search

Zhendong-Wang / Prompt-Diffusion

Official PyTorch implementation of the paper "In-Context Learning Unlocked for Diffusion Models"
Apache License 2.0
371 stars 9 forks source link

Training a new task #13

Open nlylmz opened 6 months ago

nlylmz commented 6 months ago

Hi, thank you for your work. It is an excellent job! I would like to learn how to train a new task with different images on the Prompt Diffusion. My task has the same structure as yours; Image 1, Image 2 and Image 3. However, the task and images are different and I am not sure if I need to do something extra to train the images on the task. Because you are using annotators from ControlNet but, for my task, it is not available. Could you please provide some steps what I should do for training my task with different images?

Zhendong-Wang commented 6 months ago

If you already have (image1, image2) forming the example pair, and image3 as the query image, then you can directly start the training with our pipeline. You need to modify the dataset file https://github.com/Zhendong-Wang/Prompt-Diffusion/blob/main/edit_dataset.py to be consisent with your dataset.

nlylmz commented 6 months ago

Thanks a lot for your reply.