upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-07-23 11:10:55 -05:00
Code Issues Packages Projects Releases Wiki Activity

ggml: gguf_init_from_callback and gguf_init_from_buffer (#22341)

Browse Source 
* ggml: implement `gguf_init_from_buffer`

* test: `gguf_init_from_buffer`

* fix: memory breakdown for a model loaded with `no_alloc` from a file is consistent with being loaded from a buffer

* fix: use `GGML_UNUSED`

Co-authored-by: Copilot <copilot@github.com>

* fix: remove `total_size` from `gguf_reader`

* fix: file offset calculation, rename `offset` to `data_offset`

Co-authored-by: Copilot <copilot@github.com>

* refactor: extract model loader bug fixes to another PR

* feat: add `gguf_init_from_callback`

* fix: always require a max expected size

* fix: change `gguf_reader_callback_t`'s `output` type to `void *`, change `max_expected_size` and offsets to `uint64_t`

* fix: harden against offset overflow in buffer read

* fix: remove seek behavior from the callback

* feat: `max_chunk_read == 0` means `SIZE_MAX`

* fix: seeking in a gguf file with no tensors

---------

Co-authored-by: Copilot <copilot@github.com>
This commit is contained in:
Gilad S.
2026-05-25 11:33:29 +02:00
committed by GitHub
parent 6c4cbdc70b
commit 66efd13375
3 changed files with 248 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
ggml/include/gguf.h
View File

@@ -76,10 +76,16 @@ extern "C" {
struct ggml_context ** ctx;
};
// callback to simulate or wrap a FILE pointer - read up to `len` bytes at `offset` into `output` and return the number of bytes read
typedef size_t (*gguf_reader_callback_t)(void * userdata, void * output, uint64_t offset, size_t len);
GGML_API struct gguf_context * gguf_init_empty(void);
GGML_API struct gguf_context * gguf_init_from_file_ptr(FILE * file, struct gguf_init_params params);
GGML_API struct gguf_context * gguf_init_from_file(const char * fname, struct gguf_init_params params);
//GGML_API struct gguf_context * gguf_init_from_buffer(..);
GGML_API struct gguf_context * gguf_init_from_buffer(const void * data, size_t size, struct gguf_init_params params);
// max_chunk_read is the maximum number of bytes that the GGUF code will read at once from the callback, a value of 0 means no limit
GGML_API struct gguf_context * gguf_init_from_callback(gguf_reader_callback_t callback, void * userdata, size_t max_chunk_read, uint64_t max_expected_size, struct gguf_init_params params);
GGML_API void gguf_free(struct gguf_context * ctx);
@@ -87,7 +93,7 @@ extern "C" {
GGML_API uint32_t gguf_get_version (const struct gguf_context * ctx);
GGML_API size_t gguf_get_alignment (const struct gguf_context * ctx);
GGML_API size_t gguf_get_data_offset(const struct gguf_context * ctx);
GGML_API size_t gguf_get_data_offset(const struct gguf_context * ctx); // padded to gguf_get_alignment if and only if the gguf_context contains at least one tensor
GGML_API int64_t gguf_get_n_kv(const struct gguf_context * ctx);
GGML_API int64_t gguf_find_key(const struct gguf_context * ctx, const char * key); // returns -1 if key is not found