cazonai/autoradio-destilado-v1

AutoRadio Destilado v1 — Dataset da Elyra 0.1是一个通过OpenRouter从Gemma 4 26B A4B蒸馏生成的合成数据集，用于训练Elyra 0.1模型。数据集包含9202个样本，格式为ChatML（system + user + assistant），总共有约6M tokens，100%为巴西葡萄牙语（PT-BR），重复率仅为0.4%，表明数据质量较高。数据集内容具有强烈的Elyra身份特征，包括Cazon AI公司的层级结构、Elyra AI和AutoRadio产品的信息，以及特定的技术知识如ESP32、MQTT、RDA5807M、KT0803L和Cazon AI的价格信息。数据集快照拍摄于2026-04-22 09:51，蒸馏过程仍在进行，未来会有v2版本更新。

AutoRadio Destilado v1 — Dataset da Elyra 0.1 is a synthetic dataset generated by distillation from Gemma 4 26B A4B via OpenRouter, used to train the Elyra 0.1 model. The dataset contains 9202 samples in ChatML format (system + user + assistant), totaling ~6M tokens, 100% in Brazilian Portuguese (PT-BR), with only 0.4% duplicates (high quality). The content has a strong Elyra identity, including the Cazon AI company hierarchy, Elyra AI, and AutoRadio product information, as well as specific technical knowledge such as ESP32, MQTT, RDA5807M, KT0803L, and Cazon AI prices. A snapshot was taken on 2026-04-22 09:51 (distillation still running — updates will come as v2).