stabilityai
/

stable-diffusion-xl-base-0.9

StableDiffusionXLPipeline

stable-diffusion

Model card Files Files and versions

Wrong num_hidden_layers of text encoders?

#27

by p1atdev - opened Jul 10, 2023

•

edited Jul 10, 2023

This model's text encoders use the last layer:

https://huggingface.co/stabilityai/stable-diffusion-xl-base-0.9/blob/main/text_encoder/config.json#L19
https://huggingface.co/stabilityai/stable-diffusion-xl-base-0.9/blob/main/text_encoder_2/config.json#L19

But stability's config seems to use the penultimate layer:

https://github.com/Stability-AI/generative-models/blob/5c10deee76adad0032b412294130090932317a87/configs/inference/sd_xl_base.yaml#L49
https://github.com/Stability-AI/generative-models/blob/5c10deee76adad0032b412294130090932317a87/configs/inference/sd_xl_base.yaml#L58

Is this a mistake? Or was it intended?

For diffusers, selecting which layer to use is done when generating, not when loading.

https://github.com/huggingface/diffusers/blob/e3d71ad89abfee3817340b2245a49eec894a1705/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L342

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment