vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llama : fix session load / save (#1263)

Browse source
This commit is contained in:
Georgi Gerganov 2023-05-01 14:54:59 +03:00 • committed by GitHub
parent b925f1f1b0
commit 70269cae37
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 96 additions and 69 deletions
Download patch file Download diff file Expand all files Collapse all files

12
llama.h
View file

@ -19,9 +19,11 @@
# define LLAMA_API
#endif
#define LLAMA_FILE_VERSION 1
#define LLAMA_FILE_MAGIC 0x67676a74 // 'ggjt' in hex
#define LLAMA_FILE_MAGIC_UNVERSIONED 0x67676d6c // pre-versioned files
#define LLAMA_FILE_VERSION 1
#define LLAMA_FILE_MAGIC 'ggjt'
#define LLAMA_FILE_MAGIC_UNVERSIONED 'ggml'
#define LLAMA_SESSION_MAGIC 'ggsn'
#define LLAMA_SESSION_VERSION 0
#ifdef __cplusplus
extern "C" {
@ -138,8 +140,8 @@ extern "C" {
LLAMA_API size_t llama_set_state_data(struct llama_context * ctx, const uint8_t * src);
// Save/load session file
LLAMA_API size_t llama_load_session_file(struct llama_context * ctx, const char * path_session, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out);
LLAMA_API size_t llama_save_session_file(struct llama_context * ctx, const char * path_session, const llama_token * tokens, size_t n_token_count);
LLAMA_API bool llama_load_session_file(struct llama_context * ctx, const char * path_session, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out);
LLAMA_API bool llama_save_session_file(struct llama_context * ctx, const char * path_session, const llama_token * tokens, size_t n_token_count);
// Run the llama inference to obtain the logits and probabilities for the next token.
// tokens + n_tokens is the provided batch of new tokens to process