Error loading model

[RayTang88]

Thank you for your wonderful work .

I unzipped all the models from Baidu Cloud and got an error when loading Stage1/model_15000.pth. The error is as follows:

Here is the md5 value of the downloaded and unzipped model:

Thanks again.

[ningshuliang]

Sorry for this issue, I will update the checkpoint today~

[ningshuliang]

This is the New link~ https://cuhko365-my.sharepoint.com/personal/115010206_link_cuhk_edu_cn/_layouts/15/onedrive.aspx?id=%2Fpersonal%2F115010206%5Flink%5Fcuhk%5Fedu%5Fcn%2FDocuments%2Fpretrain%5Fmodels%2Ezip&parent=%2Fpersonal%2F115010206%5Flink%5Fcuhk%5Fedu%5Fcn%2FDocuments&ga=1

[RayTang88]

Thanks for your reply. When I tried to click on the link you sent, I logged into my Microsoft account with my email, and the following problem occurred. It said that I did not have sufficient permissions? Could you please upload it to Baidu Cloud again? In addition, has the clip of OpenAI been fine-tuned separately? Is it consistent with the one on HuggingFace? If it is consistent, maybe you don’t need to upload it to save your storage space.

[ningshuliang]

I have upload to baiduyun again, you can check now. [cid:8256282b-1ea4-4025-a68a-f99f36b8f8f6]

---

发件人: RayTang88 @.> 发送时间: 2024年8月13日 16:03 收件人: ningshuliang/PICTURE @.> 抄送: Shuliang Ning (SSE, 115010206) @.>; Comment @.> 主题: Re: [ningshuliang/PICTURE] Error loading model (Issue #9)

Thanks for your reply. When I tried to click on the link you sent, I logged into my Microsoft account with my email, and the following problem occurred. It said that I did not have sufficient permissions? Could you please upload it to Baidu Cloud again? image.png (view on web)https://github.com/user-attachments/assets/c918e816-d6c2-4c91-a1ef-b724490aae68 In addition, has the clip of OpenAI been fine-tuned separately? Is it consistent with the one on HuggingFace? If it is consistent, maybe you don’t need to upload it to save your storage space.

― Reply to this email directly, view it on GitHubhttps://github.com/ningshuliang/PICTURE/issues/9#issuecomment-2285612939, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AIONUHDQSNFCUBKMZPWD26DZRG4W7AVCNFSM6AAAAABMF6PZVSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEOBVGYYTEOJTHE. You are receiving this because you commented.Message ID: @.***>

[RayTang88]

Ok，Thank you!

[ningshuliang]

why reopen?

[RayTang88]

Sorry, I still have some problems after changing the model.

I tested the sha265 of the model and it is the same as the value on huggingface.

[ningshuliang]

I think you need to change the version of clip model? I know somebody has successed in load my pretrain model. Please check again.

[RayTang88]

Ok, I'm checking it！

[RayTang88]

Hi， I modified the code in Stage1_Text_to_Parsing[/test.py] line:267 After modification, it is as follows: model.load_state_dict(torch.load('./../pretrain_models/Stage1/model_15000.pth'), strict=False)， It works fine and gives the same result as you provided. So I will close this issue now, thank you very much.

[ningshuliang]

hh, it seems that you need to check the path .

[RayTang88]

I adjusted my last comment. I think it might be because I added “strict=False“ that it worked.Is this the correct modification?