search

huggingface / diffusers

🤗 Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch and FLAX.
https://huggingface.co/docs/diffusers
Apache License 2.0
26.1k stars 5.38k forks source link

low attention to prompt in SDXL #5816

Closed saeedkhanehgir closed 10 months ago

saeedkhanehgir commented 11 months ago

Hi, One of the difference between DALLE3 and SDXL is that SDXL pay less attention to prompt, Is there a way to solve this problem? I don't Know. for example changing the text encoder to other can help to solve this problem ? Thanks

yiyixuxu commented 11 months ago

You can't change to a different text encoder that the model wasn't trained on at inference time

I think SDXL is still the best we have in term of prompt following. cc @sayakpaul

sayakpaul commented 11 months ago

You can do prompt-weighting: https://huggingface.co/docs/diffusers/main/en/using-diffusers/weighted_prompts.

Also, for questions, please post them on our Discord forum. Issues are supposed to be about bugs, feature requests, API design RFCs, etc.

saeedkhanehgir commented 11 months ago

thanks a bunch @yiyixuxu @sayakpaul

I have another question. I have to use Compel to solve long sequence length problem . I want to know that can Compel decrease attention to prompt (instead of normal use)?

github-actions[bot] commented 10 months ago

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.