Training on small datasets

[j0seo]

Thank you for releasing this amazing work!

I am curious about how well it performs on relatively small datasets. Could you share your thoughts on whether we can expect good generalization performance when training on relatively small datasets (<30K)?

[xduzhangjiayu]

@j0seo
Here, https://huggingface.co/docs/diffusers/v0.20.0/en/api/pipelines/controlnet_sdxl
It says: "The ControlNet learns task-specific conditions in an end-to-end way, and the learning is robust even when the training dataset is small (< 50k)"
What do you think? Have you tried with small and large datasets and compare the results?