diff options
| author | Kawrakow <iwankawrakow@gmail.com> | 2025-05-15 16:02:39 +0300 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-05-15 16:02:39 +0300 |
| commit | 3d92d7f802b332927669f01bfa51ebbb56e868ba (patch) | |
| tree | c3913f67e36492c723cc47fe512078ee0dd19d59 /ggml/src/ggml-common.h | |
| parent | 3f8c865b920df844ba0cb4ba53c1ccce8874b045 (diff) | |
Adding IQ5_KS - 5.25 bpw quants (#422)
* iq5_ks: basics
* iq5_ks: quantize
* iq5_ks: CUDA dequantize works
* iq5_ks: dot product works on CUDA
* iq5_ks: MMQ works
* iq5_ks: Zen4
* iq5_ks: AVX2
But is is not quite right, just like iq4_k, iq5_k, iq6_k, iq4_ks.
All these need fixing on AVX2.
* iq5_ks: NEON
* iq5_ks: Metal dequantize
* iq5_ks: Metal dot product
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'ggml/src/ggml-common.h')
| -rw-r--r-- | ggml/src/ggml-common.h | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/ggml/src/ggml-common.h b/ggml/src/ggml-common.h index 5a6417fc..1c2d1b17 100644 --- a/ggml/src/ggml-common.h +++ b/ggml/src/ggml-common.h @@ -687,6 +687,13 @@ typedef struct { } block_iq6_k; static_assert(sizeof(block_iq6_k) == sizeof(ggml_half) + sizeof(uint16_t) + QK_K/2 + QK_K/4 + QK_K/16, "wrong iq6_k block size/padding"); +typedef struct { + uint8_t scales[QK_K/32]; + uint8_t qs[QK_K/2]; + uint8_t qh[QK_K/8]; +} block_iq5_ks; +static_assert(sizeof(block_iq5_ks) == QK_K/32 + QK_K/2 + QK_K/8, "wrong iq5_ks block size/padding"); + #endif // GGML_COMMON_DECL #endif // GGML_COMMON_DECL |