sd.cpp.gui.wx.git

sd.cpp.gui.wx

wxWidgets based Stable Diffusion C++ GUi

Created by Ferenc Szontágh szf@fsociety.hu @2024-01-30

No licence yet...

- Prepare

install vcpkg on your system...
configure the vcpkg in CMakePresets.json and CMakeUserPresets.json
run configure, itt will be ~15mins to download and configure the dependencies

- Externals!

in the external forlder, you can find the precompuled stable-diffusion.cpp and ggml as submodulde to it.
The default libs are configured with only CUDA, the release is Debug and Release. To use with another version (for example cpu only), you need to compile and overwrite

Build stable-diffusion.cpp

git clone --recursive https://github.com/leejet/stable-diffusion.cpp
cd stable-diffusion.cpp
git pull origin master
git submodule init
git submodule update

mkdir build
cd build
#with CUDA 
cmake .. -DSD_CUBLAS=ON -DBUILD_SHARED_LIBS=ON

#build
cmake --build . --config Release
#or Debug
cmake --build . --config Debug

#test
./bin/sd -m ../models/sd-v1-4.ckpt -p "a lovely cat"

- TO-DOs

See @git.spamming.hu ticketing system

Hash generation from binary

Repo: okdshin/PicoSHA2

RAW file: picosha2.h

Example:
C++ std::ifstream f("file.txt", std::ios::binary); std::vector<unsigned char> s(picosha2::k_digest_size); picosha2::hash256(f, s.begin(), s.end());

restclient-cpp

Curl wrapper to civiati api:
repo

installed via vcpkg as cmake dependency

CivitAi API docs

- IDE

Use the vscode... Install the extensions which is recommended by the workspace

Table Of Thread signal tokens (W.I.P)

thread tokens	payload	desc	comment
MODEL_LOAD_START	QM::QueueItem	model load started	only occurs, when no model loaded, or the job have another model
MODEL_LOAD_DONE	sd_ctx*	model is loaded	the model is loaded and the pointer returning as payload. The pointer will be reused in the next etap
MODEL_LOAD_ERROR	QM::QueueItem	model is not loaded, something went wrong at the backend
GENERATION_ERROR	QM::QueueItem
QUEUE	QM::QueueItem	simple queue msgs, the string is always a number
GENERATION_START	QM::QueueItem	the generation just started
GENERATION_PROGRESS		for progressbar, if sometimes the sd.cpp implements some progress handling	not used, while no progress info from the backend
GENERATION_DONE	sd_image_t *	everything is done, images saved to the output. Payload is the sd_image_t pointer

table generated with https://tabletomarkdown.com/convert-spreadsheet-to-markdown/