wip

Author: slaren

Makefile

@@ -65,7 +65,7 @@ test: $(TEST_TARGETS)
 			./$$test_target; \
 		fi; \
 		if [ $$? -ne 0 ]; then \
-			printf 'Test $$test_target FAILED!\n\n' $$test_target; \
+			printf 'Test %s FAILED!\n\n' $$test_target; \
 			failures=$$(( failures + 1 )); \
 		else \
 			printf 'Test %s passed.\n\n' $$test_target; \

ggml-alloc.c

@@ -736,6 +736,10 @@ void ggml_allocr_set_parse_seq(ggml_allocr_t alloc, const int * list, int n) {
 }
 
 void ggml_allocr_free(ggml_allocr_t alloc) {
+    if (alloc == NULL) {
+        return;
+    }
+
     ggml_gallocr_free(alloc->galloc);
     ggml_tallocr_free(alloc->talloc);
     free(alloc);

ggml-backend.c

@@ -505,7 +505,7 @@ static ggml_backend_graph_plan_t ggml_backend_cpu_graph_plan_create(ggml_backend
     struct ggml_backend_plan_cpu * cpu_plan = malloc(sizeof(struct ggml_backend_plan_cpu));
 
     cpu_plan->cplan = ggml_graph_plan(cgraph, cpu_ctx->n_threads);
-    cpu_plan->cgraph = *cgraph;
+    cpu_plan->cgraph = *cgraph; // FIXME: deep copy
 
     if (cpu_plan->cplan.work_size > 0) {
         cpu_plan->cplan.work_data = malloc(cpu_plan->cplan.work_size);

ggml-cuda.cu

@@ -7057,6 +7057,7 @@ inline void ggml_cuda_op_upscale(
 
     (void) src1;
     (void) dst;
+    (void) src1_dd;
 }
 
 inline void ggml_cuda_op_pad(
@@ -7073,6 +7074,7 @@ inline void ggml_cuda_op_pad(
 
     (void) src1;
     (void) dst;
+    (void) src1_dd;
 }
 
 inline void ggml_cuda_op_rms_norm(
@@ -8958,7 +8960,7 @@ void ggml_cuda_transform_tensor(void * data, struct ggml_tensor * tensor) {
 
         char * buf;
         CUDA_CHECK(cudaMalloc(&buf, size));
-        char * buf_host = (char*)data + offset_split;
+        char * buf_host = (char *)data + offset_split;
 
         // set padding to 0 to avoid possible NaN values
         if (size > original_size) {

ggml.c

@@ -2383,20 +2383,8 @@ size_t ggml_get_mem_size(const struct ggml_context * ctx) {
 size_t ggml_get_max_tensor_size(const struct ggml_context * ctx) {
     size_t max_size = 0;
 
-    struct ggml_object * obj = ctx->objects_begin;
-
-    while (obj != NULL) {
-        if (obj->type == GGML_OBJECT_TENSOR) {
-            struct ggml_tensor * tensor = (struct ggml_tensor *) ((char *) ctx->mem_buffer + obj->offs);
-
-            const size_t size = ggml_nbytes(tensor);
-
-            if (max_size < size) {
-                max_size = size;
-            }
-        }
-
-        obj = obj->next;
+    for (struct ggml_tensor * tensor = ggml_get_first_tensor(ctx); tensor != NULL; tensor = ggml_get_next_tensor(ctx, tensor)) {
+        max_size = MAX(max_size, ggml_nbytes(tensor));
     }
 
     return max_size;
@@ -3093,7 +3081,7 @@ struct ggml_tensor * ggml_view_tensor(
     return result;
 }
 
-struct ggml_tensor * ggml_get_first_tensor(struct ggml_context * ctx) {
+struct ggml_tensor * ggml_get_first_tensor(const struct ggml_context * ctx) {
     struct ggml_object * obj = ctx->objects_begin;
 
     char * const mem_buffer = ctx->mem_buffer;
@@ -3109,7 +3097,7 @@ struct ggml_tensor * ggml_get_first_tensor(struct ggml_context * ctx) {
     return NULL;
 }
 
-struct ggml_tensor * ggml_get_next_tensor(struct ggml_context * ctx, struct ggml_tensor * tensor) {
+struct ggml_tensor * ggml_get_next_tensor(const struct ggml_context * ctx, struct ggml_tensor * tensor) {
     struct ggml_object * obj = (struct ggml_object *) ((char *)tensor - GGML_OBJECT_SIZE);
     obj = obj->next;
 
@@ -19179,6 +19167,10 @@ char * gguf_get_tensor_name(const struct gguf_context * ctx, int i) {
     return ctx->infos[i].name.data;
 }
 
+enum ggml_type gguf_get_tensor_type(const struct gguf_context * ctx, int i) {
+    return ctx->infos[i].type;
+}
+
 // returns the index
 static int gguf_get_or_add_key(struct gguf_context * ctx, const char * key) {
     const int idx = gguf_find_key(ctx, key);

ggml.h

@@ -729,8 +729,8 @@ extern "C" {
     GGML_API struct ggml_tensor * ggml_view_tensor(struct ggml_context * ctx, struct ggml_tensor * src);
 
     // Context tensor enumeration and lookup
-    GGML_API struct ggml_tensor * ggml_get_first_tensor(struct ggml_context * ctx);
-    GGML_API struct ggml_tensor * ggml_get_next_tensor (struct ggml_context * ctx, struct ggml_tensor * tensor);
+    GGML_API struct ggml_tensor * ggml_get_first_tensor(const struct ggml_context * ctx);
+    GGML_API struct ggml_tensor * ggml_get_next_tensor (const struct ggml_context * ctx, struct ggml_tensor * tensor);
     GGML_API struct ggml_tensor * ggml_get_tensor(struct ggml_context * ctx, const char * name);
 
     GGML_API struct ggml_tensor * ggml_set_zero(struct ggml_tensor * tensor);
@@ -2123,10 +2123,11 @@ extern "C" {
     GGML_API const void * gguf_get_arr_data(const struct gguf_context * ctx, int key_id);
     GGML_API const char * gguf_get_arr_str (const struct gguf_context * ctx, int key_id, int i);
 
-    GGML_API int    gguf_get_n_tensors    (const struct gguf_context * ctx);
-    GGML_API int    gguf_find_tensor      (const struct gguf_context * ctx, const char * name);
-    GGML_API size_t gguf_get_tensor_offset(const struct gguf_context * ctx, int i);
-    GGML_API char * gguf_get_tensor_name  (const struct gguf_context * ctx, int i);
+    GGML_API int            gguf_get_n_tensors    (const struct gguf_context * ctx);
+    GGML_API int            gguf_find_tensor      (const struct gguf_context * ctx, const char * name);
+    GGML_API size_t         gguf_get_tensor_offset(const struct gguf_context * ctx, int i);
+    GGML_API char *         gguf_get_tensor_name  (const struct gguf_context * ctx, int i);
+    GGML_API enum ggml_type gguf_get_tensor_type  (const struct gguf_context * ctx, int i);
 
     // overrides existing values or adds a new one
     GGML_API void gguf_set_val_u8  (struct gguf_context * ctx, const char * key, uint8_t  val);