meet problems when text2img sampling

[order-a-lemonade]

I am using this pretrained model for text2img sampling:

and i get a result like this:

when i use this pretrained model:

the result is normal, like this:

is there any suggestion?