Open fritol opened 6 months ago
Hi, thank you for your interest in our work.
For the inference speed, LoRA Composite does require more processing time. Typically, the time increase should be linear, so if you're combining k LoRAs, you'd expect it to take about k times longer. The 30x slower performance you're experiencing seems unusual. Could you share more about your experiments? Additionally, LoRA Switch should have comparable inference speed to LoRA Merge.
when i run SDXL it takes few secs per image with this pipeline it takes OVER a MINUTE per image??!