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

Change according to the review

Browse source
This commit is contained in:
Howard Su 2023-06-27 08:17:35 +08:00
parent e67b15f5c2
commit 74fe5fc1ea
7 changed files with 14 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

2
examples/common.cpp
View file

@ -110,7 +110,7 @@ bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
invalid_param = true; invalid_param = true;
break; break;
} }
params.seed = std::stoi(argv[i]); params.seed = std::stoul(argv[i]);
} else if (arg == "-t" || arg == "--threads") { } else if (arg == "-t" || arg == "--threads") {
if (++i >= argc) { if (++i >= argc) {
invalid_param = true; invalid_param = true;

2
examples/embedding/embedding.cpp
View file

@ -24,7 +24,7 @@ int main(int argc, char ** argv) {
fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT); fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT);
if (params.seed == -1) { if (params.seed == LLAMA_DEFAULT_SEED) {
params.seed = time(NULL); params.seed = time(NULL);
} }

2
examples/main/main.cpp
View file

@ -94,7 +94,7 @@ int main(int argc, char ** argv) {
fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT); fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT);
if (params.seed == -1) { if (params.seed == LLAMA_DEFAULT_SEED) {
params.seed = time(NULL); params.seed = time(NULL);
} }

2
examples/perplexity/perplexity.cpp
View file

@ -136,7 +136,7 @@ int main(int argc, char ** argv) {
fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT); fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT);
if (params.seed == -1) { if (params.seed == LLAMA_DEFAULT_SEED) {
params.seed = time(NULL); params.seed = time(NULL);
} }

2
examples/train-text-from-scratch/train-text-from-scratch.cpp
View file

@ -3034,7 +3034,7 @@ int main(int argc, char ** argv) {
return 1; return 1;
} }
if (params.seed == -1) { if (params.seed == LLAMA_DEFAULT_SEED) {
params.seed = time(NULL); params.seed = time(NULL);
} }
printf("%s: seed: %u\n", __func__, params.seed); printf("%s: seed: %u\n", __func__, params.seed);

4
llama.cpp
View file

@ -938,7 +938,7 @@ static bool kv_cache_init(
struct llama_context_params llama_context_default_params() { struct llama_context_params llama_context_default_params() {
struct llama_context_params result = { struct llama_context_params result = {
/*.seed =*/ (unsigned int)-1, /*.seed =*/ LLAMA_DEFAULT_SEED,
/*.n_ctx =*/ 512, /*.n_ctx =*/ 512,
/*.n_batch =*/ 512, /*.n_batch =*/ 512,
/*.gpu_layers =*/ 0, /*.gpu_layers =*/ 0,
@ -2692,7 +2692,7 @@ struct llama_context * llama_new_context_with_model(
llama_context * ctx = new llama_context(*model, model->vocab); llama_context * ctx = new llama_context(*model, model->vocab);
if (params.seed < 0) { if (params.seed == LLAMA_DEFAULT_SEED) {
params.seed = time(NULL); params.seed = time(NULL);
} }

12
llama.h
View file

@ -46,6 +46,8 @@
#define LLAMA_SESSION_MAGIC LLAMA_FILE_MAGIC_GGSN #define LLAMA_SESSION_MAGIC LLAMA_FILE_MAGIC_GGSN
#define LLAMA_SESSION_VERSION 1 #define LLAMA_SESSION_VERSION 1
#define LLAMA_DEFAULT_SEED 0xFFFFFFFF
#if defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST) || defined(GGML_USE_METAL) #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. // Defined when llama.cpp is compiled with support for offloading model layers to GPU.
#define LLAMA_SUPPORTS_GPU_OFFLOAD #define LLAMA_SUPPORTS_GPU_OFFLOAD
@ -81,11 +83,11 @@ extern "C" {
typedef void (*llama_progress_callback)(float progress, void *ctx); typedef void (*llama_progress_callback)(float progress, void *ctx);
struct llama_context_params { struct llama_context_params {
unsigned int seed; // RNG seed, -1 for random uint32_t seed; // RNG seed, -1 for random
int n_ctx; // text context int32_t n_ctx; // text context
int n_batch; // prompt processing batch size int32_t n_batch; // prompt processing batch size
int n_gpu_layers; // number of layers to store in VRAM int32_t n_gpu_layers; // number of layers to store in VRAM
int main_gpu; // the GPU that is used for scratch and small tensors 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 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 // called with a progress value between 0 and 1, pass NULL to disable
llama_progress_callback progress_callback; llama_progress_callback progress_callback;