artventuredev
commented
11 months ago Hi, my apologies for the delayed response, I was quite busy latetly. As I understand, you should have only 1 LoRA activated, but when queuing multiple tasks with different LoRAs, the results seem to have multiple LoRAs applied.
Since it's already a while ago, are you still experiencing this issue? Can you confirm this is not happening when run the tasks directly?
Greetings :)
Whilst scheduling a stack of jobs, interchanging Checkpoints (using the dropdown) & Loras (inserted into the prompt), the generations become scatterred as they do when too many full strength Loras are piled onto the one generation.
The metadata of the images shows only one Lora active & i can replicate the later generations with only that one Lora in the prompt box, until i reload the webUI. This seemingly 'wipes' the memory of other Loras & correctly uses just the one in the prompt, making the generation data false creating totally different images :-/
The Loras seem to be stacking somewhere in memory with each new job & i fail to replicate it even using multiple Loras that i can see are being applied on the single generation, its just not the same.
I am possibly using the extension incorrectly & personally causing this but i did not find any information to reflect that & as such I will attach the line of generations i replicated the bug with, all are labled with their job# + Model-Lora & have all their metadata
ps: 1 & 4 are identical as i copied what was scheduled verbatim :)
pps: I was enamoured by job 11 & spent hours to ensure my replication of these "Gilded 'Apopunkalyptical' Angels"! Grateful I am able to do what i do:) <3