alihan/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2023-11-04 02:52:44 +03:00
Code Issues Packages Projects Releases Wiki Activity

Use Accelerate framework on Apple silicon

Browse Source 
Huge performance improvement in the Encode (almost x2 on MacBook M1 Pro)

Also various extra optimizations:

- Multi-threaded NORM operator
- Faster GELU via F16 cast
This commit is contained in:
Georgi Gerganov
2022-10-17 21:44:16 +03:00
parent 130b5c02d6
commit 72d967bce4
5 changed files with 217 additions and 121 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
whisper.cpp
View File

@@ -15,7 +15,7 @@
#include <vector>
#define USE_FLASH_ATTN
#define USE_FLASH_FF
//#define USE_FLASH_FF
// available whisper models
enum e_model {
@@ -148,11 +148,11 @@ static const std::map<e_model, size_t> MEM_REQ_ENCODE = {
};
static const std::map<e_model, size_t> MEM_REQ_ENCODE_LAYER = {
{ MODEL_TINY, 64ull*MB },
{ MODEL_BASE, 84ull*MB },
{ MODEL_SMALL, 128ull*MB },
{ MODEL_MEDIUM, 172ull*MB },
{ MODEL_LARGE, 216ull*MB },
{ MODEL_TINY, 104ull*MB },
{ MODEL_BASE, 138ull*MB },
{ MODEL_SMALL, 208ull*MB },
{ MODEL_MEDIUM, 280ull*MB },
{ MODEL_LARGE, 354ull*MB },
};
static const std::map<e_model, size_t> MEM_REQ_DECODE = {