Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-07 05:21:23 +02:00
Code Issues Packages Projects Releases Wiki Activity

refactor of added images done

Browse Source
This commit is contained in:
Concedo
2025-10-02 21:22:35 +08:00
parent df87da4694
commit f2ad0b78d4
1 changed files with 39 additions and 86 deletions
Download Patch File Download Diff File Expand all files Collapse all files
otherarch/sdcpp/sdtype_adapter.cpp
+39 -86
View File
@@ -674,18 +674,16 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
//for img2img
sd_image_t input_image = {0,0,0,nullptr};
std::vector<sd_image_t> extraimage_references;
extraimage_references.reserve(max_extra_images);
std::vector<std::vector<uint8_t>> extraimage_buffers;
extraimage_buffers.reserve(max_extra_images);
std::vector<sd_image_t> kontext_imgs;
std::vector<sd_image_t> wan_imgs;
std::vector<sd_image_t> photomaker_imgs;
bool is_wan = (loadedsdver == SDVersion::VERSION_WAN2 || loadedsdver == SDVersion::VERSION_WAN2_2_I2V || loadedsdver == SDVersion::VERSION_WAN2_2_TI2V);
bool is_kontext = (loadedsdver==SDVersion::VERSION_FLUX && !loaded_model_is_chroma(sd_ctx));
int nx, ny, nc;
int img2imgW = sd_params->width; //for img2img input
int img2imgH = sd_params->height;
int img2imgC = 3; // Assuming RGB image
//because the reference image can be larger than the output image, allocate at least enough for 1024x1024
const int imgMemNeed = std::max(img2imgW * img2imgH * img2imgC + 512, 1024 * 1024 * img2imgC + 512);
std::vector<std::vector<uint8_t>> resized_extraimage_bufs(max_extra_images, std::vector<uint8_t>(imgMemNeed));
std::string ts = get_timestamp_str();
if(!sd_is_quiet)
@@ -707,52 +705,45 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
}
input_extraimage_buffers.clear();
}
extraimage_buffers.clear();
extraimage_references.clear();
for(int i=0;i<extra_image_data.size() && i<max_extra_images;++i)
{
int nx2, ny2, nc2;
int desiredchannels = 3;
extraimage_buffers.push_back(kcpp_base64_decode(extra_image_data[i]));
input_extraimage_buffers.push_back(stbi_load_from_memory(extraimage_buffers[i].data(), extraimage_buffers[i].size(), &nx2, &ny2, &nc2, desiredchannels));
// Resize the image
float aspect_ratio = static_cast<float>(nx2) / ny2;
int desiredWidth = nx2;
int desiredHeight = ny2;
int smallestsrcdim = std::min(img2imgW,img2imgH);
if(desiredWidth > desiredHeight)
if(is_wan)
{
desiredWidth = smallestsrcdim;
desiredHeight = smallestsrcdim / aspect_ratio;
} else {
desiredHeight = smallestsrcdim;
desiredWidth = smallestsrcdim * aspect_ratio;
uint8_t * loaded = load_image_from_b64(extra_image_data[i],nx2,ny2,img2imgW,img2imgH,3);
if(loaded)
{
input_extraimage_buffers.push_back(loaded);
sd_image_t extraimage_reference;
extraimage_reference.width = nx2;
extraimage_reference.height = ny2;
extraimage_reference.channel = desiredchannels;
extraimage_reference.data = loaded;
wan_imgs.push_back(extraimage_reference);
}
}
//round dims to 64
desiredWidth = roundnearest(16,desiredWidth);
desiredHeight = roundnearest(16,desiredHeight);
desiredWidth = std::clamp(desiredWidth,64,1024);
desiredHeight = std::clamp(desiredHeight,64,1024);
if(!sd_is_quiet && sddebugmode==1)
else if (is_kontext || photomaker_enabled)
{
printf("Resize Extraimg: %dx%d to %dx%d\n",nx2,ny2,desiredWidth,desiredHeight);
uint8_t * loaded = load_image_from_b64(extra_image_data[i],nx2,ny2);
if(loaded)
{
input_extraimage_buffers.push_back(loaded);
sd_image_t extraimage_reference;
extraimage_reference.width = nx2;
extraimage_reference.height = ny2;
extraimage_reference.channel = desiredchannels;
extraimage_reference.data = loaded;
if(is_kontext)
{
kontext_imgs.push_back(extraimage_reference);
}
else
{
photomaker_imgs.push_back(extraimage_reference);
}
}
}
int resok = stbir_resize_uint8(input_extraimage_buffers[i], nx2, ny2, 0, resized_extraimage_bufs[i].data(), desiredWidth, desiredHeight, 0, desiredchannels);
if (!resok) {
printf("\nKCPP SD: resize extra image failed!\n");
output.data = "";
output.animated = 0;
output.status = 0;
return output;
}
sd_image_t extraimage_reference;
extraimage_reference.width = desiredWidth;
extraimage_reference.height = desiredHeight;
extraimage_reference.channel = desiredchannels;
extraimage_reference.data = resized_extraimage_bufs[i].data();
extraimage_references.push_back(extraimage_reference);
}
//ensure prompt has img keyword, otherwise append it
@@ -765,48 +756,10 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
sd_params->prompt = "person " + sd_params->prompt;
}
}
}
std::vector<sd_image_t> reference_imgs;
std::vector<sd_image_t> wan_imgs;
bool is_wan = (loadedsdver == SDVersion::VERSION_WAN2 || loadedsdver == SDVersion::VERSION_WAN2_2_I2V || loadedsdver == SDVersion::VERSION_WAN2_2_TI2V);
bool is_kontext = (loadedsdver==SDVersion::VERSION_FLUX && !loaded_model_is_chroma(sd_ctx));
if(extra_image_data.size()>0)
{
if(is_kontext)
{
for(int i=0;i<extra_image_data.size();++i)
{
reference_imgs.push_back(extraimage_references[i]);
}
if(!sd_is_quiet && sddebugmode==1)
{
printf("\nImage Gen: Using %d reference images\n",reference_imgs.size());
}
}
if(is_wan)
{
for(int i=0;i<extra_image_data.size();++i)
{
wan_imgs.push_back(extraimage_references[i]);
}
if(!sd_is_quiet && sddebugmode==1)
{
printf("\nImage Gen: Using %d video reference images\n",wan_imgs.size());
}
}
}
std::vector<sd_image_t> photomaker_imgs;
if(photomaker_enabled && extra_image_data.size()>0)
{
for(int i=0;i<extra_image_data.size();++i)
{
photomaker_imgs.push_back(extraimage_references[i]);
}
if(!sd_is_quiet && sddebugmode==1)
{
printf("\nPhotomaker: Using %d reference images\n",photomaker_imgs.size());
printf("\nImageGen References: Kontext=%d Wan=%d Photomaker=%d\n",kontext_imgs.size(),wan_imgs.size(),photomaker_imgs.size());
}
}
@@ -829,8 +782,8 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
params.vae_tiling_params.enabled = dotile;
params.batch_count = 1;
params.ref_images = reference_imgs.data();
params.ref_images_count = reference_imgs.size();
params.ref_images = kontext_imgs.data();
params.ref_images_count = kontext_imgs.size();
params.pm_params.id_images = photomaker_imgs.data();
params.pm_params.id_images_count = photomaker_imgs.size();