From 707d08792701de5a517c45c6c16bf4d816c00d88 Mon Sep 17 00:00:00 2001 From: Iwan Kawrakow Date: Mon, 24 Jun 2024 16:42:30 +0200 Subject: Bitnet: tiny bity faster 1.625 bpw variant on Metal We get 70.7 t/s for TG-128 vs 69.5 t/s before. --- examples/quantize/quantize.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'examples') diff --git a/examples/quantize/quantize.cpp b/examples/quantize/quantize.cpp index a5ffb2b2..0a61a083 100644 --- a/examples/quantize/quantize.cpp +++ b/examples/quantize/quantize.cpp @@ -26,7 +26,7 @@ static const std::vector QUANT_OPTIONS = { { "IQ2_M", LLAMA_FTYPE_MOSTLY_IQ2_M, " 2.7 bpw quantization", }, { "IQ1_S", LLAMA_FTYPE_MOSTLY_IQ1_S, " 1.56 bpw quantization", }, { "IQ1_M", LLAMA_FTYPE_MOSTLY_IQ1_M, " 1.75 bpw quantization", }, - { "IQ1_BN", LLAMA_FTYPE_MOSTLY_IQ1_BN, " 1.75 bpw quantization (Bitnet)", }, + { "IQ1_BN", LLAMA_FTYPE_MOSTLY_IQ1_BN, " 1.62 bpw quantization (Bitnet)", }, { "IQ2_BN", LLAMA_FTYPE_MOSTLY_IQ2_BN, " 2.00 bpw quantization (Bitnet)", }, { "Q2_K", LLAMA_FTYPE_MOSTLY_Q2_K, " 2.63G, +0.6717 ppl @ LLaMA-v1-7B", }, { "Q2_K_S", LLAMA_FTYPE_MOSTLY_Q2_K_S, " 2.16G, +9.0634 ppl @ LLaMA-v1-7B", }, -- cgit v1.2.3