aditya/llama.cpp
1
0
Fork 0
mirror of https://git.adityakumar.xyz/llama.cpp.git synced 2024-11-09 15:29:43 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

ggml : add ggml_tensor_overhead()

Browse source
This commit is contained in:
Georgi Gerganov 2023-05-27 16:19:56 +03:00
parent 83c54e6da5
commit 93618031c7
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
2 changed files with 15 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

12
ggml.c
View file

@ -3808,6 +3808,10 @@ enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype) {
return wtype; return wtype;
} }
size_t ggml_tensor_overhead(void) {
return GGML_OBJECT_SIZE + GGML_TENSOR_SIZE + 16;
}
static inline bool ggml_is_transposed(const struct ggml_tensor * tensor) { static inline bool ggml_is_transposed(const struct ggml_tensor * tensor) {
return tensor->nb[0] > tensor->nb[1]; return tensor->nb[0] > tensor->nb[1];
} }
@ -14527,6 +14531,14 @@ void ggml_graph_reset(struct ggml_cgraph * cgraph) {
} }
struct ggml_tensor * ggml_get_tensor_by_name(struct ggml_cgraph * cgraph, const char * name) { struct ggml_tensor * ggml_get_tensor_by_name(struct ggml_cgraph * cgraph, const char * name) {
for (int i = 0; i < cgraph->n_leafs; i++) {
struct ggml_tensor * leaf = cgraph->leafs[i];
if (strcmp(leaf->name, name) == 0) {
return leaf;
}
}
for (int i = 0; i < cgraph->n_nodes; i++) { for (int i = 0; i < cgraph->n_nodes; i++) {
struct ggml_tensor * node = cgraph->nodes[i]; struct ggml_tensor * node = cgraph->nodes[i];

6
ggml.h
View file

@ -380,9 +380,6 @@ extern "C" {
static const size_t GGML_TENSOR_SIZE = sizeof(struct ggml_tensor); static const size_t GGML_TENSOR_SIZE = sizeof(struct ggml_tensor);
// use this to compute the memory overhead of a tensor
static const size_t GGML_TENSOR_OVERHEAD = (GGML_OBJECT_SIZE + GGML_TENSOR_SIZE + 16);
// computation graph // computation graph
struct ggml_cgraph { struct ggml_cgraph {
int n_nodes; int n_nodes;
@ -444,6 +441,9 @@ extern "C" {
// TODO: temporary until model loading of ggml examples is refactored // TODO: temporary until model loading of ggml examples is refactored
GGML_API enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype); GGML_API enum ggml_type ggml_ftype_to_ggml_type(enum ggml_ftype ftype);
// use this to compute the memory overhead of a tensor
GGML_API size_t ggml_tensor_overhead(void);
// main // main
GGML_API struct ggml_context * ggml_init(struct ggml_init_params params); GGML_API struct ggml_context * ggml_init(struct ggml_init_params params);