ComfyUI-3D-Pack icon indicating copy to clipboard operation
ComfyUI-3D-Pack copied to clipboard

Installation does not work. fix please <3

Open quantumbalance opened this issue 1 year ago • 5 comments

Hey. this seems so cool but alas it wouldn't install with the portable version or the non portable version. although i am a dumb ape. please fix. thank you

quantumbalance avatar Mar 07 '24 04:03 quantumbalance

First Install VS 2019 and C++ desktop tools Second Install VS 2022 and C++ desktop tools Next Install Cuda 12.1

When I use install_windows_portable_win_py311_cu121.bat

1 - xformers-0.0.24-cp311-cp311-win_amd64.whl (205.4 MB) 2 - torch-2.2.0%2Bcu121-cp311-cp311-win_amd64.whl (2454.8 MB) 3 - (error) - torchvision 0.17.1+cu121 requires torch==2.2.1+cu121, but you have torch 2.2.0+cu121 which is incompatible.

I try miniconda - same result I try anaconda3 - same result I try manual install from Issue #22 - the same result

I spend 8 hours straight . Help please

Falkonar avatar Mar 08 '24 14:03 Falkonar

First Install VS 2019 and C++ desktop tools Second Install VS 2022 and C++ desktop tools Next Install Cuda 12.1

When I use install_windows_portable_win_py311_cu121.bat

1 - xformers-0.0.24-cp311-cp311-win_amd64.whl (205.4 MB) 2 - torch-2.2.0%2Bcu121-cp311-cp311-win_amd64.whl (2454.8 MB) 3 - (error) - torchvision 0.17.1+cu121 requires torch==2.2.1+cu121, but you have torch 2.2.0+cu121 which is incompatible.

I try miniconda - same result I try anaconda3 - same result I try manual install from Issue #22 - the same result

I spend 8 hours straight . Help please

Hi friend, apologies for bad package management, so many things has happened to me recently, I'm doing my best to use my free time to do what I can to help 👍

Having said that, in your case, however, it seems the problem is caused by newest version of torch not compatible with old version of torchvision, that problem can be simply solved by:

Reinstall torch==2.2.0+cu121 torchvision==0.17.0+cu121:

  • If you are using ComfyUI embed Python env: Path to your ComfyUI python embed python.exe path -s -m pip install --force-reinstall torch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 --index-url https://download.pytorch.org/whl/cu121
  • If you are using Miniconda/Conda, after activate your conda environment: pip install --force-reinstall torch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 --index-url https://download.pytorch.org/whl/cu121

MrForExample avatar Mar 09 '24 09:03 MrForExample

First Install VS 2019 and C++ desktop tools Second Install VS 2022 and C++ desktop tools Next Install Cuda 12.1 When I use install_windows_portable_win_py311_cu121.bat 1 - xformers-0.0.24-cp311-cp311-win_amd64.whl (205.4 MB) 2 - torch-2.2.0%2Bcu121-cp311-cp311-win_amd64.whl (2454.8 MB) 3 - (error) - torchvision 0.17.1+cu121 requires torch==2.2.1+cu121, but you have torch 2.2.0+cu121 which is incompatible. I try miniconda - same result I try anaconda3 - same result I try manual install from Issue #22 - the same result I spend 8 hours straight . Help please

Hi friend, apologies for bad package management, so many things has happened to me recently, I'm doing my best to use my free time to do what I can to help 👍

Having said that, in your case, however, it seems the problem is caused by newest version of torch not compatible with old version of torchvision, that problem can be simply solved by:

Reinstall torch==2.2.0+cu121 torchvision==0.17.0+cu121:

  • If you are using ComfyUI embed Python env: Path to your ComfyUI python embed python.exe path -s -m pip install --force-reinstall torch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 --index-url https://download.pytorch.org/whl/cu121
  • If you are using Miniconda/Conda, after activate your conda environment: pip install --force-reinstall torch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 --index-url https://download.pytorch.org/whl/cu121

1 (python) run_nvidia_gpu.bat result - No module named 'rembg' (IMPORT FAILED) 2 (miniconda) run_nvidia_gpu_miniconda.bat relust - No module named 'chardet'

Press any key to continue . . .

Please create the new setup instructions for noobs... . And no luck running on Debian too (((

Falkonar avatar Mar 09 '24 13:03 Falkonar

What i normally do as a quick check before throwing in the gauntlet and asking. If python says "No module named 'xxxxx'" then i do a pip install xxxxx

Then either that works or your stuck. But here I am sure it will solve your problem. Was missing for me too.

The trick above with pip install --force-reinstall torch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 --index-url https://download.pytorch.org/whl/cu121 and then pip install rembg solved it.

I did have to a pip install --force-reinstall fsspec==2023.5.0 as there was a huggingface red text on imcompatibility.

and a pip install xformers as it wasnt there

zewolf5 avatar Mar 09 '24 13:03 zewolf5

What i normally do as a quick check before throwing in the gauntlet and asking. If python says "No module named 'xxxxx'" then i do a pip install xxxxx

Then either that works or your stuck. But here I am sure it will solve your problem. Was missing for me too.

The trick above with pip install --force-reinstall torch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 --index-url https://download.pytorch.org/whl/cu121 and then pip install rembg solved it.

I did have to a pip install --force-reinstall fsspec==2023.5.0 as there was a huggingface red text on imcompatibility.

and a pip install xformers as it wasnt there

Thank you so much for chewing me up. I'm really trying, but I wasn't attentive enough. The thing is that half a year ago I was just a 3D artist, but today I'm studying computer vision to keep up with the times. The endless errors are very nerve-wracking, but it's an endless buzz when the code works. Thanks again.

Falkonar avatar Mar 09 '24 14:03 Falkonar