Use ref_images parameter for Kontext images

Author: wbruna

otherarch/sdcpp/sdtype_adapter.cpp

@@ -687,16 +687,16 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
         }
     }
 
-    std::vector<sd_image_t> kontext_imgs;
+    std::vector<sd_image_t> reference_imgs;
     if(extra_image_data.size()>0 && loadedsdver==SDVersion::VERSION_FLUX && !loaded_model_is_chroma(sd_ctx))
     {
         for(int i=0;i<extra_image_data.size();++i)
         {
-            kontext_imgs.push_back(extraimage_references[i]);
+            reference_imgs.push_back(extraimage_references[i]);
         }
         if(!sd_is_quiet && sddebugmode==1)
         {
-            printf("\nFlux Kontext: Using %d reference images\n",kontext_imgs.size());
+            printf("\nFlux Kontext: Using %d reference images\n",reference_imgs.size());
         }
     }
 
@@ -741,9 +741,10 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
     params.guidance.slg.layer_end = sd_params->skip_layer_end;
     params.guidance.slg.scale = sd_params->slg_scale;
 
+    params.ref_images = reference_imgs.data();
+    params.ref_images_count = reference_imgs.size();
+
     kcpp_img_gen_params_t extra_params = {};
-    extra_params.kontext_imgs = kontext_imgs.data();
-    extra_params.kontext_img_count = kontext_imgs.size();
     extra_params.photomaker_references = photomaker_imgs.data();
     extra_params.photomaker_reference_count = photomaker_imgs.size();
 

otherarch/sdcpp/stable-diffusion.cpp

@@ -2288,22 +2288,6 @@ sd_image_t* generate_image(sd_ctx_t* sd_ctx, const sd_img_gen_params_t* sd_img_g
         LOG_INFO("encode_first_stage completed, taking %.2fs", (t1 - t0) * 1.0f / 1000);
     }
 
-    if (kcpp_img_gen_params) {
-        for (int i = 0; i < kcpp_img_gen_params->kontext_img_count; i++) {
-            ggml_tensor* img = ggml_new_tensor_4d(work_ctx, GGML_TYPE_F32, kcpp_img_gen_params->kontext_imgs[i].width, kcpp_img_gen_params->kontext_imgs[i].height, 3, 1);
-            sd_image_to_tensor(kcpp_img_gen_params->kontext_imgs[i].data, img);
-
-            ggml_tensor* latent = NULL;
-            if (!sd_ctx->sd->use_tiny_autoencoder) {
-                ggml_tensor* moments = sd_ctx->sd->encode_first_stage(work_ctx, img);
-                latent               = sd_ctx->sd->get_first_stage_encoding(work_ctx, moments);
-            } else {
-                latent = sd_ctx->sd->encode_first_stage(work_ctx, img);
-            }
-            ref_latents.push_back(latent);
-        }
-    }
-
     sd_image_t* result_images = generate_image_internal(sd_ctx,
                                                         work_ctx,
                                                         init_latent,

otherarch/sdcpp/stable-diffusion.h

@@ -189,8 +189,6 @@ typedef struct {
 } sd_img_gen_params_t;
 
 typedef struct {
-    sd_image_t* kontext_imgs;
-    int kontext_img_count;
     sd_image_t* photomaker_references;
     int photomaker_reference_count;
 } kcpp_img_gen_params_t;