I'm using lora + quantops + int8-tensorwise chroma. There is barely any effect on the image. The same exact model + the int8-fast node has much more expected impact on the outputs. Here even setting 2.0 strength is hardly working. I compared using the same sampling, seed and the outputs are much closer to the model without any lora applied.
I'm using lora + quantops + int8-tensorwise chroma. There is barely any effect on the image. The same exact model + the int8-fast node has much more expected impact on the outputs. Here even setting 2.0 strength is hardly working. I compared using the same sampling, seed and the outputs are much closer to the model without any lora applied.