aboutsummaryrefslogtreecommitdiff
path: root/llama.h
diff options
context:
space:
mode:
authorGeorgi Gerganov <ggerganov@gmail.com>2023-05-01 14:54:59 +0300
committerGitHub <noreply@github.com>2023-05-01 14:54:59 +0300
commit70269cae37538461ff816e714afbb3ebcdcdc26b (patch)
tree448ee5f5f8e93816fc7e60c7b05b29668536884a /llama.h
parentb925f1f1b082319ee69943f8d1a83ac9b6ff09ca (diff)
llama : fix session load / save (#1263)
Diffstat (limited to 'llama.h')
-rw-r--r--llama.h12
1 files changed, 7 insertions, 5 deletions
diff --git a/llama.h b/llama.h
index 9fbba76..2f6ce8d 100644
--- a/llama.h
+++ b/llama.h
@@ -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