Closed gimmel closed 1 year ago
Installation ran without errors, but I can’t start gui.py because of a missing module.
Traceback (most recent call last): File "/Users/mkaiser/sd-gui/gui.py", line 2, in import torch ModuleNotFoundError: No module named 'torch'
I already installed pytorch-nightly but it did not help.
pip install diffusers --force-reinstall --no-cache-dir
I tried this already, but it does not help. There is still the same error:
ModuleNotFoundError: No module named 'torch'
Make sure conda ml is activated
conda activate ml
cd sd-gui
python gui.py
conda info --env
doesn't show a ml environment. So there must be a problem in the beginning of the installation. I deleted everything and started over.
~ % /bin/bash -c "$(curl -fsSL https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh)"
Please run using "bash" or "sh", but not "." or "source"\n/bin/bash: line 11: return: can only `return' from a function or sourced script
But I can start the installation, which ends in
WARNING: md5sum mismatch of tar archive
expected: 41abe40ea68b8cba7668f9d0110eab4f
got: 0961cd918a612915577461e6ac5696f9
Unpacking payload ...
/bin/bash: line 409: /Users/mkaiser/miniconda3/conda.exe: cannot execute binary file
/bin/bash: line 411: /Users/mkaiser/miniconda3/conda.exe: cannot execute binary file
So I downloaded https://docs.conda.io/en/latest/miniconda.html https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh and installed it manually, which worked without errors.
Then I tried to create them environment:
conda create -n ml python=3.8.8
which results in
Collecting package metadata (current_repodata.json): done
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: failed
PackagesNotFoundError: The following packages are not available from current channels:
- python=3.8.8
Current channels:
- https://repo.anaconda.com/pkgs/main/osx-arm64
- https://repo.anaconda.com/pkgs/main/noarch
- https://repo.anaconda.com/pkgs/r/osx-arm64
- https://repo.anaconda.com/pkgs/r/noarch
Do I have to install python 3.8.8? My actual python version is 3.10.6.
WARNING: md5sum mismatch of tar archive
expected: 41abe40ea68b8cba7668f9d0110eab4f
got: 0961cd918a612915577461e6ac5696f9
Unpacking payload ...
/bin/bash: line 409: /Users/mkaiser/miniconda3/conda.exe: cannot execute binary file
/bin/bash: line 411: /Users/mkaiser/miniconda3/conda.exe: cannot execute binary file
You have installed the wrong conda version, it says conda.exe which is for windows. You need for macOS arm
Step 1
Completely remove miniconda
rm -rf ~/opt/miniconda3
Clean additional configurations with:
rm -rf ~/.condarc ~/.conda
Remove the path statement by using the vi terminal editor (delete everything inside this file. (to delete lines click D multiple times on your keyboard for each line
vi ~/.bash_profile
Once done, close vi terminal editor by pressing shift, colon :
, and typewq
hit enter.
Step 2 Install miniconda from this command
curl https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh -o Miniconda3-latest-MacOSX-arm64.sh
/bin/bash Miniconda3-latest-MacOSX-arm64.sh
Step 4
Create and activate a new conda environment named ml
python -V
this command will show you which version is installed on your computer
conda create -n ml python=YOUR VERSION
example: conda create -n ml python=3.9.13
conda activate ml
Install the needed Python packages
conda install pytorch torchvision torchaudio -c pytorch-nightly
conda install transformers
conda install -c conda-forge diffusers
conda install ftfy
Install git and git-lfs via Homebrew
brew install git git-lfs
Clone this repo and create an output folder
git clone https://github.com/FahimF/sd-gui.git
cd sd-gui
mkdir output
Clone the Hugging Face model repo - you will need the Hugging Face user and password for this step
git lfs install
git clone https://huggingface.co/CompVis/stable-diffusion-v1-4
Thank you. I already installed Miniconda3-latest-MacOSX-arm64.sh manually, but I did it again the way you recommended. I also removed /opt/miniconda3 (~/opt/miniconda3 did not exist).
But I still can’t create the condo environment:
~ % python -V
Python 3.10.6
~ % conda create -n ml python=3.10.6
Collecting package metadata (current_repodata.json): done
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: failed
PackagesNotFoundError: The following packages are not available from current channels:
- python=3.10.6
Current channels:
- https://repo.anaconda.com/pkgs/main/osx-arm64
- https://repo.anaconda.com/pkgs/main/noarch
- https://repo.anaconda.com/pkgs/r/osx-arm64
- https://repo.anaconda.com/pkgs/r/noarch
Conda seems not to use the installed python, but installs it from the available channels. I solved it by using the conda-forge channel:
conda create --channel conda-forge -n ml python=3.10.6
Now that the conda environment is finally working, the next problems occur.
~ % conda activate ml
(ml) ~ % conda install pytorch torchvision torchaudio -c pytorch-nightly
There are conflicts between the packages.
Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: failed with repodata from current_repodata.json, will retry with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Solving environment: /
Found conflicts! Looking for incompatible packages.
This can take several minutes. Press CTRL-C to abort.
failed
UnsatisfiableError: The following specifications were found to be incompatible with each other:
Output in format: Requested package -> Available versions
Package pytorch conflicts for:
pytorch
torchvision -> pytorch[version='1.10.2|1.13.0.dev20220917|1.13.0.dev20220918']
torchaudio -> pytorch==1.13.0.dev20220917
Package setuptools conflicts for:
python=3.10.6 -> pip -> setuptools
torchvision -> setuptools
Package typing-extensions conflicts for:
torchvision -> pytorch==1.10.2 -> typing-extensions
pytorch -> typing-extensions
reinstall torch
conda install pytorch torchvision torchaudio -c pytorch-nightly --force-reinstall
I also would recommend python version 3.9.13. (anything under 3.10 version, for apple sillicon)
I'm closing this issue since there has been no updates on this.
Installation ran without errors, but I can’t start gui.py because of a missing module.
Traceback (most recent call last): File "/Users/mkaiser/sd-gui/gui.py", line 2, in
import torch
ModuleNotFoundError: No module named 'torch'
I already installed pytorch-nightly but it did not help.