跳到正文
RCreddit.com·
暂不在当前实时榜单

I trained a 44M parameter quantized LLM from scratch on 45B tokens. It ships in 19.8 MB and runs at ~1,900 tok/s on CPU. [P]

AI 摘要

A developer trained a 44M parameter quantized LLM, SHADOW-250M, from scratch on 45B tokens, resulting in a 19.8 MB model that runs at ~1,900 tok/s on CPU. While SHADOW-250M performed lower on standard benchmarks like ARC-Easy (0.307) and PIQA (0.570) compared to Supra-50M-Reasoning, it demonstrated superior performance in generating direct and accurate answers to various questions, including jokes, math problems, and date calculations, where Supra often struggled or provided irrelevant information.

为什么是这条

This report uniquely details a quantized LLM that, unlike others, prioritizes direct, accurate answers over benchmark scores, showcasing its practical utility in specific tasks.

时间与来源

时间显示为 UTC

显示时区:UTC

本地时区尚不可用,暂时显示 UTC。

收录当时偏移:UTC+02026年9月15日 18:01 UTC

收录
2026年9月15日 18:01
来源类型
开发者社区

本站未收录正文。

前往源站阅读 →
来源·reddit.com