Noise Result

Question

Noise Result

simonmecho opened this issue 5 months ago · 12 comments

Hello, I'm trying the workflow using fp8 and also the gguf version, but in both case I just got result of noise only, what was the case for that? please see the attached screenshot. Thanks!

Answer 1 · 2024-09-03T11:15:16.000Z

It's a bit strange indeed. Is there any error message in the terminal?
or try .bin format model instead.

Answer 2 · 2024-09-03T12:27:09.000Z

Actually I've already tried the .bin model, and in both cases there is no errors on the console, which is quite weird...

Answer 3 · 2024-09-03T12:42:17.000Z

这是什么情况？？？？怎么解决

Answer 4 · 2024-09-03T17:05:53.000Z

uugf model or the normal one?

Answer 5 · 2024-09-04T13:11:34.000Z

And the generation take forever is really slow

Answer 6 · 2024-09-04T13:39:03.000Z

try using non-uugf format model with hyper-flux lora.

Answer 7 · 2024-09-04T13:46:13.000Z

Answer 8 · 2024-09-04T14:19:24.000Z

It looks quite ok, have you ever tried our example workflow?

Answer 9 · 2024-09-04T14:20:39.000Z

Yes same result

Answer 10 · 2024-09-04T15:10:24.000Z

You guys using Lora with clip port or only use load only Lora weight?

Answer 11 · 2024-09-04T15:13:57.000Z

I was using model only load lora, I bypass the lora node and the result is the same, the ksampler take a lot of time to render.

Loader 445.20s
Ksampler 1943.20s

Answer 12 · 2024-09-04T15:37:03.000Z

What kind of gpu are you using? And what is your PyTorch version? Rafael Nolasco ***@***.***>於2024年9月4日 23:14寫道： I was using model only load lora, I bypass the lora node and the result is the same, the ksampler take a lot of time to render. Loader 445.20s Ksampler 1943.20s —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: ***@***.***>