rename define to match upstream

Author: LostRuins

CMakeLists.txt

@@ -96,7 +96,7 @@ if (LLAMA_CUBLAS)
 
         add_compile_definitions(GGML_USE_LLAMAFILE)
         add_compile_definitions(GGML_USE_CUDA)
-        add_compile_definitions(SD_USE_CUBLAS)
+        add_compile_definitions(SD_USE_CUDA)
 
         if (LLAMA_CUDA_F16 OR LLAMA_CUDA_DMMV_F16)
             add_compile_definitions(GGML_CUDA_F16)
@@ -177,7 +177,7 @@ if (LLAMA_HIPBLAS)
         list(APPEND GGML_SOURCES_ROCM ${SRCS})
         file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/mmq*.cu")
         list(APPEND GGML_SOURCES_ROCM ${SRCS})
-        add_compile_definitions(GGML_USE_HIP GGML_USE_CUDA SD_USE_CUBLAS)
+        add_compile_definitions(GGML_USE_HIP GGML_USE_CUDA SD_USE_CUDA)
         add_library(ggml-rocm ${GGML_SOURCES_CUDA})
 
         file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-vec*q4_0-q4_0.cu")

Makefile

@@ -83,7 +83,7 @@ CLBLAST_FLAGS = -DGGML_USE_CLBLAST
 FAILSAFE_FLAGS = -DUSE_FAILSAFE
 VULKAN_FLAGS = -DGGML_USE_VULKAN -DSD_USE_VULKAN
 ifdef LLAMA_CUBLAS
-	CUBLAS_FLAGS = -DGGML_USE_CUDA -DSD_USE_CUBLAS
+	CUBLAS_FLAGS = -DGGML_USE_CUDA -DSD_USE_CUDA
 else
 	CUBLAS_FLAGS =
 endif
@@ -177,7 +177,7 @@ OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml/src/ggml-cuda/templat
 OBJS_CUDA_TEMP_INST += $(patsubst %.cu,%.o,$(wildcard ggml/src/ggml-cuda/template-instances/fattn-vec*f16-f16.cu))
 
 ifdef LLAMA_CUBLAS
-	CUBLAS_FLAGS = -DGGML_USE_CUDA -DSD_USE_CUBLAS -I/usr/local/cuda/include -I/opt/cuda/include -I$(CUDA_PATH)/targets/x86_64-linux/include
+	CUBLAS_FLAGS = -DGGML_USE_CUDA -DSD_USE_CUDA -I/usr/local/cuda/include -I/opt/cuda/include -I$(CUDA_PATH)/targets/x86_64-linux/include
 	CUBLASLD_FLAGS = -lcuda -lcublas -lcudart -lcublasLt -lpthread -ldl -lrt -L/usr/local/cuda/lib64 -L/opt/cuda/lib64 -L$(CUDA_PATH)/targets/x86_64-linux/lib -L$(CUDA_PATH)/lib64/stubs -L/usr/local/cuda/targets/aarch64-linux/lib -L/usr/local/cuda/targets/sbsa-linux/lib -L/usr/lib/wsl/lib
 	CUBLAS_OBJS = ggml-cuda.o ggml_v3-cuda.o ggml_v2-cuda.o ggml_v2-cuda-legacy.o
 	CUBLAS_OBJS += $(patsubst %.cu,%.o,$(filter-out ggml/src/ggml-cuda/ggml-cuda.cu, $(wildcard ggml/src/ggml-cuda/*.cu)))
@@ -256,7 +256,7 @@ ifdef DETECT_ROCWMMA
 	HIPFLAGS   += -DGGML_HIP_ROCWMMA_FATTN -I$(dir $(DETECT_ROCWMMA))
 endif
 
-	HIPFLAGS   += -DGGML_USE_HIP -DGGML_HIP_NO_VMM -DGGML_USE_CUDA -DSD_USE_CUBLAS $(shell $(ROCM_PATH)/bin/hipconfig -C)
+	HIPFLAGS   += -DGGML_USE_HIP -DGGML_HIP_NO_VMM -DGGML_USE_CUDA -DSD_USE_CUDA $(shell $(ROCM_PATH)/bin/hipconfig -C)
 	HIPLDFLAGS    += -L$(ROCM_PATH)/lib -Wl,-rpath=$(ROCM_PATH)/lib
 	HIPLDFLAGS    += -L$(ROCM_PATH)/lib64 -Wl,-rpath=$(ROCM_PATH)/lib64
 	HIPLDFLAGS    += -lhipblas -lamdhip64 -lrocblas

koboldcpp.py

@@ -3134,7 +3134,9 @@ def do_POST(self):
                 return
             try:
                 genparams = json.loads(body)
-                schema = genparams.get('schema', {})
+                schema = genparams.get('schema', None)
+                if not schema:
+                    schema = genparams
                 decoded = convert_json_to_gbnf(schema)
                 response_body = (json.dumps({"result": decoded,"success":(True if decoded else False)}).encode())
             except Exception as e:

otherarch/sdcpp/ggml_extend.hpp

@@ -28,7 +28,7 @@
 
 #include "model.h"
 
-#ifdef SD_USE_CUBLAS
+#ifdef SD_USE_CUDA
 #include "ggml-cuda.h"
 #endif
 
@@ -782,7 +782,7 @@ __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_VULKAN) && !defined(SD_USE_SYCL)
+#if defined(SD_USE_FLASH_ATTENTION) && !defined(SD_USE_CUDA) && !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];
@@ -938,7 +938,7 @@ __STATIC_INLINE__ struct ggml_tensor* ggml_nn_group_norm(struct ggml_context* ct
 }
 
 __STATIC_INLINE__ void ggml_backend_tensor_get_and_sync(ggml_backend_t backend, const struct ggml_tensor* tensor, void* data, size_t offset, size_t size) {
-#if defined(SD_USE_CUBLAS) || defined(SD_USE_SYCL)
+#if defined(SD_USE_CUDA) || defined(SD_USE_SYCL)
     if (!ggml_backend_is_cpu(backend)) {
         ggml_backend_tensor_get_async(backend, tensor, data, offset, size);
         ggml_backend_synchronize(backend);

otherarch/sdcpp/stable-diffusion.cpp

@@ -161,7 +161,7 @@ class StableDiffusionGGML {
                         bool diffusion_flash_attn) {
         use_tiny_autoencoder = taesd_path.size() > 0;
         std::string taesd_path_fixed = taesd_path;
-#ifdef SD_USE_CUBLAS
+#ifdef SD_USE_CUDA
         LOG_DEBUG("Using CUDA backend");
         backend = ggml_backend_cuda_init(0);
 #endif

otherarch/sdcpp/upscaler.cpp

@@ -15,7 +15,7 @@ struct UpscalerGGML {
     }
 
     bool load_from_file(const std::string& esrgan_path) {
-#ifdef SD_USE_CUBLAS
+#ifdef SD_USE_CUDA
         LOG_DEBUG("Using CUDA backend");
         backend = ggml_backend_cuda_init(0);
 #endif