Fine-tune textdiffuser-2 inpainting with customized dataset

[shihuai]

Thank you for open sourcing this project, we are also very interested in your project. Now, we want to fine-tune the model with customized dataset, but there are some problem when train the model:

1. Did you fine-tune textdiffuser-2 inpainting with multilingual dataset ?
2. In the paper, How many samples when you train textdiffuser-2 inpainting ?
3. How many samples we should prepare when fine-tune with multilingual dataset ?