unified ggml_numa_strategy enum and fixed text alignment in server.cpp example

common/common.h

@@ -76,7 +76,7 @@ struct gpt_params {
     float   yarn_beta_slow        = 1.0f;  // YaRN high correction dim
     int32_t yarn_orig_ctx         = 0;     // YaRN original context length
     int32_t rope_scaling_type     = LLAMA_ROPE_SCALING_UNSPECIFIED;
-    ggml_numa_strategies numa     = GGML_NUMA_STRATEGY_DISABLED;
+    ggml_numa_strategy numa       = GGML_NUMA_STRATEGY_DISABLED;
 
     // // sampling parameters
     struct llama_sampling_params sparams;

examples/server/server.cpp

@@ -1855,10 +1855,10 @@ static void server_print_usage(const char *argv0, const gpt_params &params,
     {
         printf("  --no-mmap                 do not memory-map model (slower load but may reduce pageouts if not using mlock)\n");
     }
-    printf("  --numa TYPE           attempt optimizations that help on some NUMA systems\n");
+    printf("  --numa TYPE               attempt optimizations that help on some NUMA systems\n");
-    printf("                          - distribute: spread execution evenly over all nodes\n");
+    printf("                              - distribute: spread execution evenly over all nodes\n");
-    printf("                          - isolate: only spawn threads on CPUs on the node that execution started on\n");
+    printf("                              - isolate: only spawn threads on CPUs on the node that execution started on\n");
-    printf("                          - numactl: use the CPU map provided my numactl\n");
+    printf("                              - numactl: use the CPU map provided my numactl\n");
     if (llama_supports_gpu_offload()) {
         printf("  -ngl N, --n-gpu-layers N\n");
         printf("                            number of layers to store in VRAM\n");

ggml.c

@@ -1954,7 +1954,7 @@ struct ggml_numa_node {
 };
 
 struct ggml_numa_nodes {
-    enum ggml_numa_strategies numa_strategy;
+    enum ggml_numa_strategy numa_strategy;
     struct ggml_numa_node nodes[GGML_NUMA_MAX_NODES];
     uint32_t n_nodes;
     uint32_t total_cpus; // hardware threads on system
@@ -2013,7 +2013,7 @@ static uint32_t ggml_get_numa_affinity(void) {
 }
 #endif
 
-void ggml_numa_init(enum ggml_numa_strategies numa_flag) {
+void ggml_numa_init(enum ggml_numa_strategy numa_flag) {
     if (g_state.numa.n_nodes > 0) {
         fprintf(stderr, "ggml_numa_init: NUMA already initialized\n");
 

ggml.h

@@ -678,7 +678,7 @@ extern "C" {
 
     GGML_API void    ggml_print_backtrace(void);
 
-    GGML_API void    ggml_numa_init(enum ggml_numa_strategies numa); // call once for better performance on NUMA systems
+    GGML_API void    ggml_numa_init(enum ggml_numa_strategy numa); // call once for better performance on NUMA systems
     GGML_API bool    ggml_is_numa(void); // true if init detected that system has >1 NUMA node
 
     GGML_API void    ggml_print_object (const struct ggml_object * obj);

llama.cpp

@@ -11171,7 +11171,7 @@ void llama_backend_init(void) {
 #endif
 }
 
-void llama_numa_init(enum ggml_numa_strategies numa) {
+void llama_numa_init(enum ggml_numa_strategy numa) {
     if (numa != GGML_NUMA_STRATEGY_DISABLED) {
         ggml_numa_init(numa);
     }

llama.h

@@ -309,7 +309,7 @@ extern "C" {
     LLAMA_API void llama_backend_init(void);
 
     //optional:
-    LLAMA_API void llama_numa_init(enum ggml_numa_strategies numa);
+    LLAMA_API void llama_numa_init(enum ggml_numa_strategy numa);
 
     // Call once at the end of the program - currently only used for MPI
     LLAMA_API void llama_backend_free(void);