Closed a-niess closed 2 months ago
The error is defined in ComfyUI_IPAdapter_plus/IPAdapterPlus.py
here in lines 180–181:
if is_faceid and not insightface:
raise Exception("insightface model is required for FaceID models")
I continue to have this same issue. Even after all the suggested fixes (#162) I am unsure how to proceed.
FaceID models require the FaceID IPadapter node. You are probably just using the wrong node
The exact same thing happend to me. Turns out I just needed to restart ComfyUI (I had installed everything while it was running).
FaceID models require the FaceID IPadapter node. You are probably just using the wrong node
I double checked and I am using that node. It seems the install of Insight is missing but it's installed, the appears to be correct and I used the onnxruntime to add to the wheel. as per the #162
do you get any error at startup
In my case, download InsightFace model will solve this problem.
I'm having exactly the same error. Any ideas?
I'm closing this as it looks like a configuration issue and not a bug
The errors I get in my RunPod environment are resolved by using the ComfyUI Manager Install PIP packages
button and completing the following installations:
pip install pillow==10.2.0 insightface onnxruntime onnxruntime-gpu
Afterward, I do not get any issues.
@philcoraz I'm glad it worked for you! There's a way to configure your RunPod so that these installations will persist in your python virtual environment, but doing the installs through the ComfyUI Manager each time is pretty straightforward.
@a-niess could you confirm which RunPod template you are using please? The install pip packages approach is giving me this error;
Failed to install 'pip install pillow==10.2.0 insightface onnxruntime onnxruntime-gpu' See terminal log.
@jgal14 I am using the ghcr.io/ai-dock/comfyui:latest-jupyter
template. The template used to be offered by default on RunPod but was removed some time around last week. If this is the template you are using, the closest equivalent is ghcr.io/ai-dock/comfyui:latest
which RunPod calls ComfyUI - AI-Dock
, so you can just edit the template by appending -jupyter
to the end of the title for the Container Image and then confirm with "Set Overrides." I am also using a network volume.
@jgal14 I am using the
ghcr.io/ai-dock/comfyui:latest-jupyter
template. The template used to be offered by default on RunPod but was removed some time around last week. If this is the template you are using, the closest equivalent isghcr.io/ai-dock/comfyui:latest
which RunPod callsComfyUI - AI-Dock
, so you can just edit the template by appending-jupyter
to the end of the title for the Container Image and then confirm with "Set Overrides." I am also using a network volume.
Awesome thanks! I'll check this out.
@a-niess I set that template up and opened the port and it's asking for a login. Where do I create these credentials?
scratch that, found it!
@jgal14 The default with this container image is
user
password
and can be configured to something else.
@jgal14 The default with this container image is
user
password
and can be configured to something else.
Thank you again
@a-niess did you use a custom image to get the IP-adapter running on this template? Models I'm adding to the standard file structure aren't showing in workflows!
@jgal14 When you connect to the Jupyter notebook via Connect to HTTP Service [Port 8888]
:
cd
into the appropriate directory for where you want to add models.
E.g., if you're adding a LoRA then cd ComfyUI/models/loras
wget paste-model-URL
E.g. wget https://civitai.com/api/download/models/135867
will download the file add-detail-xl.safetensors
from the Detail Tweaker XL page on CivitAI.135867
without a .ckpt
or .safetensors
extension. In that case, rename the file with the extension so that it will appear in your ComfyUI nodes. To do so in the same terminal and directory,
mv current-name target-name
E.g. mv 135867 add-detail-xl.safetensors
Note: You can always use ls
to see the contents of the current directory (not sure how familiar you may be with command line.
Models installed as above will only persist if you are using a network volume in RunPod. If you are not using a network volume (found under RunPod's Storage tab on the left), then these models will not be there next time you spin up your pod.
@jgal14 When you connect to the Jupyter notebook via
Connect to HTTP Service [Port 8888]
:
- launch a new terminal
cd
into the appropriate directory for where you want to add models. E.g., if you're adding a LoRA thencd ComfyUI/models/loras
- copy the download URL of the model from its source. E.g., if on CivitAI or HF, copy the right-click on the download button/link for the model, then copy link address
- back in the jupyter notebook terminal once you are in the appropriate directory from step 1,
wget paste-model-URL
E.g.wget https://civitai.com/api/download/models/135867
will download the fileadd-detail-xl.safetensors
from the Detail Tweaker XL page on CivitAI.- In some cases, the model will only download as
135867
without a.ckpt
or.safetensors
extension. In that case, rename the file with the extension so that it will appear in your ComfyUI nodes. To do so in the same terminal and directory,mv current-name target-name
E.g.mv 135867 add-detail-xl.safetensors
Note: You can always use
ls
to see the contents of the current directory (not sure how familiar you may be with command line.Models installed as above will only persist if you are using a network volume in RunPod. If you are not using a network volume (found under RunPod's Storage tab on the left), then these models will not be there next time you spin up your pod.
Awesome thank you for the detailed response, this is up and running now.
Now, onto a serverless deployment!
I have installed all the models, FaceID, LoRAs, etc. in the proper directories per the Installation instructions. I am only working with SDXL and have omitted the SD1.5 installations.
I have also installed insightface via
pip install insightface
and onnxruntime viapip install onnxruntime
andpip install onnxruntime-gpu
within the/workspace/ComfyUI
directory. I am running ComfyUI via RunPod with theComfyUI - Jupyter
template.I appreciate any help others may be able to offer on this error: