A frankensteinian amalgamation of notebooks, models and techniques for the generation of AI Art and Animations.
Follow these steps for the first time that you are running Disco Diffusion from Windows.
From Anaconda Powershell Prompt:
This command will allow you to use conda
from a “regular” powershell session.
conda init powershell
exit
From your VS Code Powershell prompt:
Note: These commands should be run from the working directory of this cloned repository.
This command will pull all dependencies needed by Disco Diffusion into a conda environment called discodiffusion
conda env create -f environment.yml
conda activate discodiffusion
Note: If you have already activated a discodiffusion
conda environment, you can refresh it if this repo has changed by typing:
conda env update --prefix discodiffusion --file environment.yml --prune
Compile pytorch3d
Note: These commands should be run from the working directory of this cloned repository.
For reason I’m not 100% clear on, pytorch3d
must be compiled in Windows. (Hence the requirement for C++ tool mentioned in Pre-requisties…)
git clone https://github.com/facebookresearch/pytorch3d.git
cd pytorch3d
python setup.py install
cd ..
Execute a test run:
The following test run will run with all defaults (so “the lighthouse run” as it is coined.) Image output and current image progress (progress.png
) will be stored in images_out
.
conda activate discodiffusion
python disco.py
Disco Diffusion
A frankensteinian amalgamation of notebooks, models and techniques for the generation of AI Art and Animations.
Changes in this Fork
disco.py
module intodd.py
so thatdisco.py
can become readable.disco.py
, parameters can be controlled by environment variables, command-line arguments, or config (YAML) files.examples/configs
for more)modifiers
andmultipliers
. (Seeexamples/configs
for more)Windows First-time Setup (Anaconda)
Follow these steps for the first time that you are running Disco Diffusion from Windows.
Pre-requisites
From Anaconda Powershell Prompt:
This command will allow you to use
conda
from a “regular” powershell session.From your VS Code Powershell prompt:
Note: These commands should be run from the working directory of this cloned repository.
This command will pull all dependencies needed by Disco Diffusion into a conda environment called
discodiffusion
Note: If you have already activated a
discodiffusion
conda environment, you can refresh it if this repo has changed by typing:conda env update --prefix discodiffusion --file environment.yml --prune
Compile
pytorch3d
Note: These commands should be run from the working directory of this cloned repository.
For reason I’m not 100% clear on,
pytorch3d
must be compiled in Windows. (Hence the requirement for C++ tool mentioned in Pre-requisties…)Execute a test run:
The following test run will run with all defaults (so “the lighthouse run” as it is coined.) Image output and current image progress (
progress.png
) will be stored inimages_out
.First-time Setup (
pip
)Windows PowerShell:
Windows Command Prompt:
Linux
Test Run
A batch run should begin. Kill with Control+C Example: