Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 09:15:18 +02:00
Code Issues Packages Projects Releases Wiki Activity

updated sdcpp, also set euler as default sampler

Browse Source
This commit is contained in:
Concedo
2024-12-01 17:00:20 +08:00
parent e93c2427b4
commit 2ba5949054
27 changed files with 1514 additions and 521 deletions
Download Patch File Download Diff File Expand all files Collapse all files
otherarch/sdcpp/diffusion_model.hpp
+21 -16
View File
@@ -17,7 +17,8 @@ struct DiffusionModel {
std::vector<struct ggml_tensor*> controls = {},
float control_strength = 0.f,
struct ggml_tensor** output = NULL,
struct ggml_context* output_ctx = NULL) = 0;
struct ggml_context* output_ctx = NULL,
std::vector<int> skip_layers = std::vector<int>()) = 0;
virtual void alloc_params_buffer() = 0;
virtual void free_params_buffer() = 0;
virtual void free_compute_buffer() = 0;
@@ -30,9 +31,10 @@ struct UNetModel : public DiffusionModel {
UNetModelRunner unet;
UNetModel(ggml_backend_t backend,
ggml_type wtype,
SDVersion version = VERSION_SD1)
: unet(backend, wtype, version) {
std::map<std::string, enum ggml_type>& tensor_types,
SDVersion version = VERSION_SD1,
bool flash_attn = false)
: unet(backend, tensor_types, "model.diffusion_model", version, flash_attn) {
}
void alloc_params_buffer() {
@@ -70,7 +72,9 @@ struct UNetModel : public DiffusionModel {
std::vector<struct ggml_tensor*> controls = {},
float control_strength = 0.f,
struct ggml_tensor** output = NULL,
struct ggml_context* output_ctx = NULL) {
struct ggml_context* output_ctx = NULL,
std::vector<int> skip_layers = std::vector<int>()) {
(void)skip_layers; // SLG doesn't work with UNet models
return unet.compute(n_threads, x, timesteps, context, c_concat, y, num_video_frames, controls, control_strength, output, output_ctx);
}
};
@@ -79,9 +83,8 @@ struct MMDiTModel : public DiffusionModel {
MMDiTRunner mmdit;
MMDiTModel(ggml_backend_t backend,
ggml_type wtype,
SDVersion version = VERSION_SD3_2B)
: mmdit(backend, wtype, version) {
std::map<std::string, enum ggml_type>& tensor_types)
: mmdit(backend, tensor_types, "model.diffusion_model") {
}
void alloc_params_buffer() {
@@ -119,8 +122,9 @@ struct MMDiTModel : public DiffusionModel {
std::vector<struct ggml_tensor*> controls = {},
float control_strength = 0.f,
struct ggml_tensor** output = NULL,
struct ggml_context* output_ctx = NULL) {
return mmdit.compute(n_threads, x, timesteps, context, y, output, output_ctx);
struct ggml_context* output_ctx = NULL,
std::vector<int> skip_layers = std::vector<int>()) {
return mmdit.compute(n_threads, x, timesteps, context, y, output, output_ctx, skip_layers);
}
};
@@ -128,9 +132,9 @@ struct FluxModel : public DiffusionModel {
Flux::FluxRunner flux;
FluxModel(ggml_backend_t backend,
ggml_type wtype,
SDVersion version = VERSION_FLUX_DEV)
: flux(backend, wtype, version) {
std::map<std::string, enum ggml_type>& tensor_types,
bool flash_attn = false)
: flux(backend, tensor_types, "model.diffusion_model", flash_attn) {
}
void alloc_params_buffer() {
@@ -168,9 +172,10 @@ struct FluxModel : public DiffusionModel {
std::vector<struct ggml_tensor*> controls = {},
float control_strength = 0.f,
struct ggml_tensor** output = NULL,
struct ggml_context* output_ctx = NULL) {
return flux.compute(n_threads, x, timesteps, context, y, guidance, output, output_ctx);
struct ggml_context* output_ctx = NULL,
std::vector<int> skip_layers = std::vector<int>()) {
return flux.compute(n_threads, x, timesteps, context, y, guidance, output, output_ctx, skip_layers);
}
};
#endif
#endif