Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

请教一下预训练权重文件 #5

Open
yang-chenyu104 opened this issue Dec 27, 2023 · 19 comments
Open

请教一下预训练权重文件 #5

yang-chenyu104 opened this issue Dec 27, 2023 · 19 comments

Comments

@yang-chenyu104
Copy link

在下载的预训练的pretrained文件夹为空,上面解析为stable diffusion模型,我需要去hugging face下载相应模型权重文件是吧,可以告诉一下是stable diffusion哪方面预训练文件,有相关链接吗

@ZYM-PKU
Copy link
Owner

ZYM-PKU commented Dec 27, 2023

你好,这里我们使用的是stable diffusion v2.0 的 inpainting 版本。如果需要训练,请把相应的512-inpainting-ema.ckpt放在pretrained文件夹内。如果需要推理,则不需要考虑pretrained文件夹。

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 27, 2023 via email

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 27, 2023 via email

@ZYM-PKU
Copy link
Owner

ZYM-PKU commented Dec 27, 2023

看见了可能是我下载出问题,不好意思,谢谢大佬解答,我想用扩散模型做一个可控文本区域编辑,可以用文本去生成任意的ocr真实数据,在这个基础上加入chinese clip,但不知道加入clip文本引导可以达到什么效果,因为中文需要的数据量更大

---原始邮件--- 发件人: @.> 发送时间: 2023年12月27日(周三) 上午10:07 收件人: @.>; 抄送: @.@.>; 主题: Re: [ZYM-PKU/UDiffText] 请教一下预训练权重文件 (Issue #5) 你好,这里我们使用的是stable diffusion v2.0 的 inpainting 版本。如果需要训练,请把相应的512-inpainting-ema.ckpt放在pretrained文件夹内。如果需要推理,则不需要考虑pretrained文件夹。 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

是的,中文的文字结构更加复杂,所以理论上比英文难做。我认为这篇工作的核心是用一种字符级别编码的理念来给模型施加更精细的条件控制。对应到中文上,也需要考虑编码器的改良设计,当然这也是我现在正在探索的方向。

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 27, 2023 via email

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 27, 2023 via email

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 28, 2023 via email

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 28, 2023 via email

@ZYM-PKU
Copy link
Owner

ZYM-PKU commented Dec 28, 2023

readme已更新,加入了数据集文件结构

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 28, 2023 via email

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 28, 2023 via email

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 28, 2023 via email

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 28, 2023 via email

@ZYM-PKU
Copy link
Owner

ZYM-PKU commented Dec 28, 2023

  1. 首先进行pretrain再进行正式train,根据./configs/pretrain.yaml 配置文件的设置,默认是5个epoch存储一次encoder模型权重。
  2. ocr_loss 为 1应该是出现了某种错误,但由于没有具体信息我没法判断原因。

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 28, 2023 via email

@ZYM-PKU
Copy link
Owner

ZYM-PKU commented Dec 29, 2023

我的邮箱是[email protected],可以直接通过邮件和我联系。

@yang-chenyu104
Copy link
Author

yang-chenyu104 commented Dec 29, 2023 via email

@lijain
Copy link

lijain commented May 31, 2024

能问下现在这个基准生成中文的字效果如何?我看你的在线demo关闭了

@ZYM-PKU
Copy link
Owner

ZYM-PKU commented May 31, 2024

这个方法没有做中文的实验,无法生成中文字

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants