From 956398b7fe8a4b669506e3ebb227232e700f080c Mon Sep 17 00:00:00 2001 From: Vitaly Chikunov Date: Mon, 3 Jun 2024 08:53:03 +0300 Subject: [PATCH] 1:3072-alt1.20240603 - Update to b3072 (2024-06-03). - The version scheme now matches the upstream build number more closely, instead of using the commit date. - Build with libcurl and OpenBLAS support. --- .gear/llama.cpp.spec | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) diff --git a/.gear/llama.cpp.spec b/.gear/llama.cpp.spec index cb198e7ae..71fe85f31 100644 --- a/.gear/llama.cpp.spec +++ b/.gear/llama.cpp.spec @@ -4,9 +4,10 @@ %set_verify_elf_method strict Name: llama.cpp -Version: 20240527 -Release: alt1 -Summary: Inference of LLaMA model in pure C/C++ +Version: 3072 +Release: alt1.20240603 +Epoch: 1 +Summary: LLM inference in C/C++ License: MIT Group: Sciences/Computer science Url: https://github.com/ggerganov/llama.cpp @@ -34,7 +35,7 @@ BuildRequires: tinyllamas-gguf }} %description -Plain C/C++ implementation (of inference of LLaMA model) without +Plain C/C++ implementation (of inference of many LLM models) without dependencies. AVX, AVX2 and AVX512 support for x86 architectures. Mixed F16/F32 precision. 1.5-bit, 2-bit, 3-bit, 4-bit, 5-bit, 6-bit, and 8-bit integer quantization for faster inference and reduced memory use. @@ -47,12 +48,13 @@ Supported models: Baichuan 1 & 2 + derivations, Aquila 1 & 2, Starcoder models, Refact, Persimmon 8B, MPT, Bloom, Yi models, StableLM models, Deepseek models, Qwen models, PLaMo-13B, Phi models, GPT-2, Orion 14B, InternLM2, - CodeShell, Gemma + CodeShell, Gemma, Mamba, Grok-1, Xverse, Command-R models, SEA-LION, + GritLM-7B + GritLM-8x7B, OLMo, GPT-NeoX + Pythia Multimodal models: LLaVA 1.5 models, BakLLaVA, Obsidian, ShareGPT4V, MobileVLM 1.7B/3B - models, Yi-VL + models, Yi-VL, Mini CPM, Moondream, Bunny NOTE 1: You will need to: @@ -150,6 +152,12 @@ llama-main -m %_datadir/tinyllamas/stories260K.gguf -p "Once upon a time" -s 55 %_datadir/%name %changelog +* Mon Jun 03 2024 Vitaly Chikunov 1:3072-alt1.20240603 +- Update to b3072 (2024-06-03). +- The version scheme now matches the upstream build number more closely, + instead of using the commit date. +- Build with libcurl and OpenBLAS support. + * Tue May 28 2024 Vitaly Chikunov 20240527-alt1 - Update to b3012 (2024-05-27).