feat: add vulkan backend support (#291)

* Fix includes and init vulkan the same as llama.cpp

* Add Windows Vulkan CI

* Updated ggml submodule

* support epsilon as a parameter for ggml_group_norm

---------

Co-authored-by: Cloudwalk <[email protected]>
Co-authored-by: Oleg Skutte <[email protected]>
Co-authored-by: leejet <[email protected]>

.github/workflows/build.yml

@@ -148,6 +148,9 @@ jobs:
   windows-latest-cmake:
     runs-on: windows-2019
 
+    env:
+      VULKAN_VERSION: 1.3.261.1
+
     strategy:
       matrix:
         include:
@@ -163,6 +166,8 @@ jobs:
             defines: "-DSD_CUBLAS=ON -DSD_BUILD_SHARED_LIBS=ON"
           - build: "rocm5.5"
             defines: '-G Ninja -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DSD_HIPBLAS=ON -DCMAKE_BUILD_TYPE=Release -DAMDGPU_TARGETS="gfx1100;gfx1102;gfx1030" -DSD_BUILD_SHARED_LIBS=ON'
+          - build: 'vulkan'
+            defines: "-DSD_VULKAN=ON -DSD_BUILD_SHARED_LIBS=ON"
     steps:
       - name: Clone
         id: checkout
@@ -192,6 +197,14 @@ jobs:
         uses: urkle/action-get-ninja@v1
         with:
           version: 1.11.1
+      - name: Install Vulkan SDK
+        id: get_vulkan
+        if: ${{ matrix.build == 'vulkan' }}
+        run: |
+          curl.exe -o $env:RUNNER_TEMP/VulkanSDK-Installer.exe -L "https://sdk.lunarg.com/sdk/download/${env:VULKAN_VERSION}/windows/VulkanSDK-${env:VULKAN_VERSION}-Installer.exe"
+          & "$env:RUNNER_TEMP\VulkanSDK-Installer.exe" --accept-licenses --default-answer --confirm-command install
+          Add-Content $env:GITHUB_ENV "VULKAN_SDK=C:\VulkanSDK\${env:VULKAN_VERSION}"
+          Add-Content $env:GITHUB_PATH "C:\VulkanSDK\${env:VULKAN_VERSION}\bin"
 
       - name: Build
         id: cmake_build

CMakeLists.txt

@@ -27,6 +27,7 @@ option(SD_BUILD_EXAMPLES             "sd: build examples" ${SD_STANDALONE})
 option(SD_CUBLAS                     "sd: cuda backend" OFF)
 option(SD_HIPBLAS                    "sd: rocm backend" OFF)
 option(SD_METAL                      "sd: metal backend" OFF)
+option(SD_VULKAN                     "sd: vulkan backend" OFF)
 option(SD_SYCL                       "sd: sycl backend" OFF)
 option(SD_FLASH_ATTN                 "sd: use flash attention for x4 less memory usage" OFF)
 option(SD_FAST_SOFTMAX               "sd: x1.5 faster softmax, indeterministic (sometimes, same seed don't generate same image), cuda only" OFF)
@@ -45,6 +46,12 @@ if(SD_METAL)
     add_definitions(-DSD_USE_METAL)
 endif()
 
+if (SD_VULKAN)
+    message("Use Vulkan as backend stable-diffusion")
+    set(GGML_VULKAN ON)
+    add_definitions(-DSD_USE_VULKAN)
+endif ()
+
 if (SD_HIPBLAS)
     message("Use HIPBLAS as backend stable-diffusion")
     set(GGML_HIPBLAS ON)

ggml_extend.hpp

@@ -32,6 +32,10 @@
 #include "ggml-metal.h"
 #endif
 
+#ifdef SD_USE_VULKAN
+#include "ggml-vulkan.h"
+#endif
+
 #ifdef SD_USE_SYCL
 #include "ggml-sycl.h"
 #endif
@@ -655,7 +659,7 @@ __STATIC_INLINE__ struct ggml_tensor* ggml_nn_attention(struct ggml_context* ctx
                                                         struct ggml_tensor* k,
                                                         struct ggml_tensor* v,
                                                         bool mask = false) {
-#if defined(SD_USE_FLASH_ATTENTION) && !defined(SD_USE_CUBLAS) && !defined(SD_USE_METAL) && !defined(SD_USE_SYCL)
+#if defined(SD_USE_FLASH_ATTENTION) && !defined(SD_USE_CUBLAS) && !defined(SD_USE_METAL) && !defined(SD_USE_VULKAN) && !defined(SD_USE_SYCL)
     struct ggml_tensor* kqv = ggml_flash_attn(ctx, q, k, v, false);  // [N * n_head, n_token, d_head]
 #else
     float d_head = (float)q->ne[0];

model.cpp

@@ -21,6 +21,10 @@
 #include "ggml-metal.h"
 #endif
 
+#ifdef SD_USE_VULKAN
+#include "ggml-vulkan.h"
+#endif
+
 #define ST_HEADER_SIZE_LEN 8
 
 uint64_t read_u64(uint8_t* buffer) {

stable-diffusion.cpp

@@ -160,6 +160,15 @@ class StableDiffusionGGML {
         ggml_backend_metal_log_set_callback(ggml_log_callback_default, nullptr);
         backend = ggml_backend_metal_init();
 #endif
+#ifdef SD_USE_VULKAN
+        LOG_DEBUG("Using Vulkan backend");
+        for (int device = 0; device < ggml_backend_vk_get_device_count(); ++device) {
+            backend = ggml_backend_vk_init(device);
+        }
+        if(!backend) {
+            LOG_WARN("Failed to initialize Vulkan backend");
+        }
+#endif
 #ifdef SD_USE_SYCL
         LOG_DEBUG("Using SYCL backend");
         backend = ggml_backend_sycl_init(0);
@@ -170,7 +179,7 @@ class StableDiffusionGGML {
             backend = ggml_backend_cpu_init();
         }
 #ifdef SD_USE_FLASH_ATTENTION
-#if defined(SD_USE_CUBLAS) || defined(SD_USE_METAL) || defined(SD_USE_SYCL)
+#if defined(SD_USE_CUBLAS) || defined(SD_USE_METAL) || defined (SD_USE_SYCL) || defined(SD_USE_VULKAN)
         LOG_WARN("Flash Attention not supported with GPU Backend");
 #else
         LOG_INFO("Flash Attention enabled");

upscaler.cpp

@@ -24,6 +24,10 @@ struct UpscalerGGML {
         ggml_backend_metal_log_set_callback(ggml_log_callback_default, nullptr);
         backend = ggml_backend_metal_init();
 #endif
+#ifdef SD_USE_VULKAN
+        LOG_DEBUG("Using Vulkan backend");
+        backend = ggml_backend_vk_init(0);
+#endif
 #ifdef SD_USE_SYCL
         LOG_DEBUG("Using SYCL backend");
         backend = ggml_backend_sycl_init(0);