Skip to content

Commit b3b0558

Browse files
committed
Update examples and README
1 parent e44a567 commit b3b0558

File tree

22 files changed

+4
-3
lines changed

22 files changed

+4
-3
lines changed

README.md

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -13,7 +13,7 @@
1313

1414
## News <img src="./docs/imgs/news_icon.png" width="40"/>
1515

16-
* [2021-12-12] PaddleNLP v2.2版本已发布!:tada: 欢迎体验更快的文本处理[FasterTokenizer](./examples/faster/faster_ernie)、更快的预训练模型[FasterERNIE](./examples/faster/faster_ernie)、更快的文本生成[FasterGeneration](./examples/faster/faster_generation);新推出『解语』名词短语标注工具[NPTag](./examples/text_to_knowledge/nptag)、超快中文小模型[PP-MiniLM](./examples/model_compression/PP-MiniLM)! 更多详细升级信息请查看[Release Note](https://github.com/PaddlePaddle/PaddleNLP/releases/tag/v2.1.0)
16+
* [2021-12-12] PaddleNLP v2.2版本已发布!:tada: 欢迎体验更快的文本处理[FasterTokenizer](./examples/faster/faster_ernie)、更快的预训练模型[FasterERNIE](./examples/faster/faster_ernie)、更快的文本生成[FasterGeneration](./examples/faster/faster_generation);新推出『解语』名词短语标注工具[NPTag](./examples/text_to_knowledge/nptag)、超快中文小模型[PP-MiniLM](./examples/model_compression/pp-minilm)! 更多详细升级信息请查看[Release Note](https://github.com/PaddlePaddle/PaddleNLP/releases/tag/v2.1.0)
1717
* [2021-12-12] 飞桨新产品**端到端问答工具**🚀[RocketQA](https://github.com/PaddlePaddle/RocketQA)全新发布!:tada:
1818

1919
## 简介
@@ -274,10 +274,11 @@ PaddleNLP提供了多粒度、多场景的NLP应用示例,面向动态图模
274274

275275
| 模型 | 简介 |
276276
| :--------------------------------------------------------- | ------------------------------------------------------------ |
277-
| [MiniLMv2](examples/model_compression/minilmv2) :star2: | 基于[MiniLMv2: Multi-Head Self-Attention Relation Distillation for Compressing Pretrained Transformers](https://arxiv.org/abs/2012.15828)论文策略的实现,是一种通用蒸馏方法。本实例以`bert-base-chinese`为教师模型,利用中文数据进行了通用蒸馏。 |
277+
| [MiniLMv2](examples/model_compression/minilmv2) | 基于[MiniLMv2: Multi-Head Self-Attention Relation Distillation for Compressing Pretrained Transformers](https://arxiv.org/abs/2012.15828)论文策略的实现,是一种通用蒸馏方法。本实例以`bert-base-chinese`为教师模型,利用中文数据进行了通用蒸馏。 |
278278
| [TinyBERT](./examples/model_compression/tinybert) | 基于论文[TinyBERT: Distilling BERT for Natural Language Understanding](https://arxiv.org/abs/1909.10351)的实现,提供了通用蒸馏和下游任务蒸馏的脚本。本实例利用开源模型`tinybert-6l-768d-v2`初始化,在GLUE的7个数据集上进行下游任务的蒸馏,最终模型参数量缩小1/2,预测速度提升2倍,同时保证模型精度几乎无损,其中精度可达教师模型`bert-base-uncased`的 98.90%。 |
279-
| [OFA-BERT](./examples/model_compression/ofa/) :star2: | 基于PaddleSlim Once-For-ALL(OFA)策略对BERT在GLUE任务的下游模型进行压缩,在精度无损的情况下可减少33%参数量,达到模型小型化的提速的效果。 |
279+
| [OFA-BERT](./examples/model_compression/ofa/) | 基于PaddleSlim Once-For-ALL(OFA)策略对BERT在GLUE任务的下游模型进行压缩,在精度无损的情况下可减少33%参数量,达到模型小型化的提速的效果。 |
280280
| [Distill-LSTM](./examples/model_compression/distill_lstm/) | 基于[Distilling Task-Specific Knowledge from BERT into Simple Neural Networks](https://arxiv.org/abs/1903.12136)论文策略的实现,将BERT中英文分类的下游模型知识通过蒸馏的方式迁移至LSTM的小模型结构中,取得比LSTM单独训练更好的效果。 |
281+
| [PP-MiniLM](examples/model_compression/pp-minilm) :star2: | 基于 PaddleSlim 通过模型蒸馏、剪裁、量化等级联模型压缩技术发布中文特色小模型 PP-MiniLM(6L768H) 及压缩方案,保证模型精度的同时模型推理速度达 BERT-base 的4.2倍,参数量相比减少52%,模型精度在中文语言理解评测基准 CLUE 高0.32。 |
281282

282283
#### 小样本学习 (Few-Shot Learning)
283284

File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.
File renamed without changes.

0 commit comments

Comments
 (0)