HONGYU OUYANG

Results 9 issues of HONGYU OUYANG

### PR types New features ### PR changes Models ### Description add more outputs & loss to CodeGen model

### PR types New features ### PR changes Models ### Description add more output and loss for both bart and mbart models

### PR types New features ### PR changes Models ### Description add more output and loss for unimo model

### PR types New features ### PR changes Models ### Description add more output & loss to unified transformer model

### PR types New features ### PR changes Models ### Description Add more outputs and loss to GPT model.

### 软件环境 ```Markdown - paddlepaddle-gpu: 2.3.2.post116 - paddlenlp: 2.4.0 ``` ### 重复问题 - [X] I have searched the existing issues ### 错误描述 ```Markdown 当使用 `GPTModel.from_pretrained("gpt2-small-en")`时无法创建模型,报错: FileNotFoundError: [Errno 2] No such...

bug

### PR types New features ### PR changes Models ### Description 1. Add embedding inputs to T5 model 2. Add `test_inputs_embeds` to `test_modeling_common.py`

在 weibo_comment/spider.py 下,如果获取json数据失败则会把 max_id_type 设置为 1 继续爬取。我验证后发现,max_id_type更改为 1 后爬取到的是按照时间降序的评论内容。

### 需求 能否支持会议(CCF 会议)查询? ### 版本信息 zotero version: 6.0.27 zotero-updateifsE version: 0.13.0 ### 已有材料 * ccf推荐国际学术刊物目录:https://www.ccf.org.cn/Academic_Evaluation/By_category/ * easyScholar里已有公开的CCF推荐会议数据集,采用模糊匹配方法![image](https://github.com/redleafnew/zotero-updateifsE/assets/40912707/32ff5564-bcb8-44af-8f4a-4a4f0c12a91b) 经测试: * easyScholar里勾选上述自定义数据集后,zotero内期刊文献可以被正确匹配并标注,而会议全部失败。 * 更正:[部分会议论文](https://scholar.google.com/scholar?hl=zh-CN&as_sdt=0%2C5&q=BLIP%3A+Bootstrapping+Language-Image+Pre-training+for+Unified+Vision-Language+Understanding+and+Generation&btnG=)从谷歌学术搜索上经过浏览器扩展保存到zotero里后,可以被green frog插件标注,而[部分会议论文](https://scholar.google.com/scholar?hl=zh-CN&as_sdt=0%2C5&q=VLMo%3A+Unified+Vision-Language+Pre-Training+with+Mixture-of-Modality-Experts&btnG=8)在谷歌学术里可以被easyScholar插件标注,却不能在zotero里被green frog标注 * 上述会议论文在 semantic scholar 搜索引擎里不能被easyScholar插件标注,而在谷歌学术搜索上可以被标注