From: Georgi Gerganov Date: Fri, 5 Jan 2024 13:36:04 +0000 (+0200) Subject: ggml : fix q2_k bpw in comments (ggml/680) X-Git-Tag: upstream/1.7.4~1168 X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=66d8f0b7f1a288562682656c020ff9c4835c2c52;p=pkg%2Fggml%2Fsources%2Fwhisper.cpp ggml : fix q2_k bpw in comments (ggml/680) --- 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