From 3d92d7f802b332927669f01bfa51ebbb56e868ba Mon Sep 17 00:00:00 2001 From: Kawrakow Date: Thu, 15 May 2025 16:02:39 +0300 Subject: 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 --- include/llama.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/include/llama.h b/include/llama.h index 0f3ae862..98b08bbd 100644 --- a/include/llama.h +++ b/include/llama.h @@ -193,6 +193,7 @@ extern "C" { LLAMA_FTYPE_MOSTLY_IQ2_KS = 147, // except 1d tensors LLAMA_FTYPE_MOSTLY_IQ4_KSS = 148, // except 1d tensors LLAMA_FTYPE_MOSTLY_Q8_KV = 149, // except 1d tensors + LLAMA_FTYPE_MOSTLY_IQ5_KS = 150, // except 1d tensors // LLAMA_FTYPE_MOSTLY_Q4_0_R8 = 202, // except 1d tensors LLAMA_FTYPE_MOSTLY_Q8_0_R8 = 207, // except 1d tensors @@ -231,7 +232,7 @@ extern "C" { LLAMA_ROPE_SCALING_TYPE_LINEAR = 1, LLAMA_ROPE_SCALING_TYPE_YARN = 2, LLAMA_ROPE_SCALING_TYPE_LONGROPE = 3, - LLAMA_ROPE_SCALING_TYPE_MAX_VALUE = LLAMA_ROPE_SCALING_TYPE_LONGROPE, + LLAMA_ROPE_SCALING_TYPE_MAX_VALUE = LLAMA_ROPE_SCALING_TYPE_LONGROPE, }; enum llama_pooling_type { -- cgit v1.2.3