Masu haɓaka Mozilla sun fito da sakin farko na llamafile, kayan aiki don ƙirƙirar abubuwan aiwatarwa na duniya don gudanar da manyan nau'ikan harshen koyon harshe (LLM). Yin amfani da llamafile, zaku iya ɗaukar fayil ɗin GGUF tare da sigogin ƙirar ƙirar injin kuma canza shi zuwa mai aiwatarwa wanda zai iya gudana akan tsarin aiki daban-daban guda shida akan hardware tare da na'urori masu sarrafawa na AMD64 da ARM64. An rubuta kayan aikin a cikin C/C++ kuma an rarraba su ƙarƙashin lasisin Apache 2.0.
Don aiwatar da samfuran, ana amfani da tsarin llama.cpp, wanda ke aiwatar da samfurin LLaMA a cikin C/C++. An tattara wannan tsarin zuwa fayil ɗin aiwatarwa na duniya a cikin tsarin APE (Actually Portable Executable), wanda aka haɗa shi da fayil ɗin samfurin. Lambar da aka tattara tana da alaƙa da ɗakin karatu na Cosmopolitan C na yau da kullun, wanda ke ba da damar ƙirƙirar tarukan aikace-aikace waɗanda ke gudana a ciki. Linux, FreeBSD, macOS, OpenBSD, NetBSD da kuma WindowsAn ajiye lambar a cikin sigar tsarin gine-ginen AMD64 da ARM64, kuma ya haɗa da abubuwan da aka haɗa don sauke lissafi zuwa GPUs na NVIDIA (ana amfani da CUDA).
source: budenet.ru
