Merge branch 'master' into master

.github/workflows/unit_test.yml

@@ -18,7 +18,7 @@ jobs:
   # This workflow contains a single job called "build"
   build:
     # The type of runner that the job will run on
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-20.04
 
     # Steps represent a sequence of tasks that will be executed as part of the job
     steps:
@@ -28,7 +28,8 @@ jobs:
       - name: Set up Python 3.6
         uses: actions/setup-python@v2
         with:
-          python-version: 3.6
+          python-version: '3.6' # 指定python版本
+
 
       # Runs a single command using the runners shell
       - name: Run a one-line script

README.cn.md

@@ -22,7 +22,12 @@
 
 随着 BERT、Megatron、GPT-3 等预训练模型在NLP领域取得瞩目的成果，越来越多团队投身到超大规模训练中，这使得训练模型的规模从亿级别发展到了千亿甚至万亿的规模。然而，这类超大规模的模型运用于实际场景中仍然有一些挑战。首先，模型参数量过大使得训练和推理速度过慢且部署成本极高；其次在很多实际场景中数据量不足的问题仍然制约着大模型在小样本场景中的应用，提高预训练模型在小样本场景的泛化性依然存在挑战。为了应对以上问题，PAI 团队推出了 EasyNLP 中文 NLP 算法框架，助力大模型快速且高效的落地。
 
-
+- [PAI-Diffusion模型来了！阿里云机器学习团队带您徜徉中文艺术海洋](https://zhuanlan.zhihu.com/p/590020134)
+- [模型精度再被提升，统一跨任务小样本学习算法 UPT 给出解法!](https://zhuanlan.zhihu.com/p/590611518)
+- [Span抽取和元学习能碰撞出怎样的新火花，小样本实体识别来告诉你!](https://zhuanlan.zhihu.com/p/590297824)
+- [算法 KECP 被顶会 EMNLP 收录，极少训练数据就能实现机器阅读理解](https://zhuanlan.zhihu.com/p/590024650)
+- [当大火的文图生成模型遇见知识图谱，AI画像趋近于真实世界](https://zhuanlan.zhihu.com/p/581870071)
+- [EasyNLP发布融合语言学和事实知识的中文预训练模型CKBERT](https://zhuanlan.zhihu.com/p/574853281)
 - [EasyNLP带你实现中英文机器阅读理解](https://zhuanlan.zhihu.com/p/568890245)
 - [跨模态学习能力再升级，EasyNLP电商文图检索效果刷新SOTA](https://zhuanlan.zhihu.com/p/568512230)
 - [EasyNLP玩转文本摘要（新闻标题）生成](https://zhuanlan.zhihu.com/p/566607127)

README.md

@@ -35,7 +35,11 @@ EasyNLP is an easy-to-use NLP development and application toolkit in PyTorch, fi
 
 We have a series of technical articles on the functionalities of EasyNLP.
 
-
+- [阿里云PAI-Diffusion功能再升级，全链路支持模型调优，平均推理速度提升75%以上](https://zhuanlan.zhihu.com/p/604483551)
+- [PAI-Diffusion模型来了！阿里云机器学习团队带您徜徉中文艺术海洋](https://zhuanlan.zhihu.com/p/590020134)
+- [模型精度再被提升，统一跨任务小样本学习算法 UPT 给出解法!](https://zhuanlan.zhihu.com/p/590611518)
+- [Span抽取和元学习能碰撞出怎样的新火花，小样本实体识别来告诉你!](https://zhuanlan.zhihu.com/p/590297824)
+- [算法 KECP 被顶会 EMNLP 收录，极少训练数据就能实现机器阅读理解](https://zhuanlan.zhihu.com/p/590024650)
 - [当大火的文图生成模型遇见知识图谱，AI画像趋近于真实世界](https://zhuanlan.zhihu.com/p/581870071)
 - [EasyNLP发布融合语言学和事实知识的中文预训练模型CKBERT](https://zhuanlan.zhihu.com/p/574853281)
 - [EasyNLP带你实现中英文机器阅读理解](https://zhuanlan.zhihu.com/p/568890245)

easynlp/appzoo/__init__.py

@@ -25,21 +25,24 @@
     "text_match.model": ['TextMatch', 'TextMatchTwoTower', 'DistillatoryTextMatch', 'FewshotSingleTowerTextMatch', 'CptFewshotSingleTowerTextMatch'],
     "data_augmentation.model": ["DataAugmentation"],
     "geep_classification.model": ["GEEPClassification"],
+    "text2video_retrieval.model": ["Text2VideoRetrieval"],
     "clip.model": ["CLIPApp"],
-#   "latent_diffusion.model": ["LatentDiffusion"],
+    "latent_diffusion.model": ["LatentDiffusion","StableDiffusion"],
     "wukong_clip.model": ["WukongCLIP"],
     "text2image_generation.model": ["TextImageGeneration", "TextImageGeneration_knowl"],
     "image2text_generation.model": ['VQGANGPTImageTextGeneration', 'CLIPGPTImageTextGeneration'], 
     "video2text_generation.model": ['CLIPGPTFrameTextGeneration'], 
     "sequence_generation.model": ["SequenceGeneration"], 
     "machine_reading_comprehension.model": ["MachineReadingComprehension"],
     "open_domain_dialogue.model": ["OpenDomainDialogue"],
+    "information_extraction.model": ["InformationExtractionModel"], 
 
     "sequence_classification.evaluator": ['SequenceClassificationEvaluator', 'SequenceMultiLabelClassificationEvaluator'],
     "sequence_labeling.evaluator": ['SequenceLabelingEvaluator'],
     "language_modeling.evaluator": ['LanguageModelingEvaluator'],
     "text_match.evaluator": ['TextMatchEvaluator'],
     "geep_classification.evaluator": ['GEEPClassificationEvaluator'],
+    "text2video_retrieval.evaluator": ['Text2VideoRetrievalEvaluator'],
     "clip.evaluator": ['CLIPEvaluator'],
     "wukong_clip.evaluator": ['WukongCLIPEvaluator'],
     "text2image_generation.evaluator": ["TextImageGenerationEvaluator"],
@@ -48,25 +51,30 @@
     "sequence_generation.evaluator": ["SequenceGenerationEvaluator"], 
     "machine_reading_comprehension.evaluator": ["MachineReadingComprehensionEvaluator"],
     "open_domain_dialogue.evaluator": ["OpenDomainDialogueEvaluator"],
+    "information_extraction.evaluator": ["InformationExtractionEvaluator"],
+    "latent_diffusion.evaluator": ["LatentDiffusionModelEvaluator"],
 
     "sequence_classification.predictor": ['SequenceClassificationPredictor', 'FewshotSequenceClassificationPredictor', 'CptFewshotSequenceClassificationPredictor'],
     "sequence_labeling.predictor": ['SequenceLabelingPredictor'],
     "feature_vectorization.predictor": ['FeatureVectorizationPredictor'],
     "text_match.predictor": ['TextMatchPredictor', 'TextMatchTwoTowerPredictor', 'FewshotSingleTowerTextMatchPredictor', 'CptFewshotSingleTowerTextMatchPredictor'],
     "data_augmentation.predictor": ['DataAugmentationPredictor'],
     "geep_classification.predictor": ['GEEPClassificationPredictor'],
+    "text2video_retrieval.predictor": ['Text2VideoRetrievalPredictor'],
     "clip.predictor": ['CLIPPredictor'],
-#   "latent_diffusion.predictor": ['LatentDiffusionPredictor'],
+    "latent_diffusion.predictor": ['LatentDiffusionPredictor'],
     "wukong_clip.predictor": ['WukongCLIPPredictor'],
     "text2image_generation.predictor": ['TextImageGenerationPredictor', 'TextImageGenerationKnowlPredictor'],
     "image2text_generation.predictor": ['VQGANGPTImageTextGenerationPredictor', 'CLIPGPTImageTextGenerationPredictor'],
     "video2text_generation.predictor": ['CLIPGPTFrameTextGenerationPredictor'],
     "sequence_generation.predictor": ['SequenceGenerationPredictor'],
     "machine_reading_comprehension.predictor": ["MachineReadingComprehensionPredictor"],
     "open_domain_dialogue.predictor": ["OpenDomainDialoguePredictor"],
+    "information_extraction.predictor": ["InformationExtractionPredictor"],
 
     "geep_classification.data": ['GEEPClassificationDataset'],
     "language_modeling.data": ['LanguageModelingDataset'],
+    "text2video_retrieval.data": ['Text2VideoRetrievalDataset'],
     "clip.data": ['CLIPDataset'],
     "wukong_clip.data": ['WukongCLIPDataset'],
     "sequence_classification.data": ['ClassificationDataset', 'DistillatoryClassificationDataset', 'FewshotSequenceClassificationDataset'],
@@ -78,6 +86,9 @@
     "sequence_generation.data": ['SequenceGenerationDataset'],
     "machine_reading_comprehension.data": ["MachineReadingComprehensionDataset"],
     "open_domain_dialogue.data": ['OpenDomainDialogueDataset'],
+    "information_extraction.data": ["InformationExtractionDataset"],
+    "latent_diffusion.data": ["LdmDataset"],
+
     "dataset": ['BaseDataset', 'GeneralDataset', 'load_dataset', 'list_datasets'],
     "api": ['get_application_dataset', 'get_application_model', 'get_application_model_for_evaluation', 'get_application_evaluator', 'get_application_predictor'],
 }
@@ -90,21 +101,24 @@
     from .text_match.model import TextMatch, TextMatchTwoTower, DistillatoryTextMatch, FewshotSingleTowerTextMatch, CptFewshotSingleTowerTextMatch
     from .data_augmentation.model import DataAugmentation
     from .geep_classification.model import GEEPClassification
+    from .text2video_retrieval.model import Text2VideoRetrieval
     from .clip.model import CLIPApp
-#   from .latent_diffusion.model import LatentDiffusion
+    from .latent_diffusion.model import LatentDiffusion,StableDiffusion
     from .wukong_clip.model import WukongCLIP
     from .text2image_generation.model import TextImageGeneration, TextImageGeneration_knowl
     from .image2text_generation.model import VQGANGPTImageTextGeneration, CLIPGPTImageTextGeneration
     from .video2text_generation.model import CLIPGPTFrameTextGeneration
     from .sequence_generation.model import SequenceGeneration
     from .machine_reading_comprehension.model import MachineReadingComprehension
     from .open_domain_dialogue.model import OpenDomainDialogue
+    from .information_extraction.model import InformationExtractionModel
 
     from .sequence_classification.evaluator import SequenceClassificationEvaluator, SequenceMultiLabelClassificationEvaluator
     from .sequence_labeling.evaluator import SequenceLabelingEvaluator
     from .language_modeling.evaluator import LanguageModelingEvaluator
     from .text_match.evaluator import TextMatchEvaluator
     from .geep_classification.evaluator import GEEPClassificationEvaluator
+    from .text2video_retrieval.evaluator import Text2VideoRetrievalEvaluator
     from .clip.evaluator import CLIPEvaluator
     from .wukong_clip.evaluator import WukongCLIPEvaluator
     from .text2image_generation.evaluator import TextImageGenerationEvaluator
@@ -113,28 +127,33 @@
     from .sequence_generation.evaluator import SequenceGenerationEvaluator
     from .machine_reading_comprehension.evaluator import MachineReadingComprehensionEvaluator
     from .open_domain_dialogue.evaluator import OpenDomainDialogueEvaluator
+    from .information_extraction.evaluator import InformationExtractionEvaluator
+    from .latent_diffusion.evaluator import LatentDiffusionModelEvaluator
 
     from .sequence_classification.predictor import SequenceClassificationPredictor, FewshotSequenceClassificationPredictor, CptFewshotSequenceClassificationPredictor
     from .sequence_labeling.predictor import SequenceLabelingPredictor
     from .feature_vectorization.predictor import FeatureVectorizationPredictor
     from .text_match.predictor import TextMatchPredictor, TextMatchTwoTowerPredictor, FewshotSingleTowerTextMatchPredictor, CptFewshotSingleTowerTextMatchPredictor
     from .data_augmentation.predictor import DataAugmentationPredictor
     from .geep_classification.predictor import GEEPClassificationPredictor
+    from .text2video_retrieval.predictor import Text2VideoRetrievalPredictor
     from .clip.predictor import CLIPPredictor
-#   from .latent_diffusion.predictor import LatentDiffusionPredictor
+    from .latent_diffusion.predictor import LatentDiffusionPredictor
     from .wukong_clip.predictor import WukongCLIPPredictor
     from .text2image_generation.predictor import TextImageGenerationPredictor, TextImageGenerationKnowlPredictor
     from .image2text_generation.predictor import VQGANGPTImageTextGenerationPredictor, CLIPGPTImageTextGenerationPredictor
     from .video2text_generation.predictor import CLIPGPTFrameTextGenerationPredictor
     from .sequence_generation.predictor import SequenceGenerationPredictor
     from .machine_reading_comprehension.predictor import MachineReadingComprehensionPredictor
     from .open_domain_dialogue.predictor import OpenDomainDialoguePredictor
+    from .information_extraction.predictor import InformationExtractionPredictor
 
     from .sequence_classification.data import ClassificationDataset, DistillatoryClassificationDataset, FewshotSequenceClassificationDataset
     from .sequence_labeling.data import SequenceLabelingDataset, SequenceLabelingAutoDataset
     from .language_modeling.data import LanguageModelingDataset
     from .text_match.data import TwoTowerDataset, SingleTowerDataset, DistillatorySingleTowerDataset, FewshotSingleTowerTextMatchDataset, SiameseDataset
     from .geep_classification.data import GEEPClassificationDataset
+    from .text2video_retrieval.data import Text2VideoRetrievalDataset
     from .clip.data import CLIPDataset
     from .wukong_clip.data import WukongCLIPDataset
     from .text2image_generation.data import TextImageDataset, TextImageKnowlDataset
@@ -143,6 +162,8 @@
     from .sequence_generation.data import SequenceGenerationDataset
     from .machine_reading_comprehension.data import MachineReadingComprehensionDataset
     from .open_domain_dialogue.data import OpenDomainDialogueDataset
+    from .information_extraction.data import InformationExtractionDataset
+    from .latent_diffusion.data import LdmDataset
 
     from .dataset import BaseDataset, GeneralDataset
     from .dataset import load_dataset, list_datasets