1. 新增人脸检测RetinaFace模型；
2. 完成Maas-cv CR标准自查
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9945188

商品显著性检测模型，依赖opencv，mmcv-full
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9909897

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9934673

    * add face 2d keypoints

![](https://cn-hangzhou.oss-cdn.aliyun-inc.com/git/force/uploads/comment/251924/40165669611078357/image.png)
fill mask pipeline 测试时间过长

这个task测了4个模型。从保证代码正确性的功能角度看，只测一个bert类(比如sbert中文），一个roberta类（veco)。减少测试的模型数量以减少测试时长。
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/10006556

    * skip sbert_en&bert_ch to save ci time

文本指导的语义分割模型，根据输入的文本信息，讲图像中对应文本描述的物体分割出来。
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9942863

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9970626

    * fix trainer unittest

接入tinynas-detection，新增tinynas object detection pipeline以及tinynas models。
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9938220

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9974293

add Deberta v2 modeling and fill_mask task, with master merged
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9966511

添加新的action-detection task
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9898947

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9938140

基于easycv上线的segformer，对应上传了5个对应的达摩院的分割模型，所以修正了tests/pipelines/easycv_pipelines/test_segmentation_pipeline.py内容让其能够便利测试
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9934634

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9969472

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9901220

    * add lstm-crf ner model code

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9872869

Add regression test for some unit tests.
Firstly, Run a baseline test to create a pickle file which contains the inputs and outputs of modules, then changes can be observed between
the latest version and the baseline file.
Some baseline files are submitted in the data/test/regression folder
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9814693

* nlp translation preprocess branch
 * pull the latest master
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9920445

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9942858

    * [Add] add document-segmentation

* dataset json file add args

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9941890

1. add child-tuning optimizer and ut
2. fix a training bug which can cause interruption after cross-evaluation
3. move model.params from cfg to default args in build_optimizer to prevent the saving of params in save_pretrained
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9891963

支持 MPLUG 模型 image-text-retrieval 任务的 pipeline 和 finetune
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9919955

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9892528

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9908976

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9862567

新增实时目标检测pipeline
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9788299

* 修复了zip文件不同打包模式下返回路径错误问题。
* 修复了替换了数据集文件重新下载时校验失败问题。
* 修复dataset oss文件在 REUSE 模式下重复下载的问题。
* 修复了csv数据集的meta json文件中某个split的meta和file字段都为''时加载所有split失败的问题。
 * 修复了不同版本datasets路径不一致的问题。

添加 gpt3 中小模型单机单卡下的 finetune 代码
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9899004

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9781849

    * support EasyCV

Maas新增FAQ问答模型
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9797053

1. MsDataset支持upload数据文件(压缩包)
2. MsDataset支持clone和upload meta data
3. 使用MsDataset.load()下载数据集，支持web端显示数据集下载计数
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9831232

video_summarization 修改test中的结果可视化
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9902499

添加 mplug 模型 caption 及 vqa 任务的 finetuning 支持
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9858028

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9851374

skip this test which will result in too much debug log for placement although debug level is canceled after this test case

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9875987

panoptic segmentation 模型接入
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9758389

新增video summarization模型的inference和finetune
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9840532

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9855254

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9856286

This reverts commit 4cc0395ac5.

add image-reid-person
        Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9818427

1. add device util to verify, create and place device
2. pipeline and trainer support update
3.  fix pipeline which use tf models does not place model to the right device

usage

```python
pipe = pipeline('damo/xxx', device='cpu')
pipe = pipeline('damo/xxx', device='gpu')
pipe = pipeline('damo/xxx', device='gpu:0')
pipe = pipeline('damo/xxx', device='gpu:2')
pipe = pipeline('damo/xxx', device='cuda')
pipe = pipeline('damo/xxx', device='cuda:1')
```
 Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9800672

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9826159

Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9781254

    * support single object tracking