You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
hi,I encountered a problem when using sdxl and controlnet together. The weight of sdxl will be replaced by the weight combined with lora.follow is my code:
import sdkit
from sdkit.generate import generate_images
from sdkit.models import load_model
from sdkit.utils import log
from PIL import Image
context = sdkit.Context()
convert an existing image into an openpose image (or skip these lines if you have a custom openpose image)
from sdkit.filter import apply_filters
input_image = Image.open("../2.png") # set your input image here <---------
canny_image = apply_filters(context, "canny", input_image) # you can apply other filters too, e.g. "canny" etc.
print(canny_image)
load SD
context.model_paths["stable-diffusion"] = "../sdxl10/stable-diffusion-xl-base-1.0/sd_xl_base_1.0_0.9vae.safetensors" # <---- SD model path here
load_model(context, "stable-diffusion")
load controlnet
context.model_paths["controlnet"] = "../sdxl10/controlnet-canny-sdxl-1.0-mid/diffusion_pytorch_model.fp16.safetensors" # <----- Controlnet model path
load_model(context, "controlnet")
generate the image
image = generate_images(
context,
"1boy, muscular, full armor, armor, shoulder plates, angry, looking at viewer, spiked hair, white hair",
seed=42,
width=1024,
height=1024,
control_image=canny_image,
)[0]
save the image
image.save("controlnet_image.jpg")
log.info("Generated images with a Controlnet!")
error code:
10:30:12.705 INFO MainThread Applying canny...
Traceback (most recent call last):
File "/media/alleniver/mydata/diffusion2/web_server/sdkit/examples/009-generate-controlnet.py", line 16, in
canny_image = apply_filters(context, "canny", input_image) # you can apply other filters too, e.g. "canny" etc.
File "/home/alleniver/anaconda3/envs/edtool/lib/python3.10/site-packages/sdkit/filter/apply_filters.py", line 39, in apply_filters
return [apply_filter_single_image(context, filters, image, **kwargs) for image in images]
File "/home/alleniver/anaconda3/envs/edtool/lib/python3.10/site-packages/sdkit/filter/apply_filters.py", line 39, in
return [apply_filter_single_image(context, filters, image, **kwargs) for image in images]
File "/home/alleniver/anaconda3/envs/edtool/lib/python3.10/site-packages/sdkit/filter/apply_filters.py", line 49, in apply_filter_single_image
image = get_filter_module(filter_type).apply(context, image, **kwargs)
File "/home/alleniver/anaconda3/envs/edtool/lib/python3.10/site-packages/sdkit/filter/apply_filters.py", line 59, in default_apply
model = context.models[filter_type]
KeyError: 'canny'
same error with 'depth' weights.
The text was updated successfully, but these errors were encountered:
hi,I encountered a problem when using sdxl and controlnet together. The weight of sdxl will be replaced by the weight combined with lora.follow is my code:
import sdkit
from sdkit.generate import generate_images
from sdkit.models import load_model
from sdkit.utils import log
from PIL import Image
context = sdkit.Context()
convert an existing image into an openpose image (or skip these lines if you have a custom openpose image)
from sdkit.filter import apply_filters
input_image = Image.open("../2.png") # set your input image here <---------
canny_image = apply_filters(context, "canny", input_image) # you can apply other filters too, e.g. "canny" etc.
print(canny_image)
load SD
context.model_paths["stable-diffusion"] = "../sdxl10/stable-diffusion-xl-base-1.0/sd_xl_base_1.0_0.9vae.safetensors" # <---- SD model path here
load_model(context, "stable-diffusion")
load controlnet
context.model_paths["controlnet"] = "../sdxl10/controlnet-canny-sdxl-1.0-mid/diffusion_pytorch_model.fp16.safetensors" # <----- Controlnet model path
load_model(context, "controlnet")
generate the image
image = generate_images(
context,
"1boy, muscular, full armor, armor, shoulder plates, angry, looking at viewer, spiked hair, white hair",
seed=42,
width=1024,
height=1024,
control_image=canny_image,
)[0]
save the image
image.save("controlnet_image.jpg")
log.info("Generated images with a Controlnet!")
error code:
10:30:12.705 INFO MainThread Applying canny...
Traceback (most recent call last):
File "/media/alleniver/mydata/diffusion2/web_server/sdkit/examples/009-generate-controlnet.py", line 16, in
canny_image = apply_filters(context, "canny", input_image) # you can apply other filters too, e.g. "canny" etc.
File "/home/alleniver/anaconda3/envs/edtool/lib/python3.10/site-packages/sdkit/filter/apply_filters.py", line 39, in apply_filters
return [apply_filter_single_image(context, filters, image, **kwargs) for image in images]
File "/home/alleniver/anaconda3/envs/edtool/lib/python3.10/site-packages/sdkit/filter/apply_filters.py", line 39, in
return [apply_filter_single_image(context, filters, image, **kwargs) for image in images]
File "/home/alleniver/anaconda3/envs/edtool/lib/python3.10/site-packages/sdkit/filter/apply_filters.py", line 49, in apply_filter_single_image
image = get_filter_module(filter_type).apply(context, image, **kwargs)
File "/home/alleniver/anaconda3/envs/edtool/lib/python3.10/site-packages/sdkit/filter/apply_filters.py", line 59, in default_apply
model = context.models[filter_type]
KeyError: 'canny'
same error with 'depth' weights.
The text was updated successfully, but these errors were encountered: