ollama/llama/patches/01-cuda.diff

diff --git a/llama/ggml-backend.c b/llama/ggml-backend.c
index 9e35ce98..179be840 100644
--- a/llama/ggml-backend.c
+++ b/llama/ggml-backend.c
@@ -87,7 +87,12 @@ void ggml_backend_buffer_free(ggml_backend_buffer_t buffer) {
     if (buffer->iface.free_buffer != NULL) {
         buffer->iface.free_buffer(buffer);
     }
+
+// TODO: this needs to be freed in cuda and hipblas backends because
+// the cuda backend implementation compiled with msvc
+#if !defined(GGML_USE_CUDA) && !defined(GGML_USE_HIPBLAS)
     free(buffer);
+#endif
 }
 
 size_t ggml_backend_buffer_get_size(ggml_backend_buffer_t buffer) {
diff --git a/llama/ggml-cuda.cu b/llama/ggml-cuda.cu
index 04b6e528..43b12bdf 100644
--- a/llama/ggml-cuda.cu
+++ b/llama/ggml-cuda.cu
@@ -392,6 +392,10 @@ GGML_CALL static bool ggml_backend_buffer_is_cuda(ggml_backend_buffer_t buffer)
 GGML_CALL static void ggml_backend_cuda_buffer_free_buffer(ggml_backend_buffer_t buffer) {
     ggml_backend_cuda_buffer_context * ctx = (ggml_backend_cuda_buffer_context *)buffer->context;
     delete ctx;
+
+    // TODO: this needs to be freed in cuda and hipblas backends because
+    // the cuda backend implementation compiled with msvc
+    free(buffer);
 }
 
 GGML_CALL static void * ggml_backend_cuda_buffer_get_base(ggml_backend_buffer_t buffer) {
@@ -3028,8 +3032,6 @@ GGML_CALL static ggml_backend_t ggml_backend_reg_cuda_init(const char * params,
     GGML_UNUSED(params);
 }
 
-extern "C" GGML_CALL int ggml_backend_cuda_reg_devices();
-
 GGML_CALL int ggml_backend_cuda_reg_devices() {
     int device_count = ggml_backend_cuda_get_device_count();
     //int device_count = 1; // DEBUG: some tools require delaying CUDA initialization
diff --git a/llama/ggml-cuda.h b/llama/ggml-cuda.h
index 5eb4af40..50b91009 100644
--- a/llama/ggml-cuda.h
+++ b/llama/ggml-cuda.h
@@ -31,6 +31,8 @@ GGML_API GGML_CALL ggml_backend_buffer_type_t ggml_backend_cuda_split_buffer_typ
 // pinned host buffer for use with the CPU backend for faster copies between CPU and GPU
 GGML_API GGML_CALL ggml_backend_buffer_type_t ggml_backend_cuda_host_buffer_type(void);
 
+GGML_API GGML_CALL int ggml_backend_cuda_reg_devices();
+
 GGML_API GGML_CALL int  ggml_backend_cuda_get_device_count(void);
 GGML_API GGML_CALL void ggml_backend_cuda_get_device_description(int device, char * description, size_t description_size);
 GGML_API GGML_CALL void ggml_backend_cuda_get_device_memory(int device, size_t * free, size_t * total);