Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Please update your install guide. #175

Closed
TCFrancis opened this issue Jun 25, 2024 · 6 comments
Closed

Please update your install guide. #175

TCFrancis opened this issue Jun 25, 2024 · 6 comments

Comments

@TCFrancis
Copy link

TCFrancis commented Jun 25, 2024

Please update your install guide. Please...

Sorry to say it, but the guide is a mess. I am pulling my hair out again and again...
I want to install the 3D-Pack on a second machine.
And a guide from the issues-forum, that worked for me a few weeks ago, is again broken. And fails to install ComfyUI-3D-Pack.

Please:

Imagine a person with a fresh install of windows.

With anaconda it is easy to start totally new and fresh, with no prior python installation.

Then write down all the comands, that are needed to install.
Starting with the right python version.

You write
Assume you have already downloaded ComfyUI & Configed your CUDA environment.

There are so narrow and forced requirements with only very specific versions working, then please, point out, how to set it up from the beginning.

I quote:
[IMPORTANT!!!]
Currently this package is only been tested in following setups:

Windows 10/11 (Tested on my laptop)
Ubuntu 23.10 https://github.com/MrForExample/ComfyUI-3D-Pack/issues/16
ComfyUI python_embed/Miniconda/Conda Python 3.11.x
Torch version >= 2.1.2+cu121

But how to get to that torch and ComfyUI version. How to get to that Pyton version 3.11.x?

You start with install version 0...
is it nessasary to start with 0, or is it an alternative to install version 1?
I do not know chocolatey... it is another unfinished rabbit hole installation fix?

Please....

As far as I understand the mandatory versions of ComfyUI-3D-Pack, only python 3.11 does work. as you mention it also unter [IMPORTANT!!!]

why does the guide talk about other python versions, even if they are greyed out?

install_windows_portable_win_py310_cu121.bat

install_windows_portable_win_py312_cu121.bat

Please....
Please...
Update the install guide...
I do not want to waste 3 days again to finally getting it working.

@TCFrancis
Copy link
Author

Following your guide at step ">pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121" I get the error messages in RED:

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
huggingface-hub 0.23.0 requires packaging>=20.9, which is not installed.
open-clip-torch 2.24.0 requires protobuf, which is not installed.
open-clip-torch 2.24.0 requires sentencepiece, which is not installed.

I will continue to list my progress here... But I guess this installation try is allready broken...

@TCFrancis
Copy link
Author

Yes, okay.... Here I am at step "
pip install -r ./ComfyUI/requirements.txt"

well there is no ComfyUI present in that folder...

You missed a step there to describe how to get ComfyUI, and in what way, shape or form...
Install package with python embedded,... Well, I follow the miniconda install guide here or not? Then I guess it is Pip install?
See how that gets confusing?

@TCFrancis
Copy link
Author

Okay... Another try... Following the path "Alternatively you can check this tutorial: Installing ComfyUI with Miniconda On Windows and Mac"

Looking at the guide there...

  1. download Miniconda3 -yep, check, i got it allready and i know how to create enviormnets with conda create... I am 90 % sure.
  2. install Miniconda3 - yes, check, see above.
  3. pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu121
    

Wait a minute here... Am I am beeing expected to skip "conda create" and "conda activate" an enviorment. is it not system-wide then? Am I messing up here?

Okay, doing it anyway... Still in the conda enviornment. all succesfully with "Requirement already satisfied"

  1. okay, here i git clone ComfyUI. I know that territory.
    Why is there suddenly a guide how to install git? Hm.... I have git... and using it since, I do not know... Many years.
    Fine, some people do not have it. I get it. But in a structured installation guide I would put git and miniconda installation under a Section "Prerequisite" not in the normal install guide.

Losing track of my thoughts, continue with my installation effort.

2."Clone the ComfyUI repository" --- Never mind, I did it allready with git clone...
Or should I double my effort here?

I jump over the step 3... I guess, i know it is allready done with git clone...

Okay... next Visual Studio Code... I have installed it allready, I know it is running.
Why in the whatever should I use the VS Code terminal... I am in the Windows Comand prompt...
Questioning my existance...
---pip install -r requirements.txt Is this for Visual Studio Code or ComfyUI...?
Where am I, Who am I? :D Sweating not only from a warm summer night, I turn on my air-ventilator... Laughing out loud... :D

I guess I stay in Windows Comamd Prompt World for pip install -r requirements.txt

Getting RED ERROR MESSAGES: ERROR:
pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
open-clip-torch 2.24.0 requires protobuf, which is not installed.
open-clip-torch 2.24.0 requires sentencepiece, which is not installed.

This should not happen, because it is a normal ComfyUI installation until now... Or not?... Looking it up on the official ComfyUI installation...

Looking at the official ComfyUI installation guide I see the folloing comand...:
pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu121
-pip install -r requirements.txt (from ComfyUI) Getting RED ERRORS

I guess this is hopeless...
conda remove --name ENV_NAME --all, Deleting the Crap broken ComfyUI folder...
Start fresh. Now with the official ComfyUI Guide... What could go wrong...? Nothing.... Or?

Find out in another episode from... How to Install ComfyUI-3d-Pack...

@TCFrancis
Copy link
Author

TCFrancis commented Jun 25, 2024

Fresh restart. Doing it the official Guide way with miniconda now: I know Python 3.11 once did work for me with ComfyUI-3D-Pack... so:
Windows Command promt:
1.conda create -n py311 python=3.11
2.conda activate py311
3.git clone https://github.com/comfyanonymous/ComfyUI
4. cd ComfyUI
5.pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu121
6. pip install -r requirements.txt

No error messages in red, all white and shiny
7, (Copying a model checkpoint into model/checkpoint - Folder.
8. python main.py
Getting an error about numpy missing....
Searching around the internet... There is a new version 2 release of numpy on the 16.06... And it is mentioned on the official ComfyUI-Issues Forum that numpy 2 will break some dependencies...
comfyanonymous/ComfyUI#3794
9. Okay... Related to every isntall of ComfyUI... So Doing as told: 10
10. pip install numpy==1.26.4
11. Fire up Comfyui with python main.py
12. No errors. So opening http:https://127.0.0.1:8188/ in browser.
13. Choosing the checkpoint from Point 7.
14. Queue Prompt... Image gets created... Oofff... Yes... YES!

@TCFrancis
Copy link
Author

Now back to ComfyUI-3D... Hm... Where do I pick up my attempt to get ComfyUI-3D-Pack working....

Where do I pick up from the install guide... Hm...
cd C:\Users\reall\Softwares\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-3D-Pack...
The install guide does not mention to git clone the repo, but whatever... I guess it is best to start with

  1. cd ComfyUI/custom_nodes
  2. git clone https://github.com/MrForExample/ComfyUI-3D-Pack
  3. Receiving objects: 7% (179/2341)... Sigh... I wish my new internet Company would install optical fiber allready...
  4. Then I get to choose between: install_miniconda.bat and pip install -r requirements.txt and pip install -r requirements_post.txt
  5. I choose the pip- route... It is split into 2 parts... So I can see, if Part 1 is failing or successfully...
  6. cd ComfyUI-3D-Pack
  7. pip install -r requirements.txt
  8. pip install -r requirements_post.txt
  9. Success.... Uff...

@TCFrancis
Copy link
Author

TCFrancis commented Jun 26, 2024

Well I guess the requirements.txt and post are fixed now... I did an install in the past and it was all messed up.

Shame on me. Could have tested it before. Sorry for this long posts...

But I wanted to point out my confusions following the install guide. And maybe someone can have a little laugh...

My suggestion is to make every step clear. Split the guide in prerequisite and the Install-Guide itself So people can follow it and do not jump around, missing steps.

I filtered it down to how I would write the guide:

Important Notes:
Currently this package is only been tested in following setups:
Windows 10/11 (Tested on my laptop)
Ubuntu 23.10 #16
ComfyUI python_embed/Miniconda/Conda Python 3.11.x
Torch version >= 2.1.2+cu121

Prerequisite:
Install git
Install Visual Studio Code
Install Miniconda

Install Guide:
CMD-Command-Prompt in a Folder, where you want ComfyUI installed

  1. conda create -n py311 python=3.11
  2. conda activate py311
  3. git clone https://github.com/comfyanonymous/ComfyUI
  4. cd ComfyUI
  5. pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu121
  6. pip install -r requirements.txt
  7. pip install numpy==1.26.4 (maybe, only temporary, because of: I don't know why this error happened comfyanonymous/ComfyUI#3794)
  8. cd custom_nodes
  9. git clone https://github.com/MrForExample/ComfyUI-3D-Pack
  10. cd ComfyUI-3D-Pack
  11. pip install -r requirements.txt
  12. pip install -r requirements_post.txt

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant