PROBLEM AI GUIDANCE SOLUTIONS

ComfyUI

1. When the “Reconnecting” is happening, never close

1. Reconnecting error the popup. It may take a minute but let it finish.

2. You can see the “Queue size: ERR:” in the menu.

This happens when Comfy isn’t connected to the
host (it never reconnected).

3. When it says “Queue size: ERR” it is not uncommon

that Comfy will throw any error… The same can be
seen if you were to completely disconnect your
Colab runtime (you would see “queue size err”)

4. Check out your Colab runtime in the top right when

the “reconnecting” is happening.

5. Sometimes your GPU gets maxed out for a minute

and it takes a second for colab to catch up.

1. The reason: This means that the workflow you are

2. Freeze and ^C symbol in terminal running is heavy and gpu you are using can not
handle it

Solution: you have to either change the runtime to

something more powerful, lower the image
resolution or lower the video frames (if you run
vid2vid workflow)

1. Incorrect format: “0”:” (dark long hair)

3. Error occurred when executing, BatchPromptSchedule
Correct format: “0”:”(dark long hair)

There shouldn’t be a space between a quotation and

the start of the prompt.
 There shouldn’t be an “enter” between the
keyframes+prompt as well.

Or you have unnecessary symbols such as ( , “ ‘ )

1. Open Comfy Manager and hit the "update all"

4. Error(s) in loading state_dict for resampler button, then restart your Comfy (close everything
and delete your runtime).

2. If the first one doesn't work it can be your

checkpoint, so just switch out your checkpoint

1. You have to move your image sequence into your

5. ‘nontype’ object has no attribute ‘movedim’ Google Drive in the following directory

/content/drive/MyDrive/ComfyUI/input/ **← needs to

have the “/” after input**

use that file path instead of your local one once you
upload the images to the drive.

2. (In the path of the batch loader instead of writing

your Google Drive URL, try to write this path:

/content/drive/MyDrive/ComfyUI/input/your folder
name.

It should work after this if all the other steps are correct.

1. This is due to the checkpoint not loading correctly,

6. Error while deserializing header : HeaderTooLarge download a different checkpoint.

1. You need to find it in the custom_nodes file >

7. Yaml issues config.yaml.example and rename it to config.yaml.
 Then open it with text editor or notepad and change
True to False, then save and restart ComfyUI. I did it
in Google Drive.

1. Always connect to a GPU first before running the

8. GPU Erros in colab Environment setup. For future errors, copy and
paste the error message and use Chatbots like Bing
chat, Bard, and GPT to solve it. You will get almost
80% your errors fixed by it. If it's still showing an
error, then post it in the Ai-Guidance channel.

1. Run cell before the local tunnel to make sure the

9. Python3: Can’t open file: Can’t open file environment is running
content/main.py (errno2) no such file or directory
(Colab)

1. Firstly open your colab and go to your dependencies

10. New KSampler issue with –torch & –xformers cell. Which should be the environment cell.

You should see something like 'install dependencies' under

you'll see '!pip install xformers' and some text replace that
text with

!pip install xformers!=0.0.18 torch==2.0.1 torchsde

einops transformers>=4.25.1 safetensors>=0.3.0
aiohttp accelerate pyyaml Pillow scipy tqdm psutil
--extra-index-url
https://download.pytorch.org/whl/cu118
--extra-index-url
https://download.pytorch.org/whl/cu117

Once you paste this run the cell and all should work again

1. Just restart the runtime, delete session, and rerun it

11. ‘SyntaxError: Unexpected non-whitespace character again, this should solve issue.
after JSON at position 4’
 1. Make sure you are using a V100 GPU with High
12. CUDA Out Of Memory Error Ram mode enabled
2. If you are using an advanced model/checkpoint, it is
likely that more vram will be consumed. I suggest
you explore lighter versions of the model or
alternative models known for efficiency
3. Check if you're not running multiple Colab instances
in the background that may be a cause of high load
on the GPU. Consider closing any
runtimes/programs or tabs you may have open
during your session
4. Clear Colab's cache
5. Restart your runtime. Sometimes a fresh runtime
can solve problems
6. Consider a lower batch size
7. Consider using smaller image resolutions

1. This happens if you download the image from

13. .png workflow won’t download OneDrive without using the download button on the
top left. If you right-click the image it won't have the
data.

1. Update your ComfyUI

14 . KSamper error no attribute

1. This happens when you either use an SDXL

15. Error occurred when executing KSsampler: checkpoint with an SD1.5 controlnet or vice versa.
So, use the proper models.
'NoneType' object has no attribute 'shape'
 The reason: This often happens when the update intervals
16. Custom Node (Import Failed) are too large or when the repository is not clean. ("git repo
is dirty" means that there are unapproved changes in the
Git repository).

Solutions:
1. Press "try to fix" and then "try to update".

2. Uninstall the custom node and install it again.

3. If you want to play around with the code, you can
apply the command "git pull" in the folder of this
custom node.

!pip git reset --hard

!pip git pull

1. Open Comfy Manager and hit the "update all" button

17. Error occurred when executing IPAdapterApply: then complete restart your comfy ( close everything
and delete your runtime.)
2. If the first one doesn't work it can be your
checkpoint, so just switch out your checkpoint

You need to download Python version 3.10.6

18. Torch incompatibility ‎
PyTorch isn't supported in the version you've installed

1. Go to venv folder in Comfy / scripts / run activate.bat

19. Error occurred when executing KSampler: tuple index 2. Exit out of comfy if you're still in it.
out of range 3. Reload comfy
 Install the “custom scripts” custom node made by
20. My embeddings doesn’t show when I type embe…. pythongssss.

1. Check on which controlnet this happens and

21. Error while deserializing header: redownload that model.
MetadataIncompleteBuffer The download was corrupted somehow.

Checkpoint issue. You either have an outdated checkpoint

22. Error occurred when executing KSampler: mat1 and or are using an SDXL checkpoint in an SD1.5 workflow
mat2 shapes can not be multiplied
Using SDXL checkpoint or LoRA with SD1.5 ControlNet
model

Open a terminal and do the following

23. Ultraytics Failed To Install
For Mac:
pip3 install --force-reinstall ultralytics==8.0.195

For Windows:
pip install --force-reinstall ultralytics==8.0.195

For Colab: Open a new cell and do:

!pip install --force-reinstall ultralytics==8.0.195

Then restart comfyui and try again.

This error means that you do not have any models

24. Error occurred when executing SAMLoader: downloaded. Download any from the list provided:
SAMLoader.load_model() missing 1 required positional sam_vit_h click to download
argument: ‘model_name’ sam_vit_l click to download
sam_vit_b click to download
 sam_hq_vit_h click to download
sam_hq_vit_l click to download
sam_hq_vit_b click to download
mobile_sam click to download
FastSAM-s click to download
FastSAM-x click to download
Put them in the folder \ComfyUI\models\sams
If you don’t have “sams” folder, create one.

Perhaps your path is wrong. Check this out: 👇🏻

25. My checkpoints won’t load!

1. If updating ComfyUI-Manager itself fails, please go

26. ComfyUI manager failed to update to the ComfyUI-Manager directory and execute the
command:

!git update-ref refs/remotes/origin/main a361cc1

&& git fetch --all && git pull

You are missing IPAdapter Models. Go to IPAdapter-plus

27. IPAdapterModelLoader error: repo in github and download one.

Then put them in the:

ComfyUI\custom_nodes\ComfyUI_IPAdapter_plus\models

Open terminal in your python_embeded

28. DWPose preprocessor error:

And type:

python.exe -m pip uninstall onnxruntime

onnxruntime-gpu
Delete the remaining empty onnxruntime folder from
 Open terminal in your python_embeded again and type
python.exe -m pip install onnxruntime-gpu

Add "--gpu-only" and "--disable-smart-memory" at the

29. SyntaxError: Unexpected non-whitespace character end of the code (like in the image highlighted) where you
launch ComfyUI

Set lerp_alpha and decay_factor parameters on both

30. Grow Mask With Blur node lerp_alpha decay_factor nodes to 1.0
error:

Type the letters “cmd” in your ComfyUI Manager directory,

31. Custom node DB is being updated error then press enter
LOOOOOOOONG RESPONSE TIME

type this code and press enter

git update-ref refs/remote/origin/main a361cc1&&git
fetch &&git pull
 Load image cap should match.
32. Size mismatch in any point in video. KSampler,
IPAdapter, ControlNet and so on. Also, don’t use FP16 ControlNet versions in longer videos.
Use regular ones. They differ significantly in size

-- OR --

If this error appears in IPAdapter Apply, check the CLIP

Vision & IPAdapter model compatibility below:

And make sure you download the proper model from the
link in the GitHub repo. Models from ComfyUI-Manager
could be deprecated or incompatible.

Local
16. Install comfyui locally https://github.com/comfyanonymous/ComfyUI

Go on this link and follow instructions

WarpFusion
1. IsADirectoryError: [Errno 21] 1. You get this because you haven’t yet generated any
video with WarpFusion and you don’t have a
settings file. The path you put in in the
“settings_path” field points to a directory. It should
point to a settings file that Warp can use. Since you
don’t have any settings file, you can leave this field
empty for now
2. AttributeError: ‘str’ object has no attribute ‘keys’ 1. This means your prompt syntax is incorrect, the
correct syntax would be:

{“frame number”: [“prompt”]

“Next frame”: [“prompt”]}
3.

4.

5.

6.

Automatic1111

Connection Error, whatever is used to bridge the browser

1. Connection errored out. to the A1111 instance needs to be restarted / reconnected.
 Use another vae like klf8-anime (it’s in the AI ammo box)
2. Blurry images

https://stable-diffusion-art.com/install-windows/
3. How to install a1111 locally
Go on this website and follow instructions

On Colab you’ll see a 🔽 icon. Click on it. Then

4. Error with capture.capture & pygnork “disconnect and delete runtime”. Run all the cells from top
to bottom.

Pinokio

TO INVESTIGATE
1. No “dialog requesting admin permission” Try to open the Pinkio environment using the administrator
permissions.
 TO INVESTIGATE
2. Installation loop There are basically 2 reasons why you might end up in this
kind of installation loop.

Either an antivirus or your firewall is blocking the

installation of needed components

Or Pinokio detects a pre-installed Anaconda version and

skips the installation of the needed components

If there is a pre-installed Anaconda version you don't need,

please uninstall it.
Deactivate your antivirus program and firewall.
Delete the miniconda folder located in .\pinokio\bin
Try to install the App again
Pinokio will now install Miniconda and the other
components again properly.

TO INVESTIGATE
3. “SOMETHING” is not recognized as an internal or Maybe creating new venv could help
external command