diff options
| author | Georgi Gerganov <ggerganov@gmail.com> | 2024-01-05 15:36:04 +0200 |
|---|---|---|
| committer | Georgi Gerganov <ggerganov@gmail.com> | 2024-01-05 18:02:06 +0200 |
| commit | d061bf9405cc5fd50792fb2dbdff9c9ea53d6bf9 (patch) | |
| tree | 7299ebda18ee51c3bd318f30164a4db1ea74026d /ggml-quants.h | |
| parent | 1bf681f90ef4cf37b36e6d604d3e30fc57eda650 (diff) | |
ggml : fix q2_k bpw in comments (ggml/680)
Diffstat (limited to 'ggml-quants.h')
| -rw-r--r-- | ggml-quants.h | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/ggml-quants.h b/ggml-quants.h index 70c12c27..62c1df6c 100644 --- a/ggml-quants.h +++ b/ggml-quants.h @@ -70,7 +70,7 @@ static_assert(sizeof(block_q8_1) == 2*sizeof(float) + QK8_1, "wrong q8_1 block s // 2-bit quantization // weight is represented as x = a * q + b // 16 blocks of 16 elements each -// Effectively 2.5625 bits per weight +// Effectively 2.625 bits per weight typedef struct { uint8_t scales[QK_K/16]; // scales and mins, quantized with 4 bits uint8_t qs[QK_K/4]; // quants |