Liang Hu

Results 10 issues of Liang Hu

我的ID: 18810577484 只想获取美股中上市的中概股的股票列表, 请问是否可以开发一个新接口, 或者在既有接口上增加一个参数, 判断我请求的是否是中概股列表

作者您好: 我看了一下chinese dependency parsing在paperswithcode上的SOTA模型 显示是这一篇Glyce: Glyph-vectors for Chinese Character Representations 它的dependency parsing的分数89 L , 90.2 U, 而本项目readme中的文档中写的dependency parsing的SOTA的准确率没有这个高, 请问readme中是参考的哪篇的数据?

![image](https://user-images.githubusercontent.com/13363151/145958286-7e5b51bb-b2f6-4b1f-939e-9b70d39e7cc3.png) ![image](https://user-images.githubusercontent.com/13363151/145958329-e318847c-27ce-49f8-b903-721bd9920996.png) 作者您好,执行pos finetune的时候出错,应该是数据集处理有点问题,我看pos和parsing都用同样的打标格式, 所以就task参数改了值,其它没改

尊敬的作者: 您好。我发现一个很有提升价值的点。 如果我直接做parsing的finetune, 然后在测试集上看效果, 发现分词的效果很不好, 照理来说不应该这样, 所以我猜测Parsing任务没有包含cws和pos的finetune, 但是照理来说parsing的标签是包含了cws和pos的信息的. 我先执行cws的finetune再执行parsing的finetune发现test集合上的分词效果很好. 所以我理解在执行更高级的任务的时候, 可以先从最低级的任务开始finetune, 而且是自动完成这个过程. 即如果finetune Parsing, 则先finetune cws, 然后finetune pos, 最后finetune parsing 如果finetune pos, 则先finetune cws, 再finetune pos 作者您怎么看?

![image](https://user-images.githubusercontent.com/13363151/145950535-5f510011-182b-4ed9-91bb-5437d9a2ee97.png) 返回,不是反悔

尊敬的作者: 您好, 请教下模型保存后的路径下的chars_vocab文件和label_vocab文件打开都是乱码, 请问如何解决? Windows10, 用记事本打开, 编码是ANSI ![image](https://user-images.githubusercontent.com/13363151/144989627-18dd924f-4d54-4188-bb7c-647a2e3732cc.png) ![image](https://user-images.githubusercontent.com/13363151/144989651-0c55a226-2b5b-4c1e-8cf7-e53ebae45622.png)

### Required prerequisites - [x] I have read the documentation . - [x] I have searched the [Issue Tracker](https://github.com/PKU-Alignment/align-anything/issues) and [Discussions](https://github.com/PKU-Alignment/align-anything/discussions) that this hasn't already been reported. (+1 or comment...

bug

I implement the finetune code myself according to the paper, but when i sft the janus-pro 7b or 1b, The loss started at around 5, dropped to approximately 4.7, and...

![image](https://github.com/user-attachments/assets/823d43fa-3d1a-475e-8bf6-ab98c29b12d5) just download the repo, and follow the readme.md ![image](https://github.com/user-attachments/assets/402550f9-3ecc-46fd-9c12-c0b1238dd1b1) why ? something i missed?

![Image](https://github.com/user-attachments/assets/e7fe07bd-34a0-4188-8aa9-50c331b0f9e8) ![Image](https://github.com/user-attachments/assets/134d40a0-02c3-4789-b4ac-aedb3da8d200) when load flux kontext using huggingface pipeline, the vae is loaded in image 1 but in official flux/src/flux/cli_kontext.py, the vae is image 2 whats the difference ??? thanks...