After call IP - Adapter unload method StableDiffusionImg2ImgPipeline(**pipe.components) error

[Honey-666]

hi!
When I used 1.5 to load the model, I enabled the security check mechanism, uninstalled after using the IP-Adapter, and generated the pipe from the previous pipe, which would throw an error with the security check parameters

diffusers/src/diffusers/loaders/ip_adapter.py

Line 209 in adcbe67

if hasattr(self, "feature_extractor") and getattr(self, "feature_extractor", None) is not None:

diffusers/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

Line 250 in adcbe67

if safety_checker is not None and feature_extractor is None:

os:
diffusers==0.26.0

this is my code:

import torch
from PIL import Image
from diffusers import StableDiffusionPipeline, StableDiffusionImg2ImgPipeline

model_path = 'C:\\work\\pythonProject\\aidazuo\\models\\Stable-diffusion\\stable-diffusion-v1-5'
ip_adapter_path = 'C:\\work\\pythonProject\\aidazuo\\models\\IP-Adapter'
ip_img_path = 'C:\\work\\pythonProject\\aidazuo\\jupyter-script\\test-img\\vermeer.png'

pipe = StableDiffusionPipeline.from_pretrained(
 model_path,
 variant="fp16",
 torch_dtype=torch.float16).to("cuda")

pipe.load_ip_adapter(ip_adapter_path, subfolder="models", weight_name="ip-adapter-plus_sd15.bin")
pipe.set_ip_adapter_scale(0.6)

ip_adapter_img = Image.open(ip_img_path)

images = pipe(
 prompt='a beautiful girl',
 ip_adapter_image=ip_adapter_img,
 negative_prompt="",
 num_inference_steps=30,
 num_images_per_prompt=1,
 width=512,
 height=512).images

pipe.unload_ip_adapter()

imagePipe = StableDiffusionImg2ImgPipeline(**pipe.components).to("cuda")

images2 = imagePipe(num_inference_steps=30,
 image=images[0],
 strength=0.5,
 prompt='a beautiful girl').images

for img in images2:
 print(img)
 img.show()

[yiyixuxu]

hey, thanks! I can reproduce this and will fix it shortly:)

[yiyixuxu]

should be fixed in this PR #6868