Closed robitronathon closed 4 months ago
I've figured out the main issue. The problem is that I use LoRAs in my wildcards. This definitely explains the issue a lot better, But is pretty unavoidable for how i use some wildcards. is there a good way to deal with this? ideally have it cache many LoRAs at once to be able to toggle them on and off instantly, but I would also settle for having a way to schedule prompts, like having 100 batches with these LoRAs, then 100 with some different LoRAs etc all set up automatically for over night generations.
going to settings and changing "Number of Lora networks to keep cached in memory" to 99 or something seems to work, problem solved.
Glad you found a solution! 👍
I've been using wildcards for a while, and for the past few months there has been a long pause between each generation. After disabling the extension the issue goes away, so it is definitely sd-dynamic-prompt's fault. Is there some setting to fix this? it seems like a ridiculously long time for a python script to pull some numbers then write a string based on that. I'll note that i do use compound wildcards, as in wildcards that call other wildcards, but even when i am only using wildcards that don't do that, I still have the 20 second or so pause every single batch. I really like the utility of this extension, but functionally doubling my generation times is really annoying.