mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-10-30 08:42:00 +00:00
a113689571
* ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
23 lines
653 B
C
23 lines
653 B
C
#pragma once
|
|
|
|
#include "ggml.h"
|
|
|
|
#ifdef __cplusplus
|
|
extern "C" {
|
|
#endif
|
|
|
|
|
|
GGML_API struct ggml_allocr * ggml_allocr_new(void * data, size_t size, size_t alignment);
|
|
GGML_API struct ggml_allocr * ggml_allocr_new_measure(size_t alignment);
|
|
|
|
GGML_API void ggml_allocr_free(struct ggml_allocr * alloc);
|
|
GGML_API bool ggml_allocr_is_measure(struct ggml_allocr * alloc);
|
|
GGML_API void ggml_allocr_reset(struct ggml_allocr * alloc);
|
|
GGML_API void ggml_allocr_alloc(struct ggml_allocr * alloc, struct ggml_tensor * tensor);
|
|
GGML_API size_t ggml_allocr_alloc_graph(struct ggml_allocr * alloc, struct ggml_cgraph * graph);
|
|
|
|
|
|
#ifdef __cplusplus
|
|
}
|
|
#endif
|