From 7cf52a49dee661f6adb7847991c6a84925999f5d Mon Sep 17 00:00:00 2001 From: Sijun He Date: Fri, 24 Jun 2022 00:36:22 +0800 Subject: [PATCH] Nezha Pytorch implementation (#17776) * wip * rebase * all tests pass * rebase * ready for PR * address comments * fix styles * add require_torch to pipeline test * remove remote image to improve CI consistency * address comments; fix tf/flax tests * address comments; fix tf/flax tests * fix tests; add alias * repo consistency tests * Update src/transformers/pipelines/visual_question_answering.py Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com> * address comments * Update src/transformers/pipelines/visual_question_answering.py Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com> * merge * wip * wip * wip * most basic tests passes * all tests pass now * relative embedding * wip * running make fixup * remove bert changes * fix doc * fix doc * fix issues * fix doc * address comments * fix CI * remove redundant copied from * address comments * fix broken test Co-authored-by: Sijun He Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com> --- README.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/nezha.mdx | 76 + src/transformers/__init__.py | 28 + src/transformers/models/__init__.py | 1 + .../models/auto/configuration_auto.py | 3 + src/transformers/models/auto/modeling_auto.py | 9 + .../models/auto/tokenization_auto.py | 1 + src/transformers/models/nezha/__init__.py | 74 + .../models/nezha/configuration_nezha.py | 110 ++ .../models/nezha/modeling_nezha.py | 1723 +++++++++++++++++ src/transformers/utils/dummy_pt_objects.py | 66 + src/transformers/utils/fx.py | 1 + tests/models/nezha/__init__.py | 0 tests/models/nezha/test_modeling_nezha.py | 478 +++++ 19 files changed, 2578 insertions(+) create mode 100644 docs/source/en/model_doc/nezha.mdx create mode 100644 src/transformers/models/nezha/__init__.py create mode 100644 src/transformers/models/nezha/configuration_nezha.py create mode 100644 src/transformers/models/nezha/modeling_nezha.py create mode 100644 tests/models/nezha/__init__.py create mode 100644 tests/models/nezha/test_modeling_nezha.py diff --git a/README.md b/README.md index b03cc35753..b9934c7134 100644 --- a/README.md +++ b/README.md @@ -299,6 +299,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. 1. **[MPNet](https://huggingface.co/docs/transformers/model_doc/mpnet)** (from Microsoft Research) released with the paper [MPNet: Masked and Permuted Pre-training for Language Understanding](https://arxiv.org/abs/2004.09297) by Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu. 1. **[MT5](https://huggingface.co/docs/transformers/model_doc/mt5)** (from Google AI) released with the paper [mT5: A massively multilingual pre-trained text-to-text transformer](https://arxiv.org/abs/2010.11934) by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel. +1. **[Nezha](https://huggingface.co/docs/transformers/main/model_doc/nezha)** (from Huawei Noah’s Ark Lab) released with the paper [NEZHA: Neural Contextualized Representation for Chinese Language Understanding](https://arxiv.org/abs/1909.00204) by Junqiu Wei, Xiaozhe Ren, Xiaoguang Li, Wenyong Huang, Yi Liao, Yasheng Wang, Jiashu Lin, Xin Jiang, Xiao Chen and Qun Liu. 1. **[Nyströmformer](https://huggingface.co/docs/transformers/model_doc/nystromformer)** (from the University of Wisconsin - Madison) released with the paper [Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention](https://arxiv.org/abs/2102.03902) by Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, Vikas Singh. 1. **[OPT](https://huggingface.co/docs/transformers/master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al. 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu. diff --git a/README_ko.md b/README_ko.md index f977b1ecc3..68e6747b38 100644 --- a/README_ko.md +++ b/README_ko.md @@ -280,6 +280,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. 1. **[MPNet](https://huggingface.co/docs/transformers/model_doc/mpnet)** (from Microsoft Research) released with the paper [MPNet: Masked and Permuted Pre-training for Language Understanding](https://arxiv.org/abs/2004.09297) by Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu. 1. **[MT5](https://huggingface.co/docs/transformers/model_doc/mt5)** (from Google AI) released with the paper [mT5: A massively multilingual pre-trained text-to-text transformer](https://arxiv.org/abs/2010.11934) by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel. +1. **[Nezha](https://huggingface.co/docs/transformers/main/model_doc/nezha)** (from Huawei Noah’s Ark Lab) released with the paper [NEZHA: Neural Contextualized Representation for Chinese Language Understanding](https://arxiv.org/abs/1909.00204) by Junqiu Wei, Xiaozhe Ren, Xiaoguang Li, Wenyong Huang, Yi Liao, Yasheng Wang, Jiashu Lin, Xin Jiang, Xiao Chen and Qun Liu. 1. **[Nyströmformer](https://huggingface.co/docs/transformers/model_doc/nystromformer)** (from the University of Wisconsin - Madison) released with the paper [Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention](https://arxiv.org/abs/2102.03902) by Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, Vikas Singh. 1. **[OPT](https://huggingface.co/docs/transformers/master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al. 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu. diff --git a/README_zh-hans.md b/README_zh-hans.md index 4605918eb2..2878aaec87 100644 --- a/README_zh-hans.md +++ b/README_zh-hans.md @@ -304,6 +304,7 @@ conda install -c huggingface transformers 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (来自 CMU/Google Brain) 伴随论文 [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) 由 Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou 发布。 1. **[MPNet](https://huggingface.co/docs/transformers/model_doc/mpnet)** (来自 Microsoft Research) 伴随论文 [MPNet: Masked and Permuted Pre-training for Language Understanding](https://arxiv.org/abs/2004.09297) 由 Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu 发布。 1. **[MT5](https://huggingface.co/docs/transformers/model_doc/mt5)** (来自 Google AI) 伴随论文 [mT5: A massively multilingual pre-trained text-to-text transformer](https://arxiv.org/abs/2010.11934) 由 Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel 发布。 +1. **[Nezha](https://huggingface.co/docs/transformers/main/model_doc/nezha)** (来自华为诺亚方舟实验室) 伴随论文 [NEZHA: Neural Contextualized Representation for Chinese Language Understanding](https://arxiv.org/abs/1909.00204) 由 Junqiu Wei, Xiaozhe Ren, Xiaoguang Li, Wenyong Huang, Yi Liao, Yasheng Wang, Jiashu Lin, Xin Jiang, Xiao Chen and Qun Liu 发布。 1. **[Nyströmformer](https://huggingface.co/docs/transformers/model_doc/nystromformer)** (来自 the University of Wisconsin - Madison) 伴随论文 [Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention](https://arxiv.org/abs/2102.03902) 由 Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, Vikas Singh 发布。 1. **[OPT](https://huggingface.co/docs/transformers/master/model_doc/opt)** (来自 Meta AI) 伴随论文 [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) 由 Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al 发布。 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (来自 Google) 伴随论文 [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) 由 Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu 发布。 diff --git a/README_zh-hant.md b/README_zh-hant.md index b4db503bc8..6b4b8e2717 100644 --- a/README_zh-hant.md +++ b/README_zh-hant.md @@ -316,6 +316,7 @@ conda install -c huggingface transformers 1. **[MobileBERT](https://huggingface.co/docs/transformers/model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. 1. **[MPNet](https://huggingface.co/docs/transformers/model_doc/mpnet)** (from Microsoft Research) released with the paper [MPNet: Masked and Permuted Pre-training for Language Understanding](https://arxiv.org/abs/2004.09297) by Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu. 1. **[MT5](https://huggingface.co/docs/transformers/model_doc/mt5)** (from Google AI) released with the paper [mT5: A massively multilingual pre-trained text-to-text transformer](https://arxiv.org/abs/2010.11934) by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel. +1. **[Nezha](https://huggingface.co/docs/transformers/main/model_doc/nezha)** (from Huawei Noah’s Ark Lab) released with the paper [NEZHA: Neural Contextualized Representation for Chinese Language Understanding](https://arxiv.org/abs/1909.00204) by Junqiu Wei, Xiaozhe Ren, Xiaoguang Li, Wenyong Huang, Yi Liao, Yasheng Wang, Jiashu Lin, Xin Jiang, Xiao Chen and Qun Liu. 1. **[Nyströmformer](https://huggingface.co/docs/transformers/model_doc/nystromformer)** (from the University of Wisconsin - Madison) released with the paper [Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention](https://arxiv.org/abs/2102.03902) by Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, Vikas Singh. 1. **[OPT](https://huggingface.co/docs/transformers/master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al. 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu. diff --git a/docs/source/en/_toctree.yml b/docs/source/en/_toctree.yml index eaa972ffcc..428ed64d9d 100644 --- a/docs/source/en/_toctree.yml +++ b/docs/source/en/_toctree.yml @@ -306,6 +306,8 @@ title: MPNet - local: model_doc/mt5 title: MT5 + - local: model_doc/nezha + title: NEZHA - local: model_doc/nystromformer title: Nyströmformer - local: model_doc/openai-gpt diff --git a/docs/source/en/index.mdx b/docs/source/en/index.mdx index 1cfd105c26..3d837f5511 100644 --- a/docs/source/en/index.mdx +++ b/docs/source/en/index.mdx @@ -122,6 +122,7 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret 1. **[MobileBERT](model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou. 1. **[MPNet](model_doc/mpnet)** (from Microsoft Research) released with the paper [MPNet: Masked and Permuted Pre-training for Language Understanding](https://arxiv.org/abs/2004.09297) by Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu. 1. **[MT5](model_doc/mt5)** (from Google AI) released with the paper [mT5: A massively multilingual pre-trained text-to-text transformer](https://arxiv.org/abs/2010.11934) by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel. +1. **[Nezha](model_doc/nezha)** (from Huawei Noah’s Ark Lab) released with the paper [NEZHA: Neural Contextualized Representation for Chinese Language Understanding](https://arxiv.org/abs/1909.00204) by Junqiu Wei, Xiaozhe Ren, Xiaoguang Li, Wenyong Huang, Yi Liao, Yasheng Wang, Jiashu Lin, Xin Jiang, Xiao Chen and Qun Liu. 1. **[Nyströmformer](model_doc/nystromformer)** (from the University of Wisconsin - Madison) released with the paper [Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention](https://arxiv.org/abs/2102.03902) by Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, Vikas Singh. 1. **[OPT](master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al. 1. **[Pegasus](model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu. @@ -248,6 +249,7 @@ Flax), PyTorch, and/or TensorFlow. | MobileBERT | ✅ | ✅ | ✅ | ✅ | ❌ | | MPNet | ✅ | ✅ | ✅ | ✅ | ❌ | | MT5 | ✅ | ✅ | ✅ | ✅ | ✅ | +| Nezha | ❌ | ❌ | ✅ | ❌ | ❌ | | Nyströmformer | ❌ | ❌ | ✅ | ❌ | ❌ | | OpenAI GPT | ✅ | ✅ | ✅ | ✅ | ❌ | | OpenAI GPT-2 | ✅ | ✅ | ✅ | ✅ | ✅ | diff --git a/docs/source/en/model_doc/nezha.mdx b/docs/source/en/model_doc/nezha.mdx new file mode 100644 index 0000000000..8b613c38eb --- /dev/null +++ b/docs/source/en/model_doc/nezha.mdx @@ -0,0 +1,76 @@ + + +# Nezha + +## Overview + +The Nezha model was proposed in [NEZHA: Neural Contextualized Representation for Chinese Language Understanding](https://arxiv.org/abs/1909.00204) by Junqiu Wei et al. + +The abstract from the paper is the following: + +*The pre-trained language models have achieved great successes in various natural language understanding (NLU) tasks +due to its capacity to capture the deep contextualized information in text by pre-training on large-scale corpora. +In this technical report, we present our practice of pre-training language models named NEZHA (NEural contextualiZed +representation for CHinese lAnguage understanding) on Chinese corpora and finetuning for the Chinese NLU tasks. +The current version of NEZHA is based on BERT with a collection of proven improvements, which include Functional +Relative Positional Encoding as an effective positional encoding scheme, Whole Word Masking strategy, +Mixed Precision Training and the LAMB Optimizer in training the models. The experimental results show that NEZHA +achieves the state-of-the-art performances when finetuned on several representative Chinese tasks, including +named entity recognition (People's Daily NER), sentence matching (LCQMC), Chinese sentiment classification (ChnSenti) +and natural language inference (XNLI).* + +This model was contributed by [sijunhe](https://huggingface.co/sijunhe). The original code can be found [here](https://github.com/huawei-noah/Pretrained-Language-Model/tree/master/NEZHA-PyTorch). + +## NezhaConfig + +[[autodoc]] NezhaConfig + +## NezhaModel + +[[autodoc]] NezhaModel + - forward + +## NezhaForPreTraining + +[[autodoc]] NezhaForPreTraining + - forward + +## NezhaForMaskedLM + +[[autodoc]] NezhaForMaskedLM + - forward + +## NezhaForNextSentencePrediction + +[[autodoc]] NezhaForNextSentencePrediction + - forward + +## NezhaForSequenceClassification + +[[autodoc]] NezhaForSequenceClassification + - forward + +## NezhaForMultipleChoice + +[[autodoc]] NezhaForMultipleChoice + - forward + +## NezhaForTokenClassification + +[[autodoc]] NezhaForTokenClassification + - forward + +## NezhaForQuestionAnswering + +[[autodoc]] NezhaForQuestionAnswering + - forward \ No newline at end of file diff --git a/src/transformers/__init__.py b/src/transformers/__init__.py index a68420b127..19c051bca0 100755 --- a/src/transformers/__init__.py +++ b/src/transformers/__init__.py @@ -255,6 +255,7 @@ _import_structure = { "models.mobilebert": ["MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "MobileBertConfig", "MobileBertTokenizer"], "models.mpnet": ["MPNET_PRETRAINED_CONFIG_ARCHIVE_MAP", "MPNetConfig", "MPNetTokenizer"], "models.mt5": ["MT5Config"], + "models.nezha": ["NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP", "NezhaConfig"], "models.nystromformer": [ "NYSTROMFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP", "NystromformerConfig", @@ -1392,6 +1393,20 @@ else: ] ) _import_structure["models.mt5"].extend(["MT5EncoderModel", "MT5ForConditionalGeneration", "MT5Model"]) + _import_structure["models.nezha"].extend( + [ + "NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST", + "NezhaForMaskedLM", + "NezhaForPreTraining", + "NezhaForNextSentencePrediction", + "NezhaForMultipleChoice", + "NezhaForQuestionAnswering", + "NezhaForSequenceClassification", + "NezhaForTokenClassification", + "NezhaModel", + "NezhaPreTrainedModel", + ] + ) _import_structure["models.nystromformer"].extend( [ "NYSTROMFORMER_PRETRAINED_MODEL_ARCHIVE_LIST", @@ -2878,6 +2893,7 @@ if TYPE_CHECKING: from .models.mobilebert import MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, MobileBertConfig, MobileBertTokenizer from .models.mpnet import MPNET_PRETRAINED_CONFIG_ARCHIVE_MAP, MPNetConfig, MPNetTokenizer from .models.mt5 import MT5Config + from .models.nezha import NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP, NezhaConfig from .models.nystromformer import NYSTROMFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP, NystromformerConfig from .models.openai import OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP, OpenAIGPTConfig, OpenAIGPTTokenizer from .models.opt import OPTConfig @@ -3831,6 +3847,18 @@ if TYPE_CHECKING: MPNetPreTrainedModel, ) from .models.mt5 import MT5EncoderModel, MT5ForConditionalGeneration, MT5Model + from .models.nezha import ( + NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST, + NezhaForMaskedLM, + NezhaForMultipleChoice, + NezhaForNextSentencePrediction, + NezhaForPreTraining, + NezhaForQuestionAnswering, + NezhaForSequenceClassification, + NezhaForTokenClassification, + NezhaModel, + NezhaPreTrainedModel, + ) from .models.nystromformer import ( NYSTROMFORMER_PRETRAINED_MODEL_ARCHIVE_LIST, NystromformerForMaskedLM, diff --git a/src/transformers/models/__init__.py b/src/transformers/models/__init__.py index bb0d7bccb7..c3eba3fcd4 100644 --- a/src/transformers/models/__init__.py +++ b/src/transformers/models/__init__.py @@ -92,6 +92,7 @@ from . import ( mobilebert, mpnet, mt5, + nezha, nystromformer, openai, opt, diff --git a/src/transformers/models/auto/configuration_auto.py b/src/transformers/models/auto/configuration_auto.py index b0f288e365..4eb52991ca 100644 --- a/src/transformers/models/auto/configuration_auto.py +++ b/src/transformers/models/auto/configuration_auto.py @@ -90,6 +90,7 @@ CONFIG_MAPPING_NAMES = OrderedDict( ("mobilebert", "MobileBertConfig"), ("mpnet", "MPNetConfig"), ("mt5", "MT5Config"), + ("nezha", "NezhaConfig"), ("nystromformer", "NystromformerConfig"), ("openai-gpt", "OpenAIGPTConfig"), ("opt", "OPTConfig"), @@ -203,6 +204,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict( ("mctct", "MCTCT_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("megatron-bert", "MEGATRON_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("mpnet", "MPNET_PRETRAINED_CONFIG_ARCHIVE_MAP"), + ("nezha", "NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("nystromformer", "NYSTROMFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("openai-gpt", "OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP"), ("opt", "OPT_PRETRAINED_CONFIG_ARCHIVE_MAP"), @@ -327,6 +329,7 @@ MODEL_NAMES_MAPPING = OrderedDict( ("mobilebert", "MobileBERT"), ("mpnet", "MPNet"), ("mt5", "MT5"), + ("nezha", "Nezha"), ("nystromformer", "Nyströmformer"), ("openai-gpt", "OpenAI GPT"), ("opt", "OPT"), diff --git a/src/transformers/models/auto/modeling_auto.py b/src/transformers/models/auto/modeling_auto.py index 528f502285..6ffb348410 100644 --- a/src/transformers/models/auto/modeling_auto.py +++ b/src/transformers/models/auto/modeling_auto.py @@ -89,6 +89,7 @@ MODEL_MAPPING_NAMES = OrderedDict( ("mobilebert", "MobileBertModel"), ("mpnet", "MPNetModel"), ("mt5", "MT5Model"), + ("nezha", "NezhaModel"), ("nystromformer", "NystromformerModel"), ("openai-gpt", "OpenAIGPTModel"), ("opt", "OPTModel"), @@ -166,6 +167,7 @@ MODEL_FOR_PRETRAINING_MAPPING_NAMES = OrderedDict( ("megatron-bert", "MegatronBertForPreTraining"), ("mobilebert", "MobileBertForPreTraining"), ("mpnet", "MPNetForMaskedLM"), + ("nezha", "NezhaForPreTraining"), ("openai-gpt", "OpenAIGPTLMHeadModel"), ("retribert", "RetriBertModel"), ("roberta", "RobertaForMaskedLM"), @@ -224,6 +226,7 @@ MODEL_WITH_LM_HEAD_MAPPING_NAMES = OrderedDict( ("megatron-bert", "MegatronBertForCausalLM"), ("mobilebert", "MobileBertForMaskedLM"), ("mpnet", "MPNetForMaskedLM"), + ("nezha", "NezhaForMaskedLM"), ("nystromformer", "NystromformerForMaskedLM"), ("openai-gpt", "OpenAIGPTLMHeadModel"), ("plbart", "PLBartForConditionalGeneration"), @@ -390,6 +393,7 @@ MODEL_FOR_MASKED_LM_MAPPING_NAMES = OrderedDict( ("megatron-bert", "MegatronBertForMaskedLM"), ("mobilebert", "MobileBertForMaskedLM"), ("mpnet", "MPNetForMaskedLM"), + ("nezha", "NezhaForMaskedLM"), ("nystromformer", "NystromformerForMaskedLM"), ("perceiver", "PerceiverForMaskedLM"), ("qdqbert", "QDQBertForMaskedLM"), @@ -479,6 +483,7 @@ MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES = OrderedDict( ("megatron-bert", "MegatronBertForSequenceClassification"), ("mobilebert", "MobileBertForSequenceClassification"), ("mpnet", "MPNetForSequenceClassification"), + ("nezha", "NezhaForSequenceClassification"), ("nystromformer", "NystromformerForSequenceClassification"), ("openai-gpt", "OpenAIGPTForSequenceClassification"), ("perceiver", "PerceiverForSequenceClassification"), @@ -529,6 +534,7 @@ MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES = OrderedDict( ("megatron-bert", "MegatronBertForQuestionAnswering"), ("mobilebert", "MobileBertForQuestionAnswering"), ("mpnet", "MPNetForQuestionAnswering"), + ("nezha", "NezhaForQuestionAnswering"), ("nystromformer", "NystromformerForQuestionAnswering"), ("qdqbert", "QDQBertForQuestionAnswering"), ("reformer", "ReformerForQuestionAnswering"), @@ -585,6 +591,7 @@ MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES = OrderedDict( ("megatron-bert", "MegatronBertForTokenClassification"), ("mobilebert", "MobileBertForTokenClassification"), ("mpnet", "MPNetForTokenClassification"), + ("nezha", "NezhaForTokenClassification"), ("nystromformer", "NystromformerForTokenClassification"), ("qdqbert", "QDQBertForTokenClassification"), ("rembert", "RemBertForTokenClassification"), @@ -620,6 +627,7 @@ MODEL_FOR_MULTIPLE_CHOICE_MAPPING_NAMES = OrderedDict( ("megatron-bert", "MegatronBertForMultipleChoice"), ("mobilebert", "MobileBertForMultipleChoice"), ("mpnet", "MPNetForMultipleChoice"), + ("nezha", "NezhaForMultipleChoice"), ("nystromformer", "NystromformerForMultipleChoice"), ("qdqbert", "QDQBertForMultipleChoice"), ("rembert", "RemBertForMultipleChoice"), @@ -640,6 +648,7 @@ MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING_NAMES = OrderedDict( ("fnet", "FNetForNextSentencePrediction"), ("megatron-bert", "MegatronBertForNextSentencePrediction"), ("mobilebert", "MobileBertForNextSentencePrediction"), + ("nezha", "NezhaForNextSentencePrediction"), ("qdqbert", "QDQBertForNextSentencePrediction"), ] ) diff --git a/src/transformers/models/auto/tokenization_auto.py b/src/transformers/models/auto/tokenization_auto.py index 76a6dee790..56881bc331 100644 --- a/src/transformers/models/auto/tokenization_auto.py +++ b/src/transformers/models/auto/tokenization_auto.py @@ -173,6 +173,7 @@ else: "MT5TokenizerFast" if is_tokenizers_available() else None, ), ), + ("nezha", ("BertTokenizer", "BertTokenizerFast" if is_tokenizers_available() else None)), ( "nystromformer", ( diff --git a/src/transformers/models/nezha/__init__.py b/src/transformers/models/nezha/__init__.py new file mode 100644 index 0000000000..9811ee3252 --- /dev/null +++ b/src/transformers/models/nezha/__init__.py @@ -0,0 +1,74 @@ +# 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_nezha": ["NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP", "NezhaConfig"], +} + +try: + if not is_torch_available(): + raise OptionalDependencyNotAvailable() +except OptionalDependencyNotAvailable: + pass +else: + _import_structure["modeling_nezha"] = [ + "NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST", + "NezhaForNextSentencePrediction", + "NezhaForMaskedLM", + "NezhaForPreTraining", + "NezhaForMultipleChoice", + "NezhaForQuestionAnswering", + "NezhaForSequenceClassification", + "NezhaForTokenClassification", + "NezhaModel", + "NezhaPreTrainedModel", + ] + + +if TYPE_CHECKING: + from .configuration_nezha import NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP, NezhaConfig + + try: + if not is_torch_available(): + raise OptionalDependencyNotAvailable() + except OptionalDependencyNotAvailable: + pass + else: + from .modeling_nezha import ( + NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST, + NezhaForMaskedLM, + NezhaForMultipleChoice, + NezhaForNextSentencePrediction, + NezhaForPreTraining, + NezhaForQuestionAnswering, + NezhaForSequenceClassification, + NezhaForTokenClassification, + NezhaModel, + NezhaPreTrainedModel, + ) + + +else: + import sys + + sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__) diff --git a/src/transformers/models/nezha/configuration_nezha.py b/src/transformers/models/nezha/configuration_nezha.py new file mode 100644 index 0000000000..eb57016cd4 --- /dev/null +++ b/src/transformers/models/nezha/configuration_nezha.py @@ -0,0 +1,110 @@ +from transformers import PretrainedConfig + + +NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP = { + "sijunhe/nezha-cn-base": "https://huggingface.co/sijunhe/nezha-cn-base/resolve/main/config.json", +} + + +class NezhaConfig(PretrainedConfig): + r""" + This is the configuration class to store the configuration of an [`NezhaModel`]. It is used to instantiate an Nezha + 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 Nezha + [sijunhe/nezha-cn-base](https://huggingface.co/sijunhe/nezha-cn-base) 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 21128): + Vocabulary size of the NEZHA model. Defines the different tokens that can be represented by the + *inputs_ids* passed to the forward method of [`NezhaModel`]. + embedding_size (`int`, optional, defaults to 128): + Dimensionality of vocabulary embeddings. + hidden_size (`int`, optional, defaults to 768): + Dimensionality of the encoder layers and the pooler layer. + num_hidden_layers (`int`, optional, defaults to 12): + Number of hidden layers in the Transformer encoder. + num_attention_heads (`int`, optional, defaults to 12): + Number of attention heads for each attention layer in the Transformer encoder. + intermediate_size (`int`, optional, defaults to 3072): + The dimensionality 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. + hidden_dropout_prob (`float`, optional, defaults to 0.1): + The dropout probability 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 512): + The maximum sequence length that this model might ever be used with. Typically set this to something large + (e.g., 512 or 1024 or 2048). + type_vocab_size (`int`, optional, defaults to 2): + The vocabulary size of the *token_type_ids* passed into [`NezhaModel`]. + 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. + classifier_dropout (`float`, optional, defaults to 0.1): + The dropout ratio for attached classifiers. + + Example: + + ```python + >>> from transformers import NezhaConfig, NezhaModel + + >>> # Initializing an Nezha configuration + >>> configuration = NezhaConfig() + + >>> # Initializing a model from the Nezha-base style configuration model + >>> model = NezhaModel(configuration) + + >>> # Accessing the model configuration + >>> configuration = model.config + ```""" + + pretrained_config_archive_map = NEZHA_PRETRAINED_CONFIG_ARCHIVE_MAP + model_type = "nezha" + + def __init__( + self, + vocab_size=21128, + embedding_size=128, + hidden_size=768, + num_hidden_layers=12, + num_attention_heads=12, + intermediate_size=3072, + hidden_act="gelu", + hidden_dropout_prob=0.1, + attention_probs_dropout_prob=0.1, + max_position_embeddings=512, + max_relative_position=64, + type_vocab_size=2, + initializer_range=0.02, + layer_norm_eps=1e-12, + classifier_dropout=0.1, + pad_token_id=0, + bos_token_id=2, + eos_token_id=3, + use_cache=True, + **kwargs + ): + super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs) + + self.vocab_size = vocab_size + self.embedding_size = embedding_size + self.hidden_size = hidden_size + self.num_hidden_layers = num_hidden_layers + self.num_attention_heads = num_attention_heads + self.hidden_act = hidden_act + self.intermediate_size = intermediate_size + self.hidden_dropout_prob = hidden_dropout_prob + self.attention_probs_dropout_prob = attention_probs_dropout_prob + self.max_position_embeddings = max_position_embeddings + self.max_relative_position = max_relative_position + self.type_vocab_size = type_vocab_size + self.initializer_range = initializer_range + self.layer_norm_eps = layer_norm_eps + self.classifier_dropout = classifier_dropout + self.use_cache = use_cache diff --git a/src/transformers/models/nezha/modeling_nezha.py b/src/transformers/models/nezha/modeling_nezha.py new file mode 100644 index 0000000000..ab37c142bc --- /dev/null +++ b/src/transformers/models/nezha/modeling_nezha.py @@ -0,0 +1,1723 @@ +# 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. +"""PyTorch Nezha model.""" + + +import math +import os +import warnings +from dataclasses import dataclass +from typing import List, Optional, Tuple, Union + +import torch +import torch.utils.checkpoint +from packaging import version +from torch import nn +from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss + +from ...activations import ACT2FN +from ...modeling_outputs import ( + BaseModelOutputWithPastAndCrossAttentions, + BaseModelOutputWithPoolingAndCrossAttentions, + MaskedLMOutput, + MultipleChoiceModelOutput, + NextSentencePredictorOutput, + QuestionAnsweringModelOutput, + SequenceClassifierOutput, + TokenClassifierOutput, +) +from ...modeling_utils import PreTrainedModel +from ...pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer +from ...utils import ( + ModelOutput, + add_code_sample_docstrings, + add_start_docstrings, + add_start_docstrings_to_model_forward, + logging, + replace_return_docstrings, +) +from .configuration_nezha import NezhaConfig + + +logger = logging.get_logger(__name__) + +_CHECKPOINT_FOR_DOC = "sijunhe/nezha-cn-base" +_CONFIG_FOR_DOC = "NezhaConfig" +_TOKENIZER_FOR_DOC = "BertTokenizer" + +NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST = [ + "sijunhe/nezha-cn-base", + "sijunhe/nezha-cn-large", + "sijunhe/nezha-base-wwm", + "sijunhe/nezha-large-wwm", + # See all Nezha models at https://huggingface.co/models?filter=nezha +] + + +def load_tf_weights_in_nezha(model, config, tf_checkpoint_path): + """Load tf checkpoints in a pytorch model.""" + try: + import re + + import numpy as np + import tensorflow as tf + except ImportError: + logger.error( + "Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see " + "https://www.tensorflow.org/install/ for installation instructions." + ) + raise + tf_path = os.path.abspath(tf_checkpoint_path) + logger.info(f"Converting TensorFlow checkpoint from {tf_path}") + # Load weights from TF model + init_vars = tf.train.list_variables(tf_path) + names = [] + arrays = [] + for name, shape in init_vars: + logger.info(f"Loading TF weight {name} with shape {shape}") + array = tf.train.load_variable(tf_path, name) + names.append(name) + arrays.append(array) + + for name, array in zip(names, arrays): + name = name.split("/") + # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v + # which are not required for using pretrained model + if any( + n in ["adam_v", "adam_m", "AdamWeightDecayOptimizer", "AdamWeightDecayOptimizer_1", "global_step"] + for n in name + ): + logger.info(f"Skipping {'/'.join(name)}") + continue + pointer = model + for m_name in name: + if re.fullmatch(r"[A-Za-z]+_\d+", m_name): + scope_names = re.split(r"_(\d+)", m_name) + else: + scope_names = [m_name] + if scope_names[0] == "kernel" or scope_names[0] == "gamma": + pointer = getattr(pointer, "weight") + elif scope_names[0] == "output_bias" or scope_names[0] == "beta": + pointer = getattr(pointer, "bias") + elif scope_names[0] == "output_weights": + pointer = getattr(pointer, "weight") + elif scope_names[0] == "squad": + pointer = getattr(pointer, "classifier") + else: + try: + pointer = getattr(pointer, scope_names[0]) + except AttributeError: + logger.info(f"Skipping {'/'.join(name)}") + continue + if len(scope_names) >= 2: + num = int(scope_names[1]) + pointer = pointer[num] + if m_name[-11:] == "_embeddings": + pointer = getattr(pointer, "weight") + elif m_name == "kernel": + array = np.transpose(array) + try: + if pointer.shape != array.shape: + raise ValueError(f"Pointer shape {pointer.shape} and array shape {array.shape} mismatched") + except AssertionError as e: + e.args += (pointer.shape, array.shape) + raise + logger.info(f"Initialize PyTorch weight {name}") + pointer.data = torch.from_numpy(array) + return model + + +class NezhaRelativePositionsEncoding(nn.Module): + """Implement the Functional Relative Position Encoding""" + + def __init__(self, length, depth, max_relative_position=127): + super().__init__() + vocab_size = max_relative_position * 2 + 1 + range_vec = torch.arange(length) + range_mat = range_vec.repeat(length).view(length, length) + distance_mat = range_mat - torch.t(range_mat) + distance_mat_clipped = torch.clamp(distance_mat, -max_relative_position, max_relative_position) + final_mat = distance_mat_clipped + max_relative_position + + embeddings_table = torch.zeros(vocab_size, depth) + position = torch.arange(0, vocab_size, dtype=torch.float).unsqueeze(1) + div_term = torch.exp(torch.arange(0, depth, 2).float() * (-math.log(10000.0) / depth)) + embeddings_table[:, 0::2] = torch.sin(position * div_term) + embeddings_table[:, 1::2] = torch.cos(position * div_term) + + flat_relative_positions_matrix = final_mat.view(-1) + one_hot_relative_positions_matrix = torch.nn.functional.one_hot( + flat_relative_positions_matrix, num_classes=vocab_size + ).float() + positions_encoding = torch.matmul(one_hot_relative_positions_matrix, embeddings_table) + my_shape = list(final_mat.size()) + my_shape.append(depth) + positions_encoding = positions_encoding.view(my_shape) + self.register_buffer("positions_encoding", positions_encoding) + + def forward(self, length): + return self.positions_encoding[:length, :length, :] + + +class NezhaEmbeddings(nn.Module): + """Construct the embeddings from word and token_type embeddings.""" + + def __init__(self, config): + super().__init__() + self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id) + self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size) + + # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load + # any TensorFlow checkpoint file + self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) + self.dropout = nn.Dropout(config.hidden_dropout_prob) + if version.parse(torch.__version__) > version.parse("1.6.0"): + self.register_buffer( + "token_type_ids", + torch.zeros((1, config.max_position_embeddings), dtype=torch.long), + persistent=False, + ) + + def forward( + self, + input_ids: Optional[torch.LongTensor] = None, + token_type_ids: Optional[torch.LongTensor] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + ) -> torch.Tensor: + if input_ids is not None: + input_shape = input_ids.size() + else: + input_shape = inputs_embeds.size()[:-1] + + seq_length = input_shape[1] + + if inputs_embeds is None: + inputs_embeds = self.word_embeddings(input_ids) + + # Setting the token_type_ids to the registered buffer in constructor where it is all zeros, which usually occurs + # when its auto-generated, registered buffer helps users when tracing the model without passing token_type_ids, solves + # issue #5664 + if token_type_ids is None: + if hasattr(self, "token_type_ids"): + buffered_token_type_ids = self.token_type_ids[:, :seq_length] + buffered_token_type_ids_expanded = buffered_token_type_ids.expand(input_shape[0], seq_length) + token_type_ids = buffered_token_type_ids_expanded + else: + token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=inputs_embeds.device) + + token_type_embeddings = self.token_type_embeddings(token_type_ids) + + embeddings = inputs_embeds + token_type_embeddings + embeddings = self.LayerNorm(embeddings) + embeddings = self.dropout(embeddings) + return embeddings + + +class NezhaSelfAttention(nn.Module): + def __init__(self, config): + super().__init__() + if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"): + raise ValueError( + f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention " + f"heads ({config.num_attention_heads})" + ) + + self.num_attention_heads = config.num_attention_heads + self.attention_head_size = int(config.hidden_size / config.num_attention_heads) + self.all_head_size = self.num_attention_heads * self.attention_head_size + + self.query = nn.Linear(config.hidden_size, self.all_head_size) + self.key = nn.Linear(config.hidden_size, self.all_head_size) + self.value = nn.Linear(config.hidden_size, self.all_head_size) + + self.dropout = nn.Dropout(config.attention_probs_dropout_prob) + self.relative_positions_encoding = NezhaRelativePositionsEncoding( + length=config.max_position_embeddings, + depth=self.attention_head_size, + max_relative_position=config.max_relative_position, + ) + self.is_decoder = config.is_decoder + + def transpose_for_scores(self, x: torch.Tensor) -> torch.Tensor: + new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size) + x = x.view(new_x_shape) + return x.permute(0, 2, 1, 3) + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.FloatTensor] = None, + head_mask: Optional[torch.FloatTensor] = None, + encoder_hidden_states: Optional[torch.FloatTensor] = None, + encoder_attention_mask: Optional[torch.FloatTensor] = None, + past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None, + output_attentions: Optional[bool] = False, + ) -> Tuple[torch.Tensor]: + mixed_query_layer = self.query(hidden_states) + + # If this is instantiated as a cross-attention module, the keys + # and values come from an encoder; the attention mask needs to be + # such that the encoder's padding tokens are not attended to. + is_cross_attention = encoder_hidden_states is not None + + if is_cross_attention and past_key_value is not None: + # reuse k,v, cross_attentions + key_layer = past_key_value[0] + value_layer = past_key_value[1] + attention_mask = encoder_attention_mask + elif is_cross_attention: + key_layer = self.transpose_for_scores(self.key(encoder_hidden_states)) + value_layer = self.transpose_for_scores(self.value(encoder_hidden_states)) + attention_mask = encoder_attention_mask + elif past_key_value is not None: + key_layer = self.transpose_for_scores(self.key(hidden_states)) + value_layer = self.transpose_for_scores(self.value(hidden_states)) + key_layer = torch.cat([past_key_value[0], key_layer], dim=2) + value_layer = torch.cat([past_key_value[1], value_layer], dim=2) + else: + key_layer = self.transpose_for_scores(self.key(hidden_states)) + value_layer = self.transpose_for_scores(self.value(hidden_states)) + + query_layer = self.transpose_for_scores(mixed_query_layer) + + 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_layer, value_layer) + + # Take the dot product between "query" and "key" to get the raw attention scores. + attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2)) + + batch_size, num_attention_heads, from_seq_length, to_seq_length = attention_scores.size() + relations_keys = self.relative_positions_encoding(to_seq_length) + query_layer_t = query_layer.permute(2, 0, 1, 3) + + query_layer_r = query_layer_t.contiguous().view( + from_seq_length, batch_size * num_attention_heads, self.attention_head_size + ) + key_position_scores = torch.matmul(query_layer_r, relations_keys.permute(0, 2, 1)) + key_position_scores_r = key_position_scores.view( + from_seq_length, batch_size, num_attention_heads, from_seq_length + ) + key_position_scores_r_t = key_position_scores_r.permute(1, 2, 0, 3) + attention_scores = attention_scores + key_position_scores_r_t + + attention_scores = attention_scores / math.sqrt(self.attention_head_size) + + if attention_mask is not None: + # Apply the attention mask is (precomputed for all layers in NezhaModel forward() function) + attention_scores = attention_scores + attention_mask + + # Normalize the attention scores to probabilities. + attention_probs = nn.functional.softmax(attention_scores, dim=-1) + + # This is actually dropping out entire tokens to attend to, which might + # seem a bit unusual, but is taken from the original Transformer paper. + attention_probs = self.dropout(attention_probs) + + # Mask heads if we want to + if head_mask is not None: + attention_probs = attention_probs * head_mask + + context_layer = torch.matmul(attention_probs, value_layer) + relations_values = self.relative_positions_encoding(to_seq_length) + attention_probs_t = attention_probs.permute(2, 0, 1, 3) + attentions_probs_r = attention_probs_t.contiguous().view( + from_seq_length, batch_size * num_attention_heads, to_seq_length + ) + value_position_scores = torch.matmul(attentions_probs_r, relations_values) + value_position_scores_r = value_position_scores.view( + from_seq_length, batch_size, num_attention_heads, self.attention_head_size + ) + value_position_scores_r_t = value_position_scores_r.permute(1, 2, 0, 3) + context_layer = context_layer + value_position_scores_r_t + + context_layer = context_layer.permute(0, 2, 1, 3).contiguous() + new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,) + context_layer = context_layer.view(new_context_layer_shape) + + outputs = (context_layer, attention_probs) if output_attentions else (context_layer,) + + if self.is_decoder: + outputs = outputs + (past_key_value,) + return outputs + + +# Copied from transformers.models.bert.modeling_bert.BertSelfOutput with Bert->Nezha +class NezhaSelfOutput(nn.Module): + def __init__(self, config): + super().__init__() + self.dense = nn.Linear(config.hidden_size, config.hidden_size) + self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) + self.dropout = nn.Dropout(config.hidden_dropout_prob) + + def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor: + hidden_states = self.dense(hidden_states) + hidden_states = self.dropout(hidden_states) + hidden_states = self.LayerNorm(hidden_states + input_tensor) + return hidden_states + + +class NezhaAttention(nn.Module): + def __init__(self, config): + super().__init__() + self.self = NezhaSelfAttention(config) + self.output = NezhaSelfOutput(config) + self.pruned_heads = set() + + def prune_heads(self, heads): + if len(heads) == 0: + return + heads, index = find_pruneable_heads_and_indices( + heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads + ) + + # Prune linear layers + self.self.query = prune_linear_layer(self.self.query, index) + self.self.key = prune_linear_layer(self.self.key, index) + self.self.value = prune_linear_layer(self.self.value, index) + self.output.dense = prune_linear_layer(self.output.dense, index, dim=1) + + # Update hyper params and store pruned heads + self.self.num_attention_heads = self.self.num_attention_heads - len(heads) + self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads + self.pruned_heads = self.pruned_heads.union(heads) + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.FloatTensor] = None, + head_mask: Optional[torch.FloatTensor] = None, + encoder_hidden_states: Optional[torch.FloatTensor] = None, + encoder_attention_mask: Optional[torch.FloatTensor] = None, + past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None, + output_attentions: Optional[bool] = False, + ) -> Tuple[torch.Tensor]: + self_outputs = self.self( + hidden_states, + attention_mask, + head_mask, + encoder_hidden_states, + encoder_attention_mask, + past_key_value, + output_attentions, + ) + attention_output = self.output(self_outputs[0], hidden_states) + outputs = (attention_output,) + self_outputs[1:] # add attentions if we output them + return outputs + + +# Copied from transformers.models.bert.modeling_bert.BertIntermediate with Bert->Nezha +class NezhaIntermediate(nn.Module): + def __init__(self, config): + super().__init__() + self.dense = nn.Linear(config.hidden_size, config.intermediate_size) + if isinstance(config.hidden_act, str): + self.intermediate_act_fn = ACT2FN[config.hidden_act] + else: + self.intermediate_act_fn = config.hidden_act + + def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: + hidden_states = self.dense(hidden_states) + hidden_states = self.intermediate_act_fn(hidden_states) + return hidden_states + + +# Copied from transformers.models.bert.modeling_bert.BertOutput with Bert->Nezha +class NezhaOutput(nn.Module): + def __init__(self, config): + super().__init__() + self.dense = nn.Linear(config.intermediate_size, config.hidden_size) + self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) + self.dropout = nn.Dropout(config.hidden_dropout_prob) + + def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor: + hidden_states = self.dense(hidden_states) + hidden_states = self.dropout(hidden_states) + hidden_states = self.LayerNorm(hidden_states + input_tensor) + return hidden_states + + +class NezhaLayer(nn.Module): + def __init__(self, config): + super().__init__() + self.chunk_size_feed_forward = config.chunk_size_feed_forward + self.seq_len_dim = 1 + self.attention = NezhaAttention(config) + self.is_decoder = config.is_decoder + self.add_cross_attention = config.add_cross_attention + if self.add_cross_attention: + if not self.is_decoder: + raise ValueError(f"{self} should be used as a decoder model if cross attention is added") + self.crossattention = NezhaAttention(config) + self.intermediate = NezhaIntermediate(config) + self.output = NezhaOutput(config) + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.FloatTensor] = None, + head_mask: Optional[torch.FloatTensor] = None, + encoder_hidden_states: Optional[torch.FloatTensor] = None, + encoder_attention_mask: Optional[torch.FloatTensor] = None, + past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None, + output_attentions: Optional[bool] = False, + ) -> Tuple[torch.Tensor]: + # 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 + self_attention_outputs = self.attention( + hidden_states, + attention_mask, + head_mask, + output_attentions=output_attentions, + past_key_value=self_attn_past_key_value, + ) + attention_output = self_attention_outputs[0] + + # if decoder, the last output is tuple of self-attn cache + if self.is_decoder: + outputs = self_attention_outputs[1:-1] + present_key_value = self_attention_outputs[-1] + else: + outputs = self_attention_outputs[1:] # add self attentions if we output attention weights + + cross_attn_present_key_value = None + if self.is_decoder and encoder_hidden_states is not None: + if not hasattr(self, "crossattention"): + raise ValueError( + f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers" + " by setting `config.add_cross_attention=True`" + ) + + # cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple + cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None + cross_attention_outputs = self.crossattention( + attention_output, + attention_mask, + head_mask, + encoder_hidden_states, + encoder_attention_mask, + cross_attn_past_key_value, + output_attentions, + ) + attention_output = cross_attention_outputs[0] + outputs = outputs + cross_attention_outputs[1:-1] # add cross attentions if we output attention weights + + # add cross-attn cache to positions 3,4 of present_key_value tuple + cross_attn_present_key_value = cross_attention_outputs[-1] + present_key_value = present_key_value + cross_attn_present_key_value + + layer_output = apply_chunking_to_forward( + self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output + ) + outputs = (layer_output,) + outputs + + # if decoder, return the attn key/values as the last output + if self.is_decoder: + outputs = outputs + (present_key_value,) + + return outputs + + def feed_forward_chunk(self, attention_output): + intermediate_output = self.intermediate(attention_output) + layer_output = self.output(intermediate_output, attention_output) + return layer_output + + +# Copied from transformers.models.bert.modeling_bert.BertEncoder with Bert->Nezha +class NezhaEncoder(nn.Module): + def __init__(self, config): + super().__init__() + self.config = config + self.layer = nn.ModuleList([NezhaLayer(config) for _ in range(config.num_hidden_layers)]) + self.gradient_checkpointing = False + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.FloatTensor] = None, + head_mask: Optional[torch.FloatTensor] = None, + encoder_hidden_states: Optional[torch.FloatTensor] = None, + encoder_attention_mask: Optional[torch.FloatTensor] = None, + past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = False, + output_hidden_states: Optional[bool] = False, + return_dict: Optional[bool] = True, + ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]: + all_hidden_states = () if output_hidden_states else None + all_self_attentions = () if output_attentions else None + all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None + + next_decoder_cache = () if use_cache else None + for i, layer_module in enumerate(self.layer): + if output_hidden_states: + all_hidden_states = all_hidden_states + (hidden_states,) + + layer_head_mask = head_mask[i] if head_mask is not None else None + past_key_value = past_key_values[i] 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): + return module(*inputs, past_key_value, output_attentions) + + return custom_forward + + layer_outputs = torch.utils.checkpoint.checkpoint( + create_custom_forward(layer_module), + hidden_states, + attention_mask, + layer_head_mask, + encoder_hidden_states, + encoder_attention_mask, + ) + else: + layer_outputs = layer_module( + hidden_states, + attention_mask, + layer_head_mask, + encoder_hidden_states, + encoder_attention_mask, + past_key_value, + output_attentions, + ) + + hidden_states = layer_outputs[0] + if use_cache: + next_decoder_cache += (layer_outputs[-1],) + if output_attentions: + all_self_attentions = all_self_attentions + (layer_outputs[1],) + if self.config.add_cross_attention: + all_cross_attentions = all_cross_attentions + (layer_outputs[2],) + + if output_hidden_states: + all_hidden_states = all_hidden_states + (hidden_states,) + + if not return_dict: + return tuple( + v + for v in [ + hidden_states, + next_decoder_cache, + all_hidden_states, + all_self_attentions, + all_cross_attentions, + ] + if v is not None + ) + return BaseModelOutputWithPastAndCrossAttentions( + last_hidden_state=hidden_states, + past_key_values=next_decoder_cache, + hidden_states=all_hidden_states, + attentions=all_self_attentions, + cross_attentions=all_cross_attentions, + ) + + +# Copied from transformers.models.bert.modeling_bert.BertPooler with Bert->Nezha +class NezhaPooler(nn.Module): + def __init__(self, config): + super().__init__() + self.dense = nn.Linear(config.hidden_size, config.hidden_size) + self.activation = nn.Tanh() + + def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: + # We "pool" the model by simply taking the hidden state corresponding + # to the first token. + first_token_tensor = hidden_states[:, 0] + pooled_output = self.dense(first_token_tensor) + pooled_output = self.activation(pooled_output) + return pooled_output + + +# Copied from transformers.models.bert.modeling_bert.BertPredictionHeadTransform with Bert->Nezha +class NezhaPredictionHeadTransform(nn.Module): + def __init__(self, config): + super().__init__() + self.dense = nn.Linear(config.hidden_size, config.hidden_size) + if isinstance(config.hidden_act, str): + self.transform_act_fn = ACT2FN[config.hidden_act] + else: + self.transform_act_fn = config.hidden_act + self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) + + def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: + hidden_states = self.dense(hidden_states) + hidden_states = self.transform_act_fn(hidden_states) + hidden_states = self.LayerNorm(hidden_states) + return hidden_states + + +# Copied from transformers.models.bert.modeling_bert.BertLMPredictionHead with Bert->Nezha +class NezhaLMPredictionHead(nn.Module): + def __init__(self, config): + super().__init__() + self.transform = NezhaPredictionHeadTransform(config) + + # The output weights are the same as the input embeddings, but there is + # an output-only bias for each token. + self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False) + + self.bias = nn.Parameter(torch.zeros(config.vocab_size)) + + # Need a link between the two variables so that the bias is correctly resized with `resize_token_embeddings` + self.decoder.bias = self.bias + + def forward(self, hidden_states): + hidden_states = self.transform(hidden_states) + hidden_states = self.decoder(hidden_states) + return hidden_states + + +# Copied from transformers.models.bert.modeling_bert.BertOnlyMLMHead with Bert->Nezha +class NezhaOnlyMLMHead(nn.Module): + def __init__(self, config): + super().__init__() + self.predictions = NezhaLMPredictionHead(config) + + def forward(self, sequence_output: torch.Tensor) -> torch.Tensor: + prediction_scores = self.predictions(sequence_output) + return prediction_scores + + +# Copied from transformers.models.bert.modeling_bert.BertOnlyNSPHead with Bert->Nezha +class NezhaOnlyNSPHead(nn.Module): + def __init__(self, config): + super().__init__() + self.seq_relationship = nn.Linear(config.hidden_size, 2) + + def forward(self, pooled_output): + seq_relationship_score = self.seq_relationship(pooled_output) + return seq_relationship_score + + +# Copied from transformers.models.bert.modeling_bert.BertPreTrainingHeads with Bert->Nezha +class NezhaPreTrainingHeads(nn.Module): + def __init__(self, config): + super().__init__() + self.predictions = NezhaLMPredictionHead(config) + self.seq_relationship = nn.Linear(config.hidden_size, 2) + + def forward(self, sequence_output, pooled_output): + prediction_scores = self.predictions(sequence_output) + seq_relationship_score = self.seq_relationship(pooled_output) + return prediction_scores, seq_relationship_score + + +class NezhaPreTrainedModel(PreTrainedModel): + """ + An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained + models. + """ + + config_class = NezhaConfig + load_tf_weights = load_tf_weights_in_nezha + base_model_prefix = "nezha" + supports_gradient_checkpointing = True + _keys_to_ignore_on_load_missing = [r"positions_encoding"] + + 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, NezhaEncoder): + module.gradient_checkpointing = value + + +@dataclass +class NezhaForPreTrainingOutput(ModelOutput): + """ + Output type of [`NezhaForPreTraining`]. + + Args: + loss (*optional*, returned when `labels` is provided, `torch.FloatTensor` of shape `(1,)`): + Total loss as the sum of the masked language modeling loss and the next sequence prediction + (classification) loss. + prediction_logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`): + Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax). + seq_relationship_logits (`torch.FloatTensor` of shape `(batch_size, 2)`): + Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation + before SoftMax). + hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`): + Tuple of `torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer) of + shape `(batch_size, sequence_length, hidden_size)`. + + Hidden-states of the model at the output of each layer plus the initial embedding outputs. + attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`): + Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length, + sequence_length)`. + + Attentions weights after the attention softmax, used to compute the weighted average in the self-attention + heads. + """ + + loss: Optional[torch.FloatTensor] = None + prediction_logits: torch.FloatTensor = None + seq_relationship_logits: torch.FloatTensor = None + hidden_states: Optional[Tuple[torch.FloatTensor]] = None + attentions: Optional[Tuple[torch.FloatTensor]] = None + + +NEZHA_START_DOCSTRING = r""" + + This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the + library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads + etc.) + + This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass. + Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage + and behavior. + + Parameters: + config ([`NezhaConfig`]): 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. +""" + +NEZHA_INPUTS_DOCSTRING = r""" + Args: + input_ids (`torch.LongTensor` of shape `({0})`): + Indices of input sequence tokens in the vocabulary. + + Indices can be obtained using [`BertTokenizer`]. 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) + token_type_ids (`torch.LongTensor` of shape `({0})`, *optional*): + Segment token indices to indicate first and second portions of the inputs. Indices are selected in `[0, + 1]`: + + - 0 corresponds to a *sentence A* token, + - 1 corresponds to a *sentence B* token. + + [What are token type IDs?](../glossary#token-type-ids) + 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. + 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 Nezha Model transformer outputting raw hidden-states without any specific head on top.", + NEZHA_START_DOCSTRING, +) +class NezhaModel(NezhaPreTrainedModel): + """ + + The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of + cross-attention is added between the self-attention layers, following the architecture described in [Attention is + all you need](https://arxiv.org/abs/1706.03762) by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, + Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin. + + To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set + to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and + `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass. + """ + + def __init__(self, config, add_pooling_layer=True): + super().__init__(config) + self.config = config + + self.embeddings = NezhaEmbeddings(config) + self.encoder = NezhaEncoder(config) + + self.pooler = NezhaPooler(config) if add_pooling_layer else None + + # Initialize weights and apply final processing + self.post_init() + + def get_input_embeddings(self): + return self.embeddings.word_embeddings + + def set_input_embeddings(self, value): + self.embeddings.word_embeddings = value + + def _prune_heads(self, heads_to_prune): + """ + Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base + class PreTrainedModel + """ + for layer, heads in heads_to_prune.items(): + self.encoder.layer[layer].attention.prune_heads(heads) + + @add_start_docstrings_to_model_forward(NEZHA_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @add_code_sample_docstrings( + processor_class=_TOKENIZER_FOR_DOC, + checkpoint=_CHECKPOINT_FOR_DOC, + output_type=BaseModelOutputWithPoolingAndCrossAttentions, + config_class=_CONFIG_FOR_DOC, + ) + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + encoder_hidden_states: Optional[torch.Tensor] = None, + encoder_attention_mask: Optional[torch.Tensor] = None, + past_key_values: Optional[List[torch.FloatTensor]] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]: + r""" + encoder_hidden_states (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*): + Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if + the model is configured as a decoder. + encoder_attention_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): + Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in + the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`: + + - 1 for tokens that are **not masked**, + - 0 for tokens that are **masked**. + past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`): + Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding. + + If `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)`. + 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 = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + if self.config.is_decoder: + use_cache = use_cache if use_cache is not None else self.config.use_cache + else: + use_cache = False + + 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_shape = input_ids.size() + elif inputs_embeds is not None: + input_shape = inputs_embeds.size()[:-1] + else: + raise ValueError("You have to specify either input_ids or inputs_embeds") + + batch_size, seq_length = input_shape + device = input_ids.device if input_ids is not None else inputs_embeds.device + + # 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 attention_mask is None: + attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device) + + if token_type_ids is None: + if hasattr(self.embeddings, "token_type_ids"): + buffered_token_type_ids = self.embeddings.token_type_ids[:, :seq_length] + buffered_token_type_ids_expanded = buffered_token_type_ids.expand(batch_size, seq_length) + token_type_ids = buffered_token_type_ids_expanded + else: + token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device) + + # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length] + # ourselves in which case we just need to make it broadcastable to all heads. + extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape) + + # If a 2D or 3D attention mask is provided for the cross-attention + # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length] + if self.config.is_decoder and encoder_hidden_states is not None: + encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size() + encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length) + if encoder_attention_mask is None: + encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device) + encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask) + else: + encoder_extended_attention_mask = None + + # Prepare head mask if needed + # 1.0 in head_mask indicate we keep the head + # attention_probs has shape bsz x n_heads x N x N + # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads] + # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length] + head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers) + + embedding_output = self.embeddings( + input_ids=input_ids, + token_type_ids=token_type_ids, + inputs_embeds=inputs_embeds, + ) + encoder_outputs = self.encoder( + embedding_output, + attention_mask=extended_attention_mask, + head_mask=head_mask, + encoder_hidden_states=encoder_hidden_states, + encoder_attention_mask=encoder_extended_attention_mask, + 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 = encoder_outputs[0] + pooled_output = self.pooler(sequence_output) if self.pooler is not None else None + + if not return_dict: + return (sequence_output, pooled_output) + encoder_outputs[1:] + + return BaseModelOutputWithPoolingAndCrossAttentions( + last_hidden_state=sequence_output, + pooler_output=pooled_output, + past_key_values=encoder_outputs.past_key_values, + hidden_states=encoder_outputs.hidden_states, + attentions=encoder_outputs.attentions, + cross_attentions=encoder_outputs.cross_attentions, + ) + + +@add_start_docstrings( + """ + Nezha Model with two heads on top as done during the pretraining: a `masked language modeling` head and a `next + sentence prediction (classification)` head. + """, + NEZHA_START_DOCSTRING, +) +class NezhaForPreTraining(NezhaPreTrainedModel): + def __init__(self, config): + super().__init__(config) + + self.nezha = NezhaModel(config) + self.cls = NezhaPreTrainingHeads(config) + + # Initialize weights and apply final processing + self.post_init() + + def get_output_embeddings(self): + return self.cls.predictions.decoder + + def set_output_embeddings(self, new_embeddings): + self.cls.predictions.decoder = new_embeddings + + @add_start_docstrings_to_model_forward(NEZHA_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @replace_return_docstrings(output_type=NezhaForPreTrainingOutput, config_class=_CONFIG_FOR_DOC) + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + labels: Optional[torch.Tensor] = None, + next_sentence_label: Optional[torch.Tensor] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple[torch.Tensor], NezhaForPreTrainingOutput]: + r""" + labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ..., + config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), + the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]` + next_sentence_label (`torch.LongTensor` of shape `(batch_size,)`, *optional*): + Labels for computing the next sequence prediction (classification) loss. Input should be a sequence + pair (see `input_ids` docstring) Indices should be in `[0, 1]`: + + - 0 indicates sequence B is a continuation of sequence A, + - 1 indicates sequence B is a random sequence. + kwargs (`Dict[str, any]`, optional, defaults to *{}*): + Used to hide legacy arguments that have been deprecated. + + Returns: + + Example: + + ```python + >>> from transformers import BertTokenizer, NezhaForPreTraining + >>> import torch + + >>> tokenizer = BertTokenizer.from_pretrained("sijunhe/nezha-cn-base") + >>> model = NezhaForPreTraining.from_pretrained("sijunhe/nezha-cn-base") + + >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt") + >>> outputs = model(**inputs) + + >>> prediction_logits = outputs.prediction_logits + >>> seq_relationship_logits = outputs.seq_relationship_logits + ``` + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + outputs = self.nezha( + input_ids, + attention_mask=attention_mask, + token_type_ids=token_type_ids, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + sequence_output, pooled_output = outputs[:2] + prediction_scores, seq_relationship_score = self.cls(sequence_output, pooled_output) + + total_loss = None + if labels is not None and next_sentence_label is not None: + loss_fct = CrossEntropyLoss() + masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) + next_sentence_loss = loss_fct(seq_relationship_score.view(-1, 2), next_sentence_label.view(-1)) + total_loss = masked_lm_loss + next_sentence_loss + + if not return_dict: + output = (prediction_scores, seq_relationship_score) + outputs[2:] + return ((total_loss,) + output) if total_loss is not None else output + + return NezhaForPreTrainingOutput( + loss=total_loss, + prediction_logits=prediction_scores, + seq_relationship_logits=seq_relationship_score, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + +@add_start_docstrings("""Nezha Model with a `language modeling` head on top.""", NEZHA_START_DOCSTRING) +class NezhaForMaskedLM(NezhaPreTrainedModel): + + _keys_to_ignore_on_load_unexpected = [r"pooler"] + _keys_to_ignore_on_load_missing = [r"predictions.decoder.bias", r"positions_encoding"] + + def __init__(self, config): + super().__init__(config) + + if config.is_decoder: + logger.warning( + "If you want to use `NezhaForMaskedLM` make sure `config.is_decoder=False` for " + "bi-directional self-attention." + ) + + self.nezha = NezhaModel(config, add_pooling_layer=False) + self.cls = NezhaOnlyMLMHead(config) + + # Initialize weights and apply final processing + self.post_init() + + def get_output_embeddings(self): + return self.cls.predictions.decoder + + def set_output_embeddings(self, new_embeddings): + self.cls.predictions.decoder = new_embeddings + + @add_start_docstrings_to_model_forward(NEZHA_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @add_code_sample_docstrings( + processor_class=_TOKENIZER_FOR_DOC, + checkpoint=_CHECKPOINT_FOR_DOC, + output_type=MaskedLMOutput, + config_class=_CONFIG_FOR_DOC, + ) + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + encoder_hidden_states: Optional[torch.Tensor] = None, + encoder_attention_mask: Optional[torch.Tensor] = None, + labels: Optional[torch.Tensor] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple[torch.Tensor], MaskedLMOutput]: + r""" + labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ..., + config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the + loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]` + """ + + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + outputs = self.nezha( + input_ids, + attention_mask=attention_mask, + token_type_ids=token_type_ids, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + encoder_hidden_states=encoder_hidden_states, + encoder_attention_mask=encoder_attention_mask, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + sequence_output = outputs[0] + prediction_scores = self.cls(sequence_output) + + masked_lm_loss = None + if labels is not None: + loss_fct = CrossEntropyLoss() # -100 index = padding token + masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) + + if not return_dict: + output = (prediction_scores,) + outputs[2:] + return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output + + return MaskedLMOutput( + loss=masked_lm_loss, + logits=prediction_scores, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + def prepare_inputs_for_generation(self, input_ids, attention_mask=None, **model_kwargs): + input_shape = input_ids.shape + effective_batch_size = input_shape[0] + + # add a dummy token + if self.config.pad_token_id is None: + raise ValueError("The PAD token should be defined for generation") + + attention_mask = torch.cat([attention_mask, attention_mask.new_zeros((attention_mask.shape[0], 1))], dim=-1) + dummy_token = torch.full( + (effective_batch_size, 1), self.config.pad_token_id, dtype=torch.long, device=input_ids.device + ) + input_ids = torch.cat([input_ids, dummy_token], dim=1) + + return {"input_ids": input_ids, "attention_mask": attention_mask} + + +@add_start_docstrings( + """Nezha Model with a `next sentence prediction (classification)` head on top.""", + NEZHA_START_DOCSTRING, +) +class NezhaForNextSentencePrediction(NezhaPreTrainedModel): + def __init__(self, config): + super().__init__(config) + + self.nezha = NezhaModel(config) + self.cls = NezhaOnlyNSPHead(config) + + # Initialize weights and apply final processing + self.post_init() + + @add_start_docstrings_to_model_forward(NEZHA_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @replace_return_docstrings(output_type=NextSentencePredictorOutput, config_class=_CONFIG_FOR_DOC) + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + labels: Optional[torch.Tensor] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + **kwargs, + ) -> Union[Tuple[torch.Tensor], NextSentencePredictorOutput]: + r""" + labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): + Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair + (see `input_ids` docstring). Indices should be in `[0, 1]`: + + - 0 indicates sequence B is a continuation of sequence A, + - 1 indicates sequence B is a random sequence. + + Returns: + + Example: + + ```python + >>> from transformers import BertTokenizer, NezhaForNextSentencePrediction + >>> import torch + + >>> tokenizer = BertTokenizer.from_pretrained("sijunhe/nezha-cn-base") + >>> model = NezhaForNextSentencePrediction.from_pretrained("sijunhe/nezha-cn-base") + + >>> prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced." + >>> next_sentence = "The sky is blue due to the shorter wavelength of blue light." + >>> encoding = tokenizer(prompt, next_sentence, return_tensors="pt") + + >>> outputs = model(**encoding, labels=torch.LongTensor([1])) + >>> logits = outputs.logits + >>> assert logits[0, 0] < logits[0, 1] # next sentence was random + ``` + """ + + if "next_sentence_label" in kwargs: + warnings.warn( + "The `next_sentence_label` argument is deprecated and will be removed in a future version, use" + " `labels` instead.", + FutureWarning, + ) + labels = kwargs.pop("next_sentence_label") + + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + outputs = self.nezha( + input_ids, + attention_mask=attention_mask, + token_type_ids=token_type_ids, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + pooled_output = outputs[1] + + seq_relationship_scores = self.cls(pooled_output) + + next_sentence_loss = None + if labels is not None: + loss_fct = CrossEntropyLoss() + next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), labels.view(-1)) + + if not return_dict: + output = (seq_relationship_scores,) + outputs[2:] + return ((next_sentence_loss,) + output) if next_sentence_loss is not None else output + + return NextSentencePredictorOutput( + loss=next_sentence_loss, + logits=seq_relationship_scores, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + +@add_start_docstrings( + """ + Nezha Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled + output) e.g. for GLUE tasks. + """, + NEZHA_START_DOCSTRING, +) +class NezhaForSequenceClassification(NezhaPreTrainedModel): + def __init__(self, config): + super().__init__(config) + self.num_labels = config.num_labels + self.config = config + + self.nezha = NezhaModel(config) + classifier_dropout = ( + config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob + ) + self.dropout = nn.Dropout(classifier_dropout) + self.classifier = nn.Linear(config.hidden_size, config.num_labels) + + # Initialize weights and apply final processing + self.post_init() + + @add_start_docstrings_to_model_forward(NEZHA_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @add_code_sample_docstrings( + processor_class=_TOKENIZER_FOR_DOC, + checkpoint=_CHECKPOINT_FOR_DOC, + output_type=SequenceClassifierOutput, + config_class=_CONFIG_FOR_DOC, + ) + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + labels: Optional[torch.Tensor] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple[torch.Tensor], SequenceClassifierOutput]: + r""" + labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): + Labels for computing the sequence classification/regression loss. Indices should be in `[0, ..., + config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If + `config.num_labels > 1` a classification loss is computed (Cross-Entropy). + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + outputs = self.nezha( + input_ids, + attention_mask=attention_mask, + token_type_ids=token_type_ids, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + pooled_output = outputs[1] + + pooled_output = self.dropout(pooled_output) + logits = self.classifier(pooled_output) + + loss = None + if labels is not None: + if self.config.problem_type is None: + if self.num_labels == 1: + self.config.problem_type = "regression" + elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int): + self.config.problem_type = "single_label_classification" + else: + self.config.problem_type = "multi_label_classification" + + if self.config.problem_type == "regression": + loss_fct = MSELoss() + if self.num_labels == 1: + loss = loss_fct(logits.squeeze(), labels.squeeze()) + else: + loss = loss_fct(logits, labels) + elif self.config.problem_type == "single_label_classification": + loss_fct = CrossEntropyLoss() + loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1)) + elif self.config.problem_type == "multi_label_classification": + loss_fct = BCEWithLogitsLoss() + loss = loss_fct(logits, labels) + if not return_dict: + output = (logits,) + outputs[2:] + return ((loss,) + output) if loss is not None else output + + return SequenceClassifierOutput( + loss=loss, + logits=logits, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + +@add_start_docstrings( + """ + Nezha Model with a multiple choice classification head on top (a linear layer on top of the pooled output and a + softmax) e.g. for RocStories/SWAG tasks. + """, + NEZHA_START_DOCSTRING, +) +class NezhaForMultipleChoice(NezhaPreTrainedModel): + def __init__(self, config): + super().__init__(config) + + self.nezha = NezhaModel(config) + classifier_dropout = ( + config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob + ) + self.dropout = nn.Dropout(classifier_dropout) + self.classifier = nn.Linear(config.hidden_size, 1) + + # Initialize weights and apply final processing + self.post_init() + + @add_start_docstrings_to_model_forward(NEZHA_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length")) + @add_code_sample_docstrings( + processor_class=_TOKENIZER_FOR_DOC, + checkpoint=_CHECKPOINT_FOR_DOC, + output_type=MultipleChoiceModelOutput, + config_class=_CONFIG_FOR_DOC, + ) + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + labels: Optional[torch.Tensor] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple[torch.Tensor], MultipleChoiceModelOutput]: + r""" + labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): + Labels for computing the multiple choice classification loss. Indices should be in `[0, ..., + num_choices-1]` where `num_choices` is the size of the second dimension of the input tensors. (See + `input_ids` above) + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1] + input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None + attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None + token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1)) if token_type_ids is not None else None + inputs_embeds = ( + inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1)) + if inputs_embeds is not None + else None + ) + + outputs = self.nezha( + input_ids, + attention_mask=attention_mask, + token_type_ids=token_type_ids, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + pooled_output = outputs[1] + print(pooled_output.shape) + pooled_output = self.dropout(pooled_output) + logits = self.classifier(pooled_output) + print(logits.shape) + print(num_choices) + reshaped_logits = logits.view(-1, num_choices) + + loss = None + if labels is not None: + loss_fct = CrossEntropyLoss() + loss = loss_fct(reshaped_logits, labels) + + if not return_dict: + output = (reshaped_logits,) + outputs[2:] + return ((loss,) + output) if loss is not None else output + + return MultipleChoiceModelOutput( + loss=loss, + logits=reshaped_logits, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + +@add_start_docstrings( + """ + Nezha Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for + Named-Entity-Recognition (NER) tasks. + """, + NEZHA_START_DOCSTRING, +) +class NezhaForTokenClassification(NezhaPreTrainedModel): + + _keys_to_ignore_on_load_unexpected = [r"pooler"] + + def __init__(self, config): + super().__init__(config) + self.num_labels = config.num_labels + + self.nezha = NezhaModel(config, add_pooling_layer=False) + classifier_dropout = ( + config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob + ) + self.dropout = nn.Dropout(classifier_dropout) + self.classifier = nn.Linear(config.hidden_size, config.num_labels) + + # Initialize weights and apply final processing + self.post_init() + + @add_start_docstrings_to_model_forward(NEZHA_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @add_code_sample_docstrings( + processor_class=_TOKENIZER_FOR_DOC, + checkpoint=_CHECKPOINT_FOR_DOC, + output_type=TokenClassifierOutput, + config_class=_CONFIG_FOR_DOC, + ) + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + labels: Optional[torch.Tensor] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple[torch.Tensor], TokenClassifierOutput]: + r""" + labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`. + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + outputs = self.nezha( + input_ids, + attention_mask=attention_mask, + token_type_ids=token_type_ids, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + sequence_output = outputs[0] + + sequence_output = self.dropout(sequence_output) + logits = self.classifier(sequence_output) + + loss = None + if labels is not None: + loss_fct = CrossEntropyLoss() + loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1)) + + if not return_dict: + output = (logits,) + outputs[2:] + return ((loss,) + output) if loss is not None else output + + return TokenClassifierOutput( + loss=loss, + logits=logits, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + +@add_start_docstrings( + """ + Nezha Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear + layers on top of the hidden-states output to compute `span start logits` and `span end logits`). + """, + NEZHA_START_DOCSTRING, +) +class NezhaForQuestionAnswering(NezhaPreTrainedModel): + + _keys_to_ignore_on_load_unexpected = [r"pooler"] + + def __init__(self, config): + super().__init__(config) + self.num_labels = config.num_labels + + self.nezha = NezhaModel(config, add_pooling_layer=False) + self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels) + + # Initialize weights and apply final processing + self.post_init() + + @add_start_docstrings_to_model_forward(NEZHA_INPUTS_DOCSTRING.format("batch_size, sequence_length")) + @add_code_sample_docstrings( + processor_class=_TOKENIZER_FOR_DOC, + checkpoint=_CHECKPOINT_FOR_DOC, + output_type=QuestionAnsweringModelOutput, + config_class=_CONFIG_FOR_DOC, + ) + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + start_positions: Optional[torch.Tensor] = None, + end_positions: Optional[torch.Tensor] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple[torch.Tensor], QuestionAnsweringModelOutput]: + r""" + start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*): + Labels for position (index) of the start of the labelled span for computing the token classification loss. + Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence + are not taken into account for computing the loss. + end_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*): + Labels for position (index) of the end of the labelled span for computing the token classification loss. + Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence + are not taken into account for computing the loss. + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + outputs = self.nezha( + input_ids, + attention_mask=attention_mask, + token_type_ids=token_type_ids, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + + sequence_output = outputs[0] + + logits = self.qa_outputs(sequence_output) + start_logits, end_logits = logits.split(1, dim=-1) + start_logits = start_logits.squeeze(-1).contiguous() + end_logits = end_logits.squeeze(-1).contiguous() + + total_loss = None + if start_positions is not None and end_positions is not None: + # If we are on multi-GPU, split add a dimension + if len(start_positions.size()) > 1: + start_positions = start_positions.squeeze(-1) + if len(end_positions.size()) > 1: + end_positions = end_positions.squeeze(-1) + # sometimes the start/end positions are outside our model inputs, we ignore these terms + ignored_index = start_logits.size(1) + start_positions = start_positions.clamp(0, ignored_index) + end_positions = end_positions.clamp(0, ignored_index) + + loss_fct = CrossEntropyLoss(ignore_index=ignored_index) + start_loss = loss_fct(start_logits, start_positions) + end_loss = loss_fct(end_logits, end_positions) + total_loss = (start_loss + end_loss) / 2 + + if not return_dict: + output = (start_logits, end_logits) + outputs[2:] + return ((total_loss,) + output) if total_loss is not None else output + + return QuestionAnsweringModelOutput( + loss=total_loss, + start_logits=start_logits, + end_logits=end_logits, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) diff --git a/src/transformers/utils/dummy_pt_objects.py b/src/transformers/utils/dummy_pt_objects.py index 847e7d87ab..3a06175fef 100644 --- a/src/transformers/utils/dummy_pt_objects.py +++ b/src/transformers/utils/dummy_pt_objects.py @@ -3130,6 +3130,72 @@ class MT5Model(metaclass=DummyObject): requires_backends(self, ["torch"]) +NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST = None + + +class NezhaForMaskedLM(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class NezhaForMultipleChoice(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class NezhaForNextSentencePrediction(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class NezhaForPreTraining(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class NezhaForQuestionAnswering(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class NezhaForSequenceClassification(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class NezhaForTokenClassification(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class NezhaModel(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class NezhaPreTrainedModel(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + NYSTROMFORMER_PRETRAINED_MODEL_ARCHIVE_LIST = None diff --git a/src/transformers/utils/fx.py b/src/transformers/utils/fx.py index 62f0c98b4c..6fed5808f8 100644 --- a/src/transformers/utils/fx.py +++ b/src/transformers/utils/fx.py @@ -115,6 +115,7 @@ _REGULAR_SUPPORTED_MODEL_NAMES_AND_TASKS = [ "megatron-bert", "mobilebert", "mt5", + "nezha", "opt", "pegasus", "plbart", diff --git a/tests/models/nezha/__init__.py b/tests/models/nezha/__init__.py new file mode 100644 index 0000000000..e69de29bb2 diff --git a/tests/models/nezha/test_modeling_nezha.py b/tests/models/nezha/test_modeling_nezha.py new file mode 100644 index 0000000000..1d8017be95 --- /dev/null +++ b/tests/models/nezha/test_modeling_nezha.py @@ -0,0 +1,478 @@ +# 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 os +import tempfile +import unittest + +from transformers import NezhaConfig, is_torch_available +from transformers.models.auto import get_values +from transformers.testing_utils import require_torch, require_torch_gpu, slow, torch_device + +from ...generation.test_generation_utils import GenerationTesterMixin +from ...test_configuration_common import ConfigTester +from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor, random_attention_mask + + +if is_torch_available(): + import torch + + from transformers import ( + MODEL_FOR_PRETRAINING_MAPPING, + NezhaForMaskedLM, + NezhaForMultipleChoice, + NezhaForNextSentencePrediction, + NezhaForPreTraining, + NezhaForQuestionAnswering, + NezhaForSequenceClassification, + NezhaForTokenClassification, + NezhaModel, + ) + from transformers.models.nezha.modeling_nezha import NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST + + +class NezhaModelTester: + def __init__( + self, + parent, + batch_size=13, + seq_length=7, + is_training=True, + use_input_mask=True, + use_token_type_ids=True, + 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): + """ + Returns a tiny configuration by default. + """ + return NezhaConfig( + 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 prepare_config_and_inputs_for_decoder(self): + ( + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + ) = self.prepare_config_and_inputs() + + config.is_decoder = True + encoder_hidden_states = floats_tensor([self.batch_size, self.seq_length, self.hidden_size]) + encoder_attention_mask = ids_tensor([self.batch_size, self.seq_length], vocab_size=2) + + return ( + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + encoder_hidden_states, + encoder_attention_mask, + ) + + def create_and_check_model( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + model = NezhaModel(config=config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids) + result = model(input_ids, token_type_ids=token_type_ids) + result = model(input_ids) + self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size)) + self.parent.assertEqual(result.pooler_output.shape, (self.batch_size, self.hidden_size)) + + def create_and_check_model_as_decoder( + self, + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + encoder_hidden_states, + encoder_attention_mask, + ): + config.add_cross_attention = True + model = NezhaModel(config) + model.to(torch_device) + model.eval() + result = model( + input_ids, + attention_mask=input_mask, + token_type_ids=token_type_ids, + encoder_hidden_states=encoder_hidden_states, + encoder_attention_mask=encoder_attention_mask, + ) + result = model( + input_ids, + attention_mask=input_mask, + token_type_ids=token_type_ids, + encoder_hidden_states=encoder_hidden_states, + ) + result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids) + self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size)) + self.parent.assertEqual(result.pooler_output.shape, (self.batch_size, self.hidden_size)) + + def create_and_check_for_masked_lm( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + model = NezhaForMaskedLM(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_for_next_sequence_prediction( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + model = NezhaForNextSentencePrediction(config=config) + model.to(torch_device) + model.eval() + result = model( + input_ids, + attention_mask=input_mask, + token_type_ids=token_type_ids, + labels=sequence_labels, + ) + self.parent.assertEqual(result.logits.shape, (self.batch_size, 2)) + + def create_and_check_for_pretraining( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + model = NezhaForPreTraining(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, + next_sentence_label=sequence_labels, + ) + self.parent.assertEqual(result.prediction_logits.shape, (self.batch_size, self.seq_length, self.vocab_size)) + self.parent.assertEqual(result.seq_relationship_logits.shape, (self.batch_size, 2)) + + def create_and_check_for_question_answering( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + model = NezhaForQuestionAnswering(config=config) + model.to(torch_device) + model.eval() + result = model( + input_ids, + attention_mask=input_mask, + token_type_ids=token_type_ids, + start_positions=sequence_labels, + end_positions=sequence_labels, + ) + self.parent.assertEqual(result.start_logits.shape, (self.batch_size, self.seq_length)) + self.parent.assertEqual(result.end_logits.shape, (self.batch_size, self.seq_length)) + + def create_and_check_for_sequence_classification( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + config.num_labels = self.num_labels + model = NezhaForSequenceClassification(config) + model.to(torch_device) + model.eval() + result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=sequence_labels) + self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_labels)) + + def create_and_check_for_token_classification( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + config.num_labels = self.num_labels + model = NezhaForTokenClassification(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.num_labels)) + + def create_and_check_for_multiple_choice( + self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels + ): + config.num_choices = self.num_choices + model = NezhaForMultipleChoice(config=config) + model.to(torch_device) + model.eval() + multiple_choice_inputs_ids = input_ids.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous() + multiple_choice_token_type_ids = token_type_ids.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous() + multiple_choice_input_mask = input_mask.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous() + result = model( + multiple_choice_inputs_ids, + attention_mask=multiple_choice_input_mask, + token_type_ids=multiple_choice_token_type_ids, + labels=choice_labels, + ) + self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_choices)) + + 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, "token_type_ids": token_type_ids, "attention_mask": input_mask} + return config, inputs_dict + + +@require_torch +class NezhaModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase): + + all_model_classes = ( + ( + NezhaModel, + NezhaForMaskedLM, + NezhaForMultipleChoice, + NezhaForNextSentencePrediction, + NezhaForPreTraining, + NezhaForQuestionAnswering, + NezhaForSequenceClassification, + NezhaForTokenClassification, + ) + if is_torch_available() + else () + ) + fx_compatible = True + + # special case for ForPreTraining model + def _prepare_for_class(self, inputs_dict, model_class, return_labels=False): + inputs_dict = super()._prepare_for_class(inputs_dict, model_class, return_labels=return_labels) + + if return_labels: + if model_class in get_values(MODEL_FOR_PRETRAINING_MAPPING): + inputs_dict["labels"] = torch.zeros( + (self.model_tester.batch_size, self.model_tester.seq_length), dtype=torch.long, device=torch_device + ) + inputs_dict["next_sentence_label"] = torch.zeros( + self.model_tester.batch_size, dtype=torch.long, device=torch_device + ) + return inputs_dict + + def setUp(self): + self.model_tester = NezhaModelTester(self) + self.config_tester = ConfigTester(self, config_class=NezhaConfig, 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_as_decoder(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs_for_decoder() + self.model_tester.create_and_check_model_as_decoder(*config_and_inputs) + + def test_model_as_decoder_with_default_input_mask(self): + # This regression test was failing with PyTorch < 1.3 + ( + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + encoder_hidden_states, + encoder_attention_mask, + ) = self.model_tester.prepare_config_and_inputs_for_decoder() + + input_mask = None + + self.model_tester.create_and_check_model_as_decoder( + config, + input_ids, + token_type_ids, + input_mask, + sequence_labels, + token_labels, + choice_labels, + encoder_hidden_states, + encoder_attention_mask, + ) + + def test_for_masked_lm(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_for_masked_lm(*config_and_inputs) + + def test_for_multiple_choice(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_for_multiple_choice(*config_and_inputs) + + def test_for_next_sequence_prediction(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_for_next_sequence_prediction(*config_and_inputs) + + def test_for_pretraining(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_for_pretraining(*config_and_inputs) + + def test_for_question_answering(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_for_question_answering(*config_and_inputs) + + def test_for_sequence_classification(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_for_sequence_classification(*config_and_inputs) + + def test_for_token_classification(self): + config_and_inputs = self.model_tester.prepare_config_and_inputs() + self.model_tester.create_and_check_for_token_classification(*config_and_inputs) + + @slow + def test_model_from_pretrained(self): + for model_name in NEZHA_PRETRAINED_MODEL_ARCHIVE_LIST[:1]: + model = NezhaModel.from_pretrained(model_name) + self.assertIsNotNone(model) + + @slow + @require_torch_gpu + def test_torchscript_device_change(self): + config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common() + for model_class in self.all_model_classes: + + # NezhaForMultipleChoice behaves incorrectly in JIT environments. + if model_class == NezhaForMultipleChoice: + return + + config.torchscript = True + model = model_class(config=config) + + inputs_dict = self._prepare_for_class(inputs_dict, model_class) + traced_model = torch.jit.trace( + model, (inputs_dict["input_ids"].to("cpu"), inputs_dict["attention_mask"].to("cpu")) + ) + + with tempfile.TemporaryDirectory() as tmp: + torch.jit.save(traced_model, os.path.join(tmp, "bert.pt")) + loaded = torch.jit.load(os.path.join(tmp, "bert.pt"), map_location=torch_device) + loaded(inputs_dict["input_ids"].to(torch_device), inputs_dict["attention_mask"].to(torch_device)) + + +@require_torch +class NezhaModelIntegrationTest(unittest.TestCase): + @slow + def test_inference_nezha_model(self): + model = NezhaModel.from_pretrained("sijunhe/nezha-cn-base") + input_ids = torch.tensor([[0, 1, 2, 3, 4, 5]]) + attention_mask = torch.tensor([[0, 1, 1, 1, 1, 1]]) + with torch.no_grad(): + output = model(input_ids, attention_mask=attention_mask)[0] + expected_shape = torch.Size((1, 6, 768)) + self.assertEqual(output.shape, expected_shape) + expected_slice = torch.tensor([[[0.0685, 0.2441, 0.1102], [0.0600, 0.1906, 0.1349], [0.0221, 0.0819, 0.0586]]]) + + self.assertTrue(torch.allclose(output[:, 1:4, 1:4], expected_slice, atol=1e-4)) + + @slow + def test_inference_nezha_masked_lm(self): + model = NezhaForMaskedLM.from_pretrained("sijunhe/nezha-cn-base") + input_ids = torch.tensor([[0, 1, 2, 3, 4, 5]]) + attention_mask = torch.tensor([[1, 1, 1, 1, 1, 1]]) + with torch.no_grad(): + output = model(input_ids, attention_mask=attention_mask)[0] + expected_shape = torch.Size((1, 6, 21128)) + self.assertEqual(output.shape, expected_shape) + expected_slice = torch.tensor( + [[-2.7939, -1.7902, -2.2189], [-2.8585, -1.8908, -2.3723], [-2.6499, -1.7750, -2.2558]] + ) + + self.assertTrue(torch.allclose(output[:, 1:4, 1:4], expected_slice, atol=1e-4))