|
|
@ -14,11 +14,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
struct gpt_params {
|
|
|
|
struct gpt_params {
|
|
|
|
int32_t seed = -1; // RNG seed
|
|
|
|
int32_t seed = -1; // RNG seed
|
|
|
|
int32_t n_threads;
|
|
|
|
int32_t n_threads = std::min(4, (int32_t) std::thread::hardware_concurrency());
|
|
|
|
int32_t n_predict = 128; // new tokens to predict
|
|
|
|
int32_t n_predict = 128; // new tokens to predict
|
|
|
|
int32_t repeat_last_n = 64; // last n tokens to penalize
|
|
|
|
int32_t repeat_last_n = 64; // last n tokens to penalize
|
|
|
|
int32_t n_ctx = 512; //context size
|
|
|
|
int32_t n_ctx = 512; //context size
|
|
|
|
|
|
|
|
|
|
|
|
// sampling parameters
|
|
|
|
// sampling parameters
|
|
|
|
int32_t top_k = 40;
|
|
|
|
int32_t top_k = 40;
|
|
|
|
float top_p = 0.95f;
|
|
|
|
float top_p = 0.95f;
|
|
|
|