structggml_tensor* c =ggml_new_tensor_4d(res_ctx, GGML_TYPE_F32,1024,2,1,1);

ggml_set_f32(c,0.5);

structggml_cplan cplan;

size_t ctx_size =10 *1024 *1024;// 10MB

{

ctx_size +=ggml_used_mem(ctx) +ggml_used_mem_of_data(ctx);

structggml_cgraph* diffusion_graph =ggml_build_forward_ctx(ctx, out);

structggml_cplancplan =ggml_graph_plan(diffusion_graph, n_threads);

cplan =ggml_graph_plan(diffusion_graph, n_threads);

ctx_size += cplan.work_size;

LOG_DEBUG("diffusion context need %.2fMB static memory, with work_size needing %.2fMB",

ggml_hold_dynamic_tensor(out);

structggml_cgraph* diffusion_graph =ggml_build_forward_ctx(ctx, out);

structggml_cplancplan =ggml_graph_plan(diffusion_graph, n_threads);

cplan =ggml_graph_plan(diffusion_graph, n_threads);

ggml_set_dynamic(ctx,false);

structggml_tensor* buf =ggml_new_tensor_1d(ctx, GGML_TYPE_I8, cplan.work_size);

true);

std::vector<int>& tokens = tokens_and_weights.first;

std::vector<float>& weights = tokens_and_weights.second;

structggml_cplan cplan;

size_t ctx_size =10 *1024 *1024;// 10MB

{

structggml_tensor* hidden_states = cond_stage_model.text_model.forward(ctx, input_ids);

structggml_cgraph cond_graph =ggml_build_forward(hidden_states);

structggml_cplancplan =ggml_graph_plan(&cond_graph, n_threads);

structggml_cgraph* cond_graph =ggml_build_forward_ctx(ctx,hidden_states);

cplan =ggml_graph_plan(cond_graph, n_threads);

ctx_size += cplan.work_size;

ctx_size +=ggml_used_mem(ctx) +ggml_used_mem_of_data(ctx);

structggml_tensor* x =ggml_dup_tensor(res_ctx,x_t);

copy_ggml_tensor(x,x_t);

structggml_cplan cplan;

size_t ctx_size =10 *1024 *1024;// 10MB

ctx_size +=ggml_used_mem(ctx) +ggml_used_mem_of_data(ctx);

structggml_cgraph* diffusion_graph =ggml_build_forward_ctx(ctx, out);

structggml_cplancplan =ggml_graph_plan(diffusion_graph, n_threads);

cplan =ggml_graph_plan(diffusion_graph, n_threads);

ctx_size += cplan.work_size;

LOG_DEBUG("diffusion context need %.2fMB static memory, with work_size needing %.2fMB",

ggml_hold_dynamic_tensor(out);

structggml_cgraph* diffusion_graph =ggml_build_forward_ctx(ctx, out);

structggml_cplancplan =ggml_graph_plan(diffusion_graph, n_threads);

cplan =ggml_graph_plan(diffusion_graph, n_threads);

ggml_set_dynamic(ctx,false);

structggml_tensor* buf =ggml_new_tensor_1d(ctx, GGML_TYPE_I8, cplan.work_size);

int64_t W = x->ne[0];

int64_t H = x->ne[1];

structggml_tensor* result =NULL;

structggml_cplan cplan;

size_t ctx_size =10 *1024 *1024;// 10MB

ctx_size +=ggml_used_mem(ctx) +ggml_used_mem_of_data(ctx);

structggml_cgraph* vae_graph =ggml_build_forward_ctx(ctx, moments);

structggml_cplancplan =ggml_graph_plan(vae_graph, n_threads);

cplan =ggml_graph_plan(vae_graph, n_threads);

ctx_size += cplan.work_size;

LOG_DEBUG("vae context need %.2fMB static memory, with work_size needing %.2fMB",

int64_t W = z->ne[0];

int64_t H = z->ne[1];

structggml_tensor* result_img =NULL;

structggml_cplan cplan;

{

float* vec = (float*)z->data;

ctx_size +=ggml_used_mem(ctx) +ggml_used_mem_of_data(ctx);

structggml_cgraph* vae_graph =ggml_build_forward_ctx(ctx, img);

structggml_cplancplan =ggml_graph_plan(vae_graph, n_threads);

cplan =ggml_graph_plan(vae_graph, n_threads);

ctx_size += cplan.work_size;

LOG_DEBUG("vae context need %.2fMB static memory, with work_size needing %.2fMB",