Merge branch 'ggerganov:master' into master

Author: MaggotHATE

ggml/include/ggml-alloc.h

@@ -24,7 +24,7 @@ GGML_API void                ggml_tallocr_alloc(struct ggml_tallocr * talloc, st
 // Graph allocator
 /*
   Example usage:
-    ggml_gallocr_t galloc = ggml_gallocr_new(ggml_bacckend_cpu_buffer_type());
+    ggml_gallocr_t galloc = ggml_gallocr_new(ggml_backend_cpu_buffer_type());
 
     // optional: create a worst-case graph and reserve the buffers to avoid reallocations
     ggml_gallocr_reserve(galloc, build_graph(max_batch));

ggml/src/ggml-cuda.cu

@@ -2448,6 +2448,7 @@ static void ggml_backend_cuda_synchronize(ggml_backend_t backend) {
     GGML_UNUSED(backend);
 }
 
+#ifdef USE_CUDA_GRAPH
 static void set_ggml_graph_node_properties(ggml_tensor * node, ggml_graph_node_properties * graph_node_properties) {
     graph_node_properties->node_address = node->data;
     graph_node_properties->node_op = node->op;
@@ -2498,6 +2499,7 @@ static bool ggml_graph_node_has_matching_properties(ggml_tensor * node, ggml_gra
 
     return true;
 }
+#endif
 
 static enum ggml_status ggml_backend_cuda_graph_compute(ggml_backend_t backend, ggml_cgraph * cgraph) {
     ggml_backend_cuda_context * cuda_ctx = (ggml_backend_cuda_context *)backend->context;

scripts/sync-ggml.last

@@ -1 +1 @@
-e5c233e5edbfcfa1d808b9293de9065035c40751
+b77f48b1efa671e094696b99fbf566aac8c87d74