You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello developers! Thank you for your contribution! I'm having problems with a github library provided by google (this library is mainly based on huggingface diffusers ): some variable names may be changed in newer versions of diffusers. Therefore, you may encounter some problems with variable names when using third-party libraries, as shown in the following figure.
A more detailed example would look like this:
However, there is no information about variable name changes in the code, for example in the source code the variable names are still encoder_hidden_states, attention_mask: (the code is in diffusers/src/diffusers/models/attention_processor.py /line 206)
I don't understand what the reason for such a problem is, it's a simple variable name problem, I guess it could be caused by the model loaded in DiffusionPipeline.from_pretrained(model_id).to(device) and the model in github not being exactly the same? Or if you can point me to where I can find the correct variable name, I'd appreciate it!
The above is my question, I would be grateful if I could get your answer!
The text was updated successfully, but these errors were encountered:
Yes for inner / non-public classes such as self.processor we want to keep the right to still do some changes to improve the code. I think for this case, we simply need to update google's prompt-to-prompt code.
A better / more long-term solution would be to add a prompt-to-prompt pipeline directly to diffusers to make sure things are properly maintained.
Hello developers! Thank you for your contribution! I'm having problems with a github library provided by google (this library is mainly based on huggingface diffusers ): some variable names may be changed in newer versions of diffusers. Therefore, you may encounter some problems with variable names when using third-party libraries, as shown in the following figure.
A more detailed example would look like this:

However, there is no information about variable name changes in the code, for example in the source code the variable names are still encoder_hidden_states, attention_mask: (the code is in diffusers/src/diffusers/models/attention_processor.py /line 206)

I don't understand what the reason for such a problem is, it's a simple variable name problem, I guess it could be caused by the model loaded in DiffusionPipeline.from_pretrained(model_id).to(device) and the model in github not being exactly the same? Or if you can point me to where I can find the correct variable name, I'd appreciate it!
The above is my question, I would be grateful if I could get your answer!
The text was updated successfully, but these errors were encountered: