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

Use unsigned for random seed (#2006)

Browse source 
* Use unsigned for random seed. Keep -1 as the value to use a time based seed.

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Howard Su 2023-06-29 21:15:15 +08:00 • committed by GitHub
parent 96a712ca1b
commit b8c8dda75f
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
10 changed files with 25 additions and 23 deletions
Download patch file Download diff file Expand all files Collapse all files

14
llama.h
View file

@ -46,6 +46,8 @@
#define LLAMA_SESSION_MAGIC LLAMA_FILE_MAGIC_GGSN
#define LLAMA_SESSION_VERSION 1
#define LLAMA_DEFAULT_SEED 0xFFFFFFFF
#if defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST) || defined(GGML_USE_METAL)
// Defined when llama.cpp is compiled with support for offloading model layers to GPU.
#define LLAMA_SUPPORTS_GPU_OFFLOAD
@ -81,11 +83,11 @@ extern "C" {
typedef void (*llama_progress_callback)(float progress, void *ctx);
struct llama_context_params {
int seed; // RNG seed, -1 for random
int n_ctx; // text context
int n_batch; // prompt processing batch size
int n_gpu_layers; // number of layers to store in VRAM
int main_gpu; // the GPU that is used for scratch and small tensors
uint32_t seed; // RNG seed, -1 for random
int32_t n_ctx; // text context
int32_t n_batch; // prompt processing batch size
int32_t n_gpu_layers; // number of layers to store in VRAM
int32_t main_gpu; // the GPU that is used for scratch and small tensors
float tensor_split[LLAMA_MAX_DEVICES]; // how to split layers across multiple GPUs
// called with a progress value between 0 and 1, pass NULL to disable
llama_progress_callback progress_callback;
@ -196,7 +198,7 @@ extern "C" {
LLAMA_API int llama_get_kv_cache_token_count(const struct llama_context * ctx);
// Sets the current rng seed.
LLAMA_API void llama_set_rng_seed(struct llama_context * ctx, int seed);
LLAMA_API void llama_set_rng_seed(struct llama_context * ctx, uint32_t seed);
// Returns the maximum size in bytes of the state (rng, logits, embedding
// and kv_cache) - will often be smaller after compacting tokens