From f64de08203aaee95ca755336de3e1db85d990198 Mon Sep 17 00:00:00 2001 From: Kawrakow Date: Wed, 4 Dec 2024 15:20:07 +0100 Subject: IQ4_XS_R4 (#123) * Adding iq4_xs_r4 This is a 1st working version on Zen4. We get PP-512(LLaMA-3.1-8B) = 226 t/s, so 16% slower than iq4_nl_x4. * iq4_xs_r4: WIP * iq4_xs_r4: Use AVX2 version for matrix x vector on Zen4 * iq4_xs_r4: NEON We get PP-512(LLaMA-3.1-8B) = 115.6 t/s on M2-Max, up from 68.2 t/s for iq4_xs! * DRY --------- Co-authored-by: Iwan Kawrakow --- examples/quantize/quantize.cpp | 1 + 1 file changed, 1 insertion(+) (limited to 'examples') diff --git a/examples/quantize/quantize.cpp b/examples/quantize/quantize.cpp index a8b4a44e..f8ce3edd 100644 --- a/examples/quantize/quantize.cpp +++ b/examples/quantize/quantize.cpp @@ -41,6 +41,7 @@ static const std::vector QUANT_OPTIONS = { { "Q3_K_L", LLAMA_FTYPE_MOSTLY_Q3_K_L, " 3.35G, +0.1764 ppl @ LLaMA-v1-7B", }, { "IQ4_NL", LLAMA_FTYPE_MOSTLY_IQ4_NL, " 4.50 bpw non-linear quantization", }, { "IQ4_NL_X4",LLAMA_FTYPE_MOSTLY_IQ4_NL_X4," 4.50 bpw non-linear quantization", }, + { "IQ4_XS_R4",LLAMA_FTYPE_MOSTLY_IQ4_XS_R4," 4.25 bpw non-linear quantization", }, { "Q4_0_R4", LLAMA_FTYPE_MOSTLY_Q4_0_R4, " 4.50 bpw quantization", }, { "Q5_0_R4", LLAMA_FTYPE_MOSTLY_Q5_0_R4, " 5.50 bpw quantization", }, { "Q6_0_R4", LLAMA_FTYPE_MOSTLY_Q6_0_R4, " 6.50 bpw quantization", }, -- cgit v1.2.3