upstream-archive/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-08-02 16:20:41 -05:00
Code Issues Packages Projects Releases Wiki Activity

Conv2D direct support (#744)

Browse Source 
* Conv2DDirect for VAE stage

* Enable only for Vulkan, reduced duplicated code

* Cmake option to use conv2d direct

* conv2d direct always on for opencl

* conv direct as a flag

* fix merge typo

* Align conv2d behavior to flash attention's

* fix readme

* add conv2d direct for controlnet

* add conv2d direct for esrgan

* clean code, use enable_conv2d_direct/get_all_blocks

* format code

---------

Co-authored-by: leejet <leejet714@gmail.com>
This commit is contained in:
Daniele
2025-08-02 17:25:17 +00:00
committed by GitHub
parent f7f05fb185
commit 5b8996f74a
11 changed files with 151 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
upscaler.cpp
View File

@@ -9,9 +9,12 @@ struct UpscalerGGML {
std::shared_ptr<ESRGAN> esrgan_upscaler;
std::string esrgan_path;
int n_threads;
bool direct = false;
UpscalerGGML(int n_threads)
: n_threads(n_threads) {
UpscalerGGML(int n_threads,
bool direct = false)
: n_threads(n_threads),
direct(direct) {
}
bool load_from_file(const std::string& esrgan_path) {
@@ -47,6 +50,9 @@ struct UpscalerGGML {
}
LOG_INFO("Upscaler weight type: %s", ggml_type_name(model_data_type));
esrgan_upscaler = std::make_shared<ESRGAN>(backend, model_loader.tensor_storages_types);
if (direct) {
esrgan_upscaler->enable_conv2d_direct();
}
if (!esrgan_upscaler->load_from_file(esrgan_path)) {
return false;
}
@@ -104,14 +110,15 @@ struct upscaler_ctx_t {
};
upscaler_ctx_t* new_upscaler_ctx(const char* esrgan_path_c_str,
int n_threads) {
int n_threads,
bool direct = false) {
upscaler_ctx_t* upscaler_ctx = (upscaler_ctx_t*)malloc(sizeof(upscaler_ctx_t));
if (upscaler_ctx == NULL) {
return NULL;
}
std::string esrgan_path(esrgan_path_c_str);
upscaler_ctx->upscaler = new UpscalerGGML(n_threads);
upscaler_ctx->upscaler = new UpscalerGGML(n_threads, direct);
if (upscaler_ctx->upscaler == NULL) {
return NULL;
}