Questions about the diversity of the results generated
I've found that with this model, giving the same prompt, with different seeds, produces very similar results. Has anyone else noticed this? What is causing it?
y
When you run Playground 2.5 locally, without webui, you have the same options or tools as the online command line tool?. Is there a webui that is clone or identical to that of the website?thx for your answers.
how could i run playground 2.5 locally with a webserver similar to the https://playground.com/canvas ??
I am using diffusers — exactly as per the code in the model card
能否支持图生图
I think it is most likely caused by overtraining, and there may be no way to alleviate this. However, if it is simply used, perhaps we can consider using LLM to automatically generate prompts, thereby achieving a certain degree of variation.