Sailor Joe
Sailor Joe
请问有对这些问题的回答记录吗
I mean the model output, not just the calculated score
When using Acc in dureader, gpt-3.5-16k's score is very low, this is my self test.  This is inconsistent with the results in the appendix of your paper
original lama repo: https://github.com/advimman/lama.Use big-lama ckpt. Original LaMa works good some way. For example, as depth controlnet's input. The LaMa results in your repo are bad and almost like blurring...
> It doesn't, unless you change the inputs like image/mask. Yes, when change input image, it requests to load new model. But in normal i2i or inpaint workflow, changing inputs...
Maybe write a custom load node to load patch() and lora() as model and add it to base model to become a inpaint model