EpicVic2193/restaurant-inventory-forecasting

高质量合成训练数据集，用于对小型LLM（Qwen2.5-1.5B）进行后训练，使其成为餐厅库存预测的专家推理引擎。数据集包含SFT（监督微调）和GRPO（强化学习）的分割，覆盖7种餐厅类型、40多种菜单项、8种天气条件和12种特殊事件等多种场景。High-quality synthetic training data for post-training a small LLM (Qwen2.5-1.5B) to become an expert reasoning engine for restaurant inventory forecasting. The dataset includes splits for SFT (supervised fine-tuning) and GRPO (reinforcement learning), covering 7 restaurant types, 40+ menu items, 8 weather conditions, and 12 special events.