diff options
| author | Iwan Kawrakow <iwan.kawrakow@gmail.com> | 2024-07-16 09:12:15 +0200 |
|---|---|---|
| committer | Iwan Kawrakow <iwan.kawrakow@gmail.com> | 2024-07-16 09:12:15 +0200 |
| commit | 52a25e307c3af8686436d977c60e9975b0900e2b (patch) | |
| tree | 17240797a176f63bf759d0a68410dc53a6c5913d /examples/llava/llava-cli.cpp | |
| parent | 6393e2682720893092f77c2a6d428a2c13ecccf7 (diff) | |
iq1bn(no lookup): Metal
In summary, compared to lookup, the multiplication based approach is
* Much better on AVX2
* Slightly better on CUDA
* Slightly worse on Metal
* Much worse on NEON
Diffstat (limited to 'examples/llava/llava-cli.cpp')
0 files changed, 0 insertions, 0 deletions