hidden_size) if config.scale_embedding else 1.0 + + self.embed_tokens = nn.Embedding(config.vocab_size, self.embed_dim, self.padding_idx) + self.embed_positions = BioGptLearnedPositionalEmbedding(config.max_position_embeddings, self.embed_dim) + + self.layers = nn.ModuleList([BioGptDecoderLayer(config) for _ in range(config.num_hidden_layers)]) + self.layer_norm = nn.LayerNorm(self.embed_dim) + + self.gradient_checkpointing = False + # Initialize weights and apply final processing + self.post_init() + + def get_input_embeddings(self): + return self.embed_tokens + + def set_input_embeddings(self, value): + self.embed_tokens = value + + # Copied from transformers.models.bart.modeling_bart.BartDecoder._prepare_decoder_attention_mask + def _prepare_decoder_attention_mask(self, attention_mask, input_shape, inputs_embeds, past_key_values_length): + # create causal mask + # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len] + combined_attention_mask = None + if input_shape[-1] > 1: + combined_attention_mask = _make_causal_mask( + input_shape, inputs_embeds.dtype, past_key_values_length=past_key_values_length + ).to(inputs_embeds.device) + + if attention_mask is not None: + # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len] + expanded_attn_mask = _expand_mask(attention_mask, inputs_embeds.dtype, tgt_len=input_shape[-1]).to( + inputs_embeds.device + ) + combined_attention_mask = ( + expanded_attn_mask if combined_attention_mask is None else expanded_attn_mask + combined_attention_mask + ) + + return combined_attention_mask + + @add_start_docstrings_to_model_forward(BIOGPT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @add_code_sample_docstrings( + processor_class=_TOKENIZER_FOR_DOC, + checkpoint=_CHECKPOINT_FOR_DOC, + output_type=BaseModelOutputWithPastAndCrossAttentions, + config_class=_CONFIG_FOR_DOC, + ) + def forward( + self, + input_ids: Optional[torch.LongTensor] = None, + attention_mask: Optional[torch.FloatTensor] = None, + head_mask: Optional[torch.FloatTensor] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple, BaseModelOutputWithPastAndCrossAttentions]: + output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + use_cache = use_cache if use_cache is not None else self.config.use_cache + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + # retrieve input_ids and inputs_embeds + if input_ids is not None and inputs_embeds is not None: + raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time") + elif input_ids is not None: + input = input_ids + input_shape = input.size() + elif inputs_embeds is not None: + input_shape = inputs_embeds.size()[:-1] + input = inputs_embeds[:, :, -1] + else: + raise ValueError("You have to specify either input_ids or inputs_embeds") + + # past_key_values_length + past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0 + + if inputs_embeds is None: + inputs_embeds = self.embed_tokens(input) * self.embed_scale + + if attention_mask is None: + attention_mask = torch.ones(inputs_embeds.shape[:2], dtype=torch.bool, device=inputs_embeds.device) + # embed positions + positions = self.embed_positions(attention_mask, past_key_values_length) + + attention_mask = self._prepare_decoder_attention_mask( + attention_mask, input_shape, inputs_embeds, past_key_values_length + ) + + hidden_states = inputs_embeds + positions + + hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training) + + all_hidden_states = () if output_hidden_states else None + all_self_attns = () if output_attentions else None + all_cross_attentions = None + next_decoder_cache = () if use_cache else None + + for idx, decoder_layer in enumerate(self.layers): + # add LayerDrop (see https://arxiv.org/abs/1909.11556 for description) + if output_hidden_states: + all_hidden_states += (hidden_states,) + dropout_probability = random.uniform(0, 1) + if self.training and (dropout_probability < self.layerdrop): + continue + + past_key_value = past_key_values[idx] if past_key_values is not None else None + + if self.gradient_checkpointing and self.training: + + if use_cache: + logger.warning( + "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..." + ) + use_cache = False + + def create_custom_forward(module): + def custom_forward(*inputs): + # None for past_key_value + return module(*inputs, output_attentions, use_cache) + + return custom_forward + + layer_outputs = torch.utils.checkpoint.checkpoint( + create_custom_forward(decoder_layer), + hidden_states, + attention_mask, + head_mask[idx] if head_mask is not None else None, + None, + ) + else: + + layer_outputs = decoder_layer( + hidden_states, + attention_mask=attention_mask, + layer_head_mask=(head_mask[idx] if head_mask is not None else None), + past_key_value=past_key_value, + output_attentions=output_attentions, + use_cache=use_cache, + ) + + hidden_states = layer_outputs[0] + + if use_cache: + next_decoder_cache += (layer_outputs[2 if output_attentions else 1],) + + if output_attentions: + all_self_attns += (layer_outputs[1],) + + # add hidden states from the last decoder layer + if output_hidden_states: + all_hidden_states += (hidden_states,) + + hidden_states = self.layer_norm(hidden_states) + + next_cache = next_decoder_cache if use_cache else None + + if not return_dict: + return tuple( + v + for v in [hidden_states, next_cache, all_hidden_states, all_self_attns, all_cross_attentions] + if v is not None + ) + return BaseModelOutputWithPastAndCrossAttentions( + last_hidden_state=hidden_states, + past_key_values=next_cache, + hidden_states=all_hidden_states, + attentions=all_self_attns, + cross_attentions=all_cross_attentions, + ) + + +@add_start_docstrings( + """BioGPT Model with a `language modeling` head on top for CLM fine-tuning.""", BIOGPT_START_DOCSTRING +) +class BioGptForCausalLM(BioGptPreTrainedModel): + _keys_to_ignore_on_load_missing = ["output_projection.weight"] + + def __init__(self, config): + super().__init__(config) + + self.biogpt = BioGptModel(config) + self.output_projection = nn.Linear(config.hidden_size, config.vocab_size, bias=False) + + # Initialize weights and apply final processing + self.post_init() + + def get_output_embeddings(self): + return self.output_projection + + def set_output_embeddings(self, new_embeddings): + self.output_projection = new_embeddings + + @add_start_docstrings_to_model_forward(BIOGPT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @add_code_sample_docstrings( + processor_class=_TOKENIZER_FOR_DOC, + checkpoint=_CHECKPOINT_FOR_DOC, + output_type=CausalLMOutputWithCrossAttentions, + config_class=_CONFIG_FOR_DOC, + ) + def forward( + self, + input_ids: Optional[torch.LongTensor] = None, + attention_mask: Optional[torch.FloatTensor] = None, + head_mask: Optional[torch.FloatTensor] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple, CausalLMOutputWithCrossAttentions]: + r""" + labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Labels for language modeling. Note that the labels **are shifted** inside the model, i.e. you can set + `labels = input_ids` Indices are selected in `[-100, 0, ..., config.vocab_size]` All labels set to `-100` + are ignored (masked), the loss is only computed for labels in `[0, ..., config.vocab_size]` + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + outputs = self.biogpt( + input_ids, + attention_mask=attention_mask, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + past_key_values=past_key_values, + use_cache=use_cache, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + sequence_output = outputs[0] + prediction_scores = self.output_projection(sequence_output) + + lm_loss = None + if labels is not None: + # we are doing next-token prediction; shift prediction scores and input ids by one + shifted_prediction_scores = prediction_scores[:, :-1, :].contiguous() + labels = labels[:, 1:].contiguous() + loss_fct = CrossEntropyLoss() + lm_loss = loss_fct(shifted_prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) + + if not return_dict: + output = (prediction_scores,) + outputs[1:] + return ((lm_loss,) + output) if lm_loss is not None else output + + return CausalLMOutputWithCrossAttentions( + loss=lm_loss, + logits=prediction_scores, + past_key_values=outputs.past_key_values, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + cross_attentions=outputs.cross_attentions, + ) + + def prepare_inputs_for_generation(self, input_ids, attention_mask, past=None, **kwargs): + + # only last token for inputs_ids if past is defined in kwargs + if past: + input_ids = input_ids[:, -1].unsqueeze(-1) + + return { + "input_ids": input_ids, + "attention_mask": attention_mask, + "past_key_values": past, + "use_cache": kwargs.get("use_cache"), + } + + @staticmethod + def _reorder_cache(past, beam_idx): + reordered_past = () + for layer_past in past: + reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),) + return reordered_past diff --git a/src/transformers/models/biogpt/tokenization_biogpt.py b/src/transformers/models/biogpt/tokenization_biogpt.py new file mode 100644 index 0000000000..405e4c8625 --- /dev/null +++ b/src/transformers/models/biogpt/tokenization_biogpt.py @@ -0,0 +1,370 @@ +# coding=utf-8 +# Copyright 2022 The HuggingFace Team and Microsoft Research AI4Science. 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 BioGPT.""" +import json +import os +from typing import List, Optional, Tuple + +from ...tokenization_utils import PreTrainedTokenizer +from ...utils import logging + + +logger = logging.get_logger(__name__) + +VOCAB_FILES_NAMES = { + "vocab_file": "vocab.json", + "merges_file": "merges.txt", +} + +PRETRAINED_VOCAB_FILES_MAP = { + "vocab_file": { + "microsoft/biogpt": "https://huggingface.co/microsoft/biogpt/resolve/main/vocab.json", + }, + "merges_file": {"microsoft/biogpt": "https://huggingface.co/microsoft/biogpt/resolve/main/merges.txt"}, +} + +PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = { + "microsoft/biogpt": 1024, +} + + +def get_pairs(word): + """ + Return set of symbol pairs in a word. word is represented as tuple of symbols (symbols being variable-length + strings) + """ + pairs = set() + prev_char = word[0] + for char in word[1:]: + pairs.add((prev_char, char)) + prev_char = char + return pairs + + +class BioGptTokenizer(PreTrainedTokenizer): + """ + Construct an FAIRSEQ Transformer tokenizer. Moses tokenization followed by Byte-Pair Encoding. + + This tokenizer inherits from [`PreTrainedTokenizer`] which contains most of the main methods. Users should refer to + this superclass for more information regarding those methods. + + Args: + vocab_file (`str`): + Path to the vocabulary file. + merges_file (`str`): + Merges file. + unk_token (`str`, *optional*, defaults to `""`): + 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. + bos_token (`str`, *optional*, defaults to `""`): + The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token. + + + + 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 `cls_token`. + + + + eos_token (`str`, *optional*, defaults to `""`): + The end of sequence token. + + + + 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 `sep_token`. + + + + sep_token (`str`, *optional*, defaults to `""`): + 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. + pad_token (`str`, *optional*, defaults to `""`): + The token used for padding, for example when batching sequences of different lengths. + """ + + 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 = ["input_ids", "attention_mask"] + + def __init__( + self, + vocab_file, + merges_file, + unk_token="", + bos_token="", + eos_token="", + sep_token="", + pad_token="", + **kwargs + ): + super().__init__( + bos_token=bos_token, + eos_token=eos_token, + sep_token=sep_token, + unk_token=unk_token, + pad_token=pad_token, + **kwargs, + ) + + try: + import sacremoses + except ImportError: + raise ImportError( + "You need to install sacremoses to use BioGptTokenizer. " + "See https://pypi.org/project/sacremoses/ for installation." + ) + + self.lang = "en" + self.sm = sacremoses + # cache of sm.MosesTokenizer instance + self.cache_moses_tokenizer = dict() + self.cache_moses_detokenizer = dict() + + """ Initialisation""" + with open(vocab_file, encoding="utf-8") as vocab_handle: + self.encoder = json.load(vocab_handle) + self.decoder = {v: k for k, v in self.encoder.items()} + with open(merges_file, encoding="utf-8") as merges_handle: + merges = merges_handle.read().split("\n")[:-1] + merges = [tuple(merge.split()[:2]) for merge in merges] + self.bpe_ranks = dict(zip(merges, range(len(merges)))) + self.cache = {} + + @property + def vocab_size(self): + """Returns vocab size""" + return len(self.encoder) + + def get_vocab(self): + return dict(self.encoder, **self.added_tokens_encoder) + + def moses_tokenize(self, text, lang): + if lang not in self.cache_moses_tokenizer: + moses_tokenizer = self.sm.MosesTokenizer(lang=lang) + self.cache_moses_tokenizer[lang] = moses_tokenizer + return self.cache_moses_tokenizer[lang].tokenize( + text, aggressive_dash_splits=True, return_str=False, escape=True + ) + + def moses_detokenize(self, tokens, lang): + if lang not in self.cache_moses_detokenizer: + moses_detokenizer = self.sm.MosesDetokenizer(lang=lang) + self.cache_moses_detokenizer[lang] = moses_detokenizer + return self.cache_moses_detokenizer[lang].detokenize(tokens) + + def bpe(self, token): + word = tuple(token[:-1]) + (token[-1] + "",) + if token in self.cache: + return self.cache[token] + pairs = get_pairs(word) + + if not pairs: + return token + "" + + while True: + bigram = min(pairs, key=lambda pair: self.bpe_ranks.get(pair, float("inf"))) + if bigram not in self.bpe_ranks: + break + first, second = bigram + new_word = [] + i = 0 + while i < len(word): + try: + j = word.index(first, i) + except ValueError: + new_word.extend(word[i:]) + break + else: + new_word.extend(word[i:j]) + i = j + + if word[i] == first and i < len(word) - 1 and word[i + 1] == second: + new_word.append(first + second) + i += 2 + else: + new_word.append(word[i]) + i += 1 + new_word = tuple(new_word) + word = new_word + if len(word) == 1: + break + else: + pairs = get_pairs(word) + word = " ".join(word) + if word == "\n ": + word = "\n" + self.cache[token] = word + return word + + def _tokenize(self, text, bypass_tokenizer=False): + """Returns a tokenized string.""" + if bypass_tokenizer: + text = text.split() + else: + text = self.moses_tokenize(text, self.lang) + + split_tokens = [] + for token in text: + if token: + split_tokens.extend([t for t in self.bpe(token).split(" ")]) + + return split_tokens + + def _convert_token_to_id(self, token): + """Converts a token (str) in an id using the vocab.""" + return self.encoder.get(token, self.encoder.get(self.unk_token)) + + def _convert_id_to_token(self, index): + """Converts an index (integer) in a token (str) using the vocab.""" + return self.decoder.get(index, self.unk_token) + + def convert_tokens_to_string(self, tokens): + """Converts a sequence of tokens (string) in a single string.""" + # remove BPE + tokens = [t.replace(" ", "").replace("", " ") for t in tokens] + tokens = "".join(tokens).split() + # detokenize + text = self.moses_detokenize(tokens, self.lang) + return text + + 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 BioGPT sequence has the following format: + + - single sequence: ` X ` + - pair of sequences: ` A B ` + + Args: + token_ids_0 (`List[int]`): + List of IDs to which the special tokens will be added. + token_ids_1 (`List[int]`, *optional*): + Optional second list of IDs for sequence pairs. + + Returns: + `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens. + """ + if token_ids_1 is None: + return [self.sep_token_id] + token_ids_0 + sep = [self.sep_token_id] + return sep + token_ids_0 + sep + token_ids_1 + + 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 (`List[int]`): + List of IDs. + token_ids_1 (`List[int]`, *optional*): + Optional second list of IDs for sequence pairs. + already_has_special_tokens (`bool`, *optional*, defaults to `False`): + Whether or not the token list is already formatted with special tokens for the model. + + Returns: + `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: + return super().get_special_tokens_mask( + token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True + ) + # no bos used in fairseq + if token_ids_1 is not None: + return [1] + ([0] * len(token_ids_0)) + [1] + ([0] * len(token_ids_1)) + return [1] + ([0] * len(token_ids_0)) + + 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 FAIRSEQ + Transformer 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 `token_ids_1` is `None`, this method only returns the first portion of the mask (0s). + + Args: + token_ids_0 (`List[int]`): + List of IDs. + token_ids_1 (`List[int]`, *optional*): + Optional second list of IDs for sequence pairs. + + Returns: + `List[int]`: List of [token type IDs](../glossary#token-type-ids) according to the given sequence(s). + """ + sep = [self.sep_token_id] + + # no bos used in fairseq + 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]: + if not os.path.isdir(save_directory): + logger.error(f"Vocabulary path ({save_directory}) should be a directory") + return + vocab_file = os.path.join( + save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"] + ) + merge_file = os.path.join( + save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["merges_file"] + ) + + with open(vocab_file, "w", encoding="utf-8") as f: + f.write(json.dumps(self.encoder, indent=2, sort_keys=True, ensure_ascii=False) + "\n") + + index = 0 + with open(merge_file, "w", encoding="utf-8") as writer: + for bpe_tokens, token_index in sorted(self.bpe_ranks.items(), key=lambda kv: kv[1]): + if index != token_index: + logger.warning( + f"Saving vocabulary to {merge_file}: BPE merge indices are not consecutive." + " Please check that the tokenizer is not corrupted!" + ) + index = token_index + writer.write(" ".join(bpe_tokens) + "\n") + index += 1 + + return vocab_file, merge_file + + def __getstate__(self): + state = self.__dict__.copy() + state["sm"] = None + return state + + def __setstate__(self, d): + self.__dict__ = d + + try: + import sacremoses + except ImportError: + raise ImportError( + "You need to install sacremoses to use XLMTokenizer. " + "See https://pypi.org/project/sacremoses/ for installation." + ) + + self.sm = sacremoses diff --git a/src/transformers/utils/dummy_pt_objects.py b/src/transformers/utils/dummy_pt_objects.py index daaefd5297..8c33cfd3df 100644 --- a/src/transformers/utils/dummy_pt_objects.py +++ b/src/transformers/utils/dummy_pt_objects.py @@ -995,6 +995,30 @@ class BigBirdPegasusPreTrainedModel(metaclass=DummyObject): requires_backends(self, ["torch"]) +BIOGPT_PRETRAINED_MODEL_ARCHIVE_LIST = None + + +class BioGptForCausalLM(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class BioGptModel(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class BioGptPreTrainedModel(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST = None diff --git a/tests/models/biogpt/__init__.py b/tests/models/biogpt/__init__.py new file mode 100644 index 0000000000..e69de29bb2 diff --git a/tests/models/biogpt/test_modeling_biogpt.py b/tests/models/biogpt/test_modeling_biogpt.py new file mode 100644 index 0000000000..b0eb86d43f --- /dev/null +++ b/tests/models/biogpt/test_modeling_biogpt.py @@ -0,0 +1,398 @@ +# coding=utf-8 +# Copyright 2022 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. +""" Testing suite for the PyTorch BioGPT model. """ + +import math +import unittest + +from transformers import BioGptConfig, is_torch_available +from transformers.testing_utils import require_torch, slow, torch_device + +from ...generation.test_utils import GenerationTesterMixin +from ...test_configuration_common import ConfigTester +from ...test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask + + +if is_torch_available(): + import torch + + from transformers import BioGptForCausalLM, BioGptModel, BioGptTokenizer + from transformers.models.biogpt.modeling_biogpt import BIOGPT_PRETRAINED_MODEL_ARCHIVE_LIST + + +class BioGptModelTester: + def __init__( + self, + parent, + batch_size=13, + seq_length=7, + is_training=True, + use_input_mask=True, + use_token_type_ids=False, + use_labels=True, + vocab_size=99, + hidden_size=32, + num_hidden_layers=5, + num_attention_heads=4, + intermediate_size=37, + hidden_act="gelu", + hidden_dropout_prob=0.1, + attention_probs_dropout_prob=0.1, + max_position_embeddings=512, + type_vocab_size=16, + type_sequence_label_size=2, + initializer_range=0.02, + num_labels=3, + num_choices=4, + scope=None, + ): + self.parent = parent + self.batch_size = batch_size + self.seq_length = seq_length + self.is_training = is_training + self.use_input_mask = use_input_mask + self.use_token_type_ids = use_token_type_ids + self.use_labels = use_labels + self.vocab_size = vocab_size + self.hidden_size = hidden_size + self.num_hidden_layers = num_hidden_layers + self.num_attention_heads = num_attention_heads + self.intermediate_size = intermediate_size + self.hidden_act = hidden_act + self.hidden_dropout_prob = hidden_dropout_prob + self.attention_probs_dropout_prob = attention_probs_dropout_prob + self.max_position_embeddings = max_position_embeddings + self.type_vocab_size = type_vocab_size + self.type_sequence_label_size = type_sequence_label_size + self.initializer_range = initializer_range + self.num_labels = num_labels + self.num_choices = num_choices + self.scope = scope + + def prepare_config_and_inputs(self): + input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size) + + input_mask = None + if self.use_input_mask: + input_mask = random_attention_mask([self.batch_size, self.seq_length]) + + token_type_ids = None + if self.use_token_type_ids: + token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size) + + sequence_labels = None + token_labels = None + choice_labels = None + if self.use_labels: + sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size) + token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels) + choice_labels = ids_tensor([self.batch_size], self.num_choices) + + config = self.get_config() + + return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + + def get_config(self): + return BioGptConfig( + vocab_size=self.vocab_size, + hidden_size=self.hidden_size, + num_hidden_layers=self.num_hidden_layers, + num_attention_heads=self.num_attention_heads, + intermediate_size=self.intermediate_size, + hidden_act=self.hidden_act, + hidden_dropout_prob=self.hidden_dropout_prob, + attention_probs_dropout_prob=self.attention_probs_dropout_prob, + max_position_embeddings=self.max_position_embeddings, + type_vocab_size=self.type_vocab_size, + is_decoder=False, + initializer_range=self.initializer_range, + ) + + def create_and_check_model( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + model = BioGptModel(config=config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=input_mask) + result = model(input_ids) + self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size)) + + def create_and_check_for_causal_lm( + self, + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + encoder_hidden_states, + encoder_attention_mask, + ): + model = BioGptForCausalLM(config=config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels) + self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size)) + + def create_and_check_biogpt_model_attention_mask_past( + self, config, input_ids, input_mask, head_mask, token_type_ids, *args + ): + model = BioGptModel(config=config) + model.to(torch_device) + model.eval() + + # create attention mask + attn_mask = torch.ones(input_ids.shape, dtype=torch.long, device=torch_device) + half_seq_length = self.seq_length // 2 + attn_mask[:, half_seq_length:] = 0 + + # first forward pass + output, past = model(input_ids, attention_mask=attn_mask).to_tuple() + + # create hypothetical next token and extent to next_input_ids + next_tokens = ids_tensor((self.batch_size, 1), config.vocab_size) + + # change a random masked slice from input_ids + random_seq_idx_to_change = ids_tensor((1,), half_seq_length).item() + 1 + random_other_next_tokens = ids_tensor((self.batch_size, 1), config.vocab_size).squeeze(-1) + input_ids[:, -random_seq_idx_to_change] = random_other_next_tokens + + # append to next input_ids and attn_mask + next_input_ids = torch.cat([input_ids, next_tokens], dim=-1) + attn_mask = torch.cat( + [attn_mask, torch.ones((attn_mask.shape[0], 1), dtype=torch.long, device=torch_device)], + dim=1, + ) + + # get two different outputs + output_from_no_past = model(next_input_ids, attention_mask=attn_mask)["last_hidden_state"] + output_from_past = model(next_tokens, past_key_values=past, attention_mask=attn_mask)["last_hidden_state"] + + # select random slice + random_slice_idx = ids_tensor((1,), output_from_past.shape[-1]).item() + output_from_no_past_slice = output_from_no_past[:, -1, random_slice_idx].detach() + output_from_past_slice = output_from_past[:, 0, random_slice_idx].detach() + + # test that outputs are equal for slice + self.parent.assertTrue(torch.allclose(output_from_past_slice, output_from_no_past_slice, atol=1e-3)) + + def create_and_check_biogpt_model_past_large_inputs( + self, config, input_ids, input_mask, head_mask, token_type_ids, *args + ): + model = BioGptModel(config=config).to(torch_device).eval() + + attention_mask = torch.ones(input_ids.shape, dtype=torch.long, device=torch_device) + + # first forward pass + outputs = model(input_ids, attention_mask=attention_mask, use_cache=True) + + output, past_key_values = outputs.to_tuple() + + # create hypothetical multiple next token and extent to next_input_ids + next_tokens = ids_tensor((self.batch_size, 3), config.vocab_size) + next_attn_mask = ids_tensor((self.batch_size, 3), 2) + + # append to next input_ids and + next_input_ids = torch.cat([input_ids, next_tokens], dim=-1) + next_attention_mask = torch.cat([attention_mask, next_attn_mask], dim=-1) + + output_from_no_past = model(next_input_ids, attention_mask=next_attention_mask)["last_hidden_state"] + output_from_past = model(next_tokens, attention_mask=next_attention_mask, past_key_values=past_key_values)[ + "last_hidden_state" + ] + + # select random slice + random_slice_idx = ids_tensor((1,), output_from_past.shape[-1]).item() + output_from_no_past_slice = output_from_no_past[:, -3:, random_slice_idx].detach() + output_from_past_slice = output_from_past[:, :, random_slice_idx].detach() + + self.parent.assertTrue(output_from_past_slice.shape[1] == next_tokens.shape[1]) + + # test that outputs are equal for slice + self.parent.assertTrue(torch.allclose(output_from_past_slice, output_from_no_past_slice, atol=1e-3)) + + def create_and_check_forward_and_backwards( + self, config, input_ids, input_mask, head_mask, token_type_ids, *args, gradient_checkpointing=False + ): + model = BioGptForCausalLM(config) + model.to(torch_device) + if gradient_checkpointing: + model.gradient_checkpointing_enable() + + result = model(input_ids, labels=input_ids) + self.parent.assertEqual(result.loss.shape, ()) + self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size)) + result.loss.backward() + + def create_and_check_biogpt_weight_initialization(self, config, *args): + model = BioGptModel(config) + model_std = model.config.initializer_range / math.sqrt(2 * model.config.num_hidden_layers) + for key in model.state_dict().keys(): + if "c_proj" in key and "weight" in key: + self.parent.assertLessEqual(abs(torch.std(model.state_dict()[key]) - model_std), 0.001) + self.parent.assertLessEqual(abs(torch.mean(model.state_dict()[key]) - 0.0), 0.01) + + def prepare_config_and_inputs_for_common(self): + config_and_inputs = self.prepare_config_and_inputs() + ( + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + ) = config_and_inputs + inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask} + return config, inputs_dict + + +@require_torch +class BioGptModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase): + + all_model_classes = (BioGptModel, BioGptForCausalLM) if is_torch_available() else () + all_generative_model_classes = (BioGptForCausalLM,) if is_torch_available() else () + test_pruning = False + + def setUp(self): + self.model_tester = BioGptModelTester(self) + self.config_tester = ConfigTester(self, config_class=BioGptConfig, hidden_size=37) + + def test_config(self): + self.config_tester.run_common_tests() + + def test_model(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_model(*config_and_inputs) + + def test_model_various_embeddings(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + for type in ["absolute", "relative_key", "relative_key_query"]: + config_and_inputs[0].position_embedding_type = type + self.model_tester.create_and_check_model(*config_and_inputs) + + def test_biogpt_model_att_mask_past(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_biogpt_model_attention_mask_past(*config_and_inputs) + + def test_biogpt_gradient_checkpointing(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_forward_and_backwards(*config_and_inputs, gradient_checkpointing=True) + + def test_biogpt_model_past_with_large_inputs(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_biogpt_model_past_large_inputs(*config_and_inputs) + + def test_biogpt_weight_initialization(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_biogpt_weight_initialization(*config_and_inputs) + + @slow + def test_batch_generation(self): + model = BioGptForCausalLM.from_pretrained("microsoft/biogpt") + model.to(torch_device) + tokenizer = BioGptTokenizer.from_pretrained("microsoft/biogpt") + + tokenizer.padding_side = "left" + + # Define PAD Token = EOS Token = 50256 + tokenizer.pad_token = tokenizer.eos_token + model.config.pad_token_id = model.config.eos_token_id + + # use different length sentences to test batching + sentences = [ + "Hello, my dog is a little", + "Today, I", + ] + + inputs = tokenizer(sentences, return_tensors="pt", padding=True) + input_ids = inputs["input_ids"].to(torch_device) + + outputs = model.generate( + input_ids=input_ids, + attention_mask=inputs["attention_mask"].to(torch_device), + ) + + inputs_non_padded = tokenizer(sentences[0], return_tensors="pt").input_ids.to(torch_device) + output_non_padded = model.generate(input_ids=inputs_non_padded) + + num_paddings = inputs_non_padded.shape[-1] - inputs["attention_mask"][-1].long().sum().cpu().item() + inputs_padded = tokenizer(sentences[1], return_tensors="pt").input_ids.to(torch_device) + output_padded = model.generate(input_ids=inputs_padded, max_length=model.config.max_length - num_paddings) + + batch_out_sentence = tokenizer.batch_decode(outputs, skip_special_tokens=True) + non_padded_sentence = tokenizer.decode(output_non_padded[0], skip_special_tokens=True) + padded_sentence = tokenizer.decode(output_padded[0], skip_special_tokens=True) + + expected_output_sentence = [ + "Hello, my dog is a little bit bigger than a little bit.", + "Today, I have a good idea of how to use the information", + ] + self.assertListEqual(expected_output_sentence, batch_out_sentence) + self.assertListEqual(expected_output_sentence, [non_padded_sentence, padded_sentence]) + + @slow + def test_model_from_pretrained(self): + for model_name in BIOGPT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]: + model = BioGptModel.from_pretrained(model_name) + self.assertIsNotNone(model) + + +@require_torch +class BioGptModelIntegrationTest(unittest.TestCase): + @slow + def test_inference_lm_head_model(self): + model = BioGptForCausalLM.from_pretrained("microsoft/biogpt") + input_ids = torch.tensor([[2, 4805, 9, 656, 21]]) + output = model(input_ids)[0] + + vocab_size = 42384 + + expected_shape = torch.Size((1, 5, vocab_size)) + self.assertEqual(output.shape, expected_shape) + + expected_slice = torch.tensor( + [[[-9.5236, -9.8918, 10.4557], [-11.0469, -9.6423, 8.1022], [-8.8664, -7.8826, 5.5325]]] + ) + + self.assertTrue(torch.allclose(output[:, :3, :3], expected_slice, atol=1e-4)) + + @slow + def test_biogpt_generation(self): + tokenizer = BioGptTokenizer.from_pretrained("microsoft/biogpt") + model = BioGptForCausalLM.from_pretrained("microsoft/biogpt") + model.to(torch_device) + + torch.manual_seed(0) + tokenized = tokenizer("COVID-19 is", return_tensors="pt").to(torch_device) + output_ids = model.generate( + **tokenized, + min_length=100, + max_length=1024, + num_beams=5, + early_stopping=True, + ) + output_str = tokenizer.decode(output_ids[0], skip_special_tokens=True) + + EXPECTED_OUTPUT_STR = ( + "COVID-19 is a global pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the" + " causative agent of coronavirus disease 2019 (COVID-19), which has spread to more than 200 countries and" + " territories, including the United States (US), Canada, Australia, New Zealand, the United Kingdom (UK)," + " and the United States of America (USA), as of March 11, 2020, with more than 800,000 confirmed cases and" + " more than 800,000 deaths." + ) + self.assertEqual(output_str, EXPECTED_OUTPUT_STR) diff --git a/tests/models/biogpt/test_tokenization_biogpt.py b/tests/models/biogpt/test_tokenization_biogpt.py new file mode 100644 index 0000000000..8ec8a248bb --- /dev/null +++ b/tests/models/biogpt/test_tokenization_biogpt.py @@ -0,0 +1,97 @@ +# coding=utf-8 +# Copyright 2022 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. + + +import json +import os +import unittest + +from transformers.models.biogpt.tokenization_biogpt import VOCAB_FILES_NAMES, BioGptTokenizer +from transformers.testing_utils import slow + +from ...test_tokenization_common import TokenizerTesterMixin + + +class BioGptTokenizationTest(TokenizerTesterMixin, unittest.TestCase): + tokenizer_class = BioGptTokenizer + test_rust_tokenizer = False + + def setUp(self): + super().setUp() + + # Adapted from Sennrich et al. 2015 and https://github.com/rsennrich/subword-nmt + vocab = [ + "l", + "o", + "w", + "e", + "r", + "s", + "t", + "i", + "d", + "n", + "w", + "r", + "t", + "lo", + "low", + "er", + "low", + "lowest", + "newer", + "wider", + "", + ] + vocab_tokens = dict(zip(vocab, range(len(vocab)))) + merges = ["l o 123", "lo w 1456", "e r 1789", ""] + + self.vocab_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["vocab_file"]) + self.merges_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["merges_file"]) + with open(self.vocab_file, "w") as fp: + fp.write(json.dumps(vocab_tokens)) + with open(self.merges_file, "w") as fp: + fp.write("\n".join(merges)) + + def get_input_output_texts(self, tokenizer): + input_text = "lower newer" + output_text = "lower newer" + return input_text, output_text + + def test_full_tokenizer(self): + """Adapted from Sennrich et al. 2015 and https://github.com/rsennrich/subword-nmt""" + tokenizer = BioGptTokenizer(self.vocab_file, self.merges_file) + + text = "lower" + bpe_tokens = ["low", "er"] + tokens = tokenizer.tokenize(text) + self.assertListEqual(tokens, bpe_tokens) + + input_tokens = tokens + [""] + input_bpe_tokens = [14, 15, 20] + self.assertListEqual(tokenizer.convert_tokens_to_ids(input_tokens), input_bpe_tokens) + + @slow + def test_sequence_builders(self): + tokenizer = BioGptTokenizer.from_pretrained("microsoft/biogpt") + + 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) + + self.assertTrue(encoded_sentence == [2] + text) + self.assertTrue(encoded_pair == [2] + text + [2] + text_2)