ggml : speed-up soft max via Accelerate + unroll

This commit is contained in:
Georgi Gerganov
2023-01-07 16:11:41 +02:00
parent d51fc3ee0a
commit d61d55cd4b
2 changed files with 114 additions and 54 deletions

View File

@ -131,7 +131,7 @@ static const std::map<std::string, std::pair<int, std::string>> g_lang = {
{ "su", { 98, "sundanese", } },
};
static const size_t MB = 1024*1024;
static const size_t MB = 3*1024*1024;
static const std::map<e_model, size_t> MEM_REQ_MODEL = {
{ MODEL_TINY, 74ull*MB },