veeraragavan410/gitechgames-dataset
收藏资源简介:
GI Tech Games数据集是一个特定领域的指令跟随数据集,基于GI Tech I Game(一家总部位于印度、业务遍及40多个国家的游戏和彩票技术公司)构建。该数据集采用Llama 2指令格式(格式为:<s>[INST] 问题 [/INST] 答案 </s>),旨在微调语言模型以回答关于GI Tech游戏产品和服务的相关问题。数据集包含155个条目,分为训练集(139行)和测试集(16行)。数据内容涵盖多个主题,包括彩票解决方案、赌场/轮盘赌、体育博彩、公司概述、服务/AI/云、拉米游戏、宾果游戏、虚拟体育、百家乐、扑克、合作伙伴/联系方式等。数据集可用于微调Llama 2、Mistral或其他指令跟随大语言模型,进行领域特定的问答任务,也适用于低资源领域适应的QLoRA/LoRA微调、游戏和彩票运营商的聊天机器人训练,以及游戏行业知识的检索增强生成(RAG)基线评估。数据来源于公开网站gitechgames.com,包括彩票、体育博彩、赌场、拉米游戏、宾果游戏、扑克、百家乐、虚拟体育和服务等产品页面。
A domain-specific instruction-following dataset built from GI Tech I Game — a gaming and lottery technology company based in India operating across 40+ countries. This dataset is formatted in the Llama 2 instruction format (<s>[INST] question [/INST] answer </s>) and is designed for fine-tuning language models to answer questions about GI Techs gaming products and services. The dataset contains 155 entries, with 139 in the train split and 16 in the test split. Topics covered include Lottery Solutions, Casino / Roulette, Sports Betting, Company Overview, Services / AI / Cloud, Rummy, Bingo, Virtual Sports, Baccarat, Poker, Partner / Contact, and others. Intended uses include fine-tuning Llama 2, Mistral, or other instruction-following LLMs for domain-specific Q&A; QLoRA / LoRA fine-tuning for low-resource domain adaptation; chatbot training for gaming and lottery operators; and RAG baseline evaluation for gaming industry knowledge. All data was sourced from the public website https://gitechgames.com, including product pages for lottery, sports betting, casino, rummy, bingo, poker, baccarat, virtual sports, and services.




