diff options
| author | Kawrakow <iwankawrakow@gmail.com> | 2024-12-03 06:15:29 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-12-03 06:15:29 +0100 |
| commit | ccec00939a30aa7762a232ac4dcadba985ef9ee4 (patch) | |
| tree | b308bd1de2b982c37fc7542374fb77efa2abb726 /include/llama.h | |
| parent | 239a344f9935cb614c6208000edb0815214286a1 (diff) | |
Q8_0_R4 (#120)
* Adding q8_0_r4
We get PP-512(LLaMA-3.1-8B) = 268 t/s on a Ryzen-7950X compared
to 175.6 t/s for Q8_0.
* q8_0_r4: NEON
We get PP-512(LLaMA-3.1-8B) = 112.6 t/s on M2-Max.
* q8_0_r4: Zen4 matrix-vector specialization
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'include/llama.h')
| -rw-r--r-- | include/llama.h | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/include/llama.h b/include/llama.h index fc034b3a..5e935533 100644 --- a/include/llama.h +++ b/include/llama.h @@ -181,6 +181,7 @@ extern "C" { LLAMA_FTYPE_MOSTLY_IQ4_KSS = 148, // except 1d tensors // LLAMA_FTYPE_MOSTLY_Q4_0_R4 = 202, // except 1d tensors + LLAMA_FTYPE_MOSTLY_Q8_0_R4 = 207, // except 1d tensors LLAMA_FTYPE_MOSTLY_IQ4_NL_X4 = 225, // except 1d tensors LLAMA_FTYPE_GUESSED = 1024, // not specified in the model file |