How to test trained stage 1 and stage 2 model?

[ElTocas]

I have trained model in specific lr images.

How can i infer the models .prh generated?

Thanks

[0x3f3f3f3fun]

I will add the code for self-trained model inference today.

[0x3f3f3f3fun]

Use this command for custom-trained model inference :)

Since DiffBIR needs to support multiple tasks, the inference code may appear somewhat complex. Feel free to ask if you have any questions about the code.

[ElTocas]

First of all, thanks for all the work you have done.

I did the process, and the result is a very black image. I just tested it for 1000 iter on 20 train images, but the result is all black.

Is there something that i missing?

[0x3f3f3f3fun]

Can you provide more detailed information?,
such as: (1) your training configuration file; (2) whether your training logs are normal; (3) your inference command.

[GraceZhuuu]

I've encountered the same problem, did you solve it ?
The training logs seem right, but inference result is all black.

i used the default training config, the training logs are normal;
checked the intermediate output, in the apple_cldm function, after 7. Decode generated latents, the output is nan.

[ruiqiyan]

First of all, thanks for all the work you have done.首先，感谢您所做的所有工作。

I did the process, and the result is a very black image. I just tested it for 1000 iter on 20 train images, but the result is all black.我做了这个过程，结果是一个非常黑的图像。我刚刚在20次火车图像上对其进行了1000次迭代的测试，但结果是黑色的。

Is there something that i missing?我想念的东西吗？

Have you found the reason?

[ruiqiyan]

I've encountered the same problem, did you solve it ?我遇到了同样的问题，您解决了吗？ The training logs seem right, but inference result is all black.训练日志似乎正确，但推断结果都是黑色的。

i used the default training config, the training logs are normal;我使用了默认的培训配置，训练日志正常； checked the intermediate output, in the apple_cldm function, after 7. Decode generated latents, the output is nan.检查了apple_cldm函数中的中间输出，在7。解码生成的潜在后，输出为NAN。

Have you found the reason?