PuLID

PuLID copied to clipboard

Some details about the implementations

9

Hi! Thanks for your extraordinary work! I have a question about the training. The paper mentioned that "We introduce a Lightning T2I branch alongside the regular diffusion branch." But in...

jy0205

如题

aytwo553

Training Script快放出来呀

3

Training Training Training Training Training Training Training Training Training Training Training Training Training

digdigdigging

Transfer the method to SD3, I encountered some confusion on train stage2

1

I am also trying to train the model, but I have some confusion regarding the loss design in the second stage. According to the paper, the lighting branch generates a...

LeonNerd

train on flux, how to get rgb image on stage 2?

2

Nice work! When train on Flux, I would ask which lightweight network should be used to obtain RGB images when training a 2-stage network? Thanks.

Shallo2512

is it human faces only?

yosun

load lora

1

I want to use this to load lora, is there any way to load flux lora in this component?

APannnnnnnnd

Pulid not working - xFormers issue

1

Here is the error message I get ApplyPulid No operator found for `memory_efficient_attention_forward` with inputs: query : shape=(1, 577, 16, 64) (torch.float16) key : shape=(1, 577, 16, 64) (torch.float16) value...

bipinpeter

F

Vadyushkaaa

代码是不是有bug，刚开始能跑，跑图几轮之后，内存不足。重新clone代码又能跑，跑几轮又不行。重新clone又能跑，真离谱啊。

1615070057