DarthAffe/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2025-12-12 13:28:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: make the immediate LoRA apply mode work better when using Vulkan (#1021)

Browse Source
This commit is contained in:
leejet 2025-11-30 12:08:25 +08:00 committed by GitHub
parent ab7e8d285e
commit bc80225336
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 18 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
ggml_extend.hpp
View File

@ -1141,6 +1141,14 @@ __STATIC_INLINE__ struct ggml_tensor* ggml_ext_ones(struct ggml_context* ctx,
} }
__STATIC_INLINE__ ggml_tensor* ggml_ext_cast_f32(ggml_context* ctx, ggml_tensor* a) { __STATIC_INLINE__ ggml_tensor* ggml_ext_cast_f32(ggml_context* ctx, ggml_tensor* a) {
#ifdef SD_USE_VULKAN
auto zero_index = ggml_get_tensor(ctx, "ggml_runner_build_in_tensor:zero_int");
auto out = ggml_reshape_1d(ctx, a, ggml_nelements(a));
out = ggml_get_rows(ctx, out, zero_index);
out = ggml_reshape(ctx, out, a);
// auto out = ggml_cast(ctx, a, GGML_TYPE_F32);
return out;
#else
auto out = ggml_reshape_2d(ctx, a, 1, ggml_nelements(a)); auto out = ggml_reshape_2d(ctx, a, 1, ggml_nelements(a));
ggml_tensor* one = ggml_ext_ones(ctx, 1, 1, 1, 1); // [1,] ggml_tensor* one = ggml_ext_ones(ctx, 1, 1, 1, 1); // [1,]
if (ggml_is_transposed(out)) { if (ggml_is_transposed(out)) {
@ -1148,7 +1156,8 @@ __STATIC_INLINE__ ggml_tensor* ggml_ext_cast_f32(ggml_context* ctx, ggml_tensor*
} else { } else {
out = ggml_mul_mat(ctx, out, one); out = ggml_mul_mat(ctx, out, one);
} }
out = ggml_reshape(ctx, out, a); out = ggml_reshape(ctx, out, a);
#endif
return out; return out;
} }
@ -1556,6 +1565,9 @@ protected:
std::vector<float> one_vec = {1.f}; std::vector<float> one_vec = {1.f};
ggml_tensor* one_tensor = nullptr; ggml_tensor* one_tensor = nullptr;
std::vector<int> zero_int_vec = {0};
ggml_tensor* zero_int_tensor = nullptr;
std::map<struct ggml_tensor*, const void*> backend_tensor_data_map; std::map<struct ggml_tensor*, const void*> backend_tensor_data_map;
std::map<std::string, struct ggml_tensor*> cache_tensor_map; // name -> tensor std::map<std::string, struct ggml_tensor*> cache_tensor_map; // name -> tensor
const std::string final_result_name = "ggml_runner_final_result_tensor"; const std::string final_result_name = "ggml_runner_final_result_tensor";
@ -1626,10 +1638,15 @@ protected:
one_tensor = ggml_new_tensor_1d(compute_ctx, GGML_TYPE_F32, 1); one_tensor = ggml_new_tensor_1d(compute_ctx, GGML_TYPE_F32, 1);
ggml_set_name(one_tensor, "ggml_runner_build_in_tensor:one"); ggml_set_name(one_tensor, "ggml_runner_build_in_tensor:one");
set_backend_tensor_data(one_tensor, one_vec.data()); set_backend_tensor_data(one_tensor, one_vec.data());
zero_int_tensor = ggml_new_tensor_1d(compute_ctx, GGML_TYPE_I32, 1);
ggml_set_name(zero_int_tensor, "ggml_runner_build_in_tensor:zero_int");
set_backend_tensor_data(zero_int_tensor, zero_int_vec.data());
} }
void prepare_build_in_tensor_after(struct ggml_cgraph* gf) { void prepare_build_in_tensor_after(struct ggml_cgraph* gf) {
ggml_build_forward_expand(gf, one_tensor); ggml_build_forward_expand(gf, one_tensor);
ggml_build_forward_expand(gf, zero_int_tensor);
} }
struct ggml_cgraph* new_graph_custom(size_t graph_size) { struct ggml_cgraph* new_graph_custom(size_t graph_size) {