about qwen-image-lora train

[LGogh]

How much data and how many training steps are needed for qwen-image-edit lora?

[Artiprocher]

@LGogh There is no fixed answer to this question, as it depends on various factors such as image resolution and image diversity. Therefore, we typically retain multiple checkpoints during training and determine the optimal stopping point through subsequent evaluation.

[LGogh]

@Artiprocher The training I conducted yielded no results at all. I have over 300 pictures of the same style and followed the official training script and parameters.

[fcchit]

@LGogh Have you made any progress? I've encountered a similar problem.