sayakpaul / stable-diffusion-keras-ft Goto Github PK

I got the following error message when running the finetune.py script:

ValueError: Unknown layer: 'PaddedConv2D'. Please ensure you are using a `keras.utils.custom_object_scope` and that this object is included in the scope. See https://www.tensorflow.org/guide/keras/save_and_serialize#registering_the_custom_object for details.

The line that got this issue is this line in the code: link
Based on my understanding, the built-in copy function cannot copy keras-cv PaddedConv2D layer. After I modified that line to this:

self.ema_diffusion_model = keras.models.clone_model(self.diffusion_model)

The script can run now.
Here are my machine specs:

• macOS
• Python 3.8.10
• tensorflow 2.11.0
• keras-cv 0.4.0

I am trying to run the google colab notebook given in the following tutorial below .
https://keras.io/examples/generative/finetune_stable_diffusion/
However I am getting out of memory error in colab environment , any idea on how to fix this @sayakpaul @deep-diver

Following: https://github.com/huggingface/diffusers/pull/1868/files

Cc @deep-diver