search

alembics / disco-diffusion

Other
7.48k stars 1.13k forks source link

Runtime error #83

Closed Rakeson closed 2 years ago

Rakeson commented 2 years ago

in 4 diffuse! SX1BIDKQMT19KST D(FYU%V AU6$C1DP}G}O2 3N28D)~40 N}N8RU6E_J6ZTUYLB%N 2G8

Rakeson commented 2 years ago

I have stuck in here for few hours, why things were OK in yesterday? (ran it for 4 to 5 hours)

Rakeson commented 2 years ago

D})M}M38_R296413XNA}8DU after multiple tests, this yellow notification will also shown up at the same time

aletts commented 2 years ago

@Rakeson People who use negative weights in their text prompts encounter that error when their weights sum to 0. Apologies for that -- the code is old, silly code which was probably written without consideration for negative weights.

When you know the problem, it's easy to work around it. For any prompts where you don't specify a weight, its weight is 1. If you've got just two prompts where one of them has a weight of -1 , then you can give the other one a weight of 2 and it'll work.

Rakeson commented 2 years ago

But wait it was fine when I typed "bla bla bla","human:-2" but when it was "bla bla bla","human:-2","no any words" and "bla bla bla","no any words","human:-2",these two kinds will show the Runtime error. Why the first time (above) was fine but its weight is -2?

Rakeson commented 2 years ago

Which this means? I should gather these elements instead of spreading them sperate?

aletts commented 2 years ago

@Rakeson As long as the prompt weights don't sum to 0, it'll run. I've seen people suggest turning the weight to -2.. and the effect of that is usually that they'll have it sum to -1 (if there were two prompts and the other is the default 1). I prefer to increase a positive weight in order to have them sum to something positive (since it seems less ridiculous), but it probably doesn't matter either way.

Rakeson commented 2 years ago

OK, I will keep trying

Rakeson commented 2 years ago

Thanks for your helping man , problem solved completely