dungnvt/vi_instruct_general_dataset_cleaned

这个数据集是`VTSNLP/instruct_general_dataset`的清理版本，专门映射为ShareGPT格式，以便与Unsloth、Axolotl和LLaMA-Factory等微调框架兼容。它采用标准ShareGPT结构，每行包含一个`conversations`列表（含`human`和`gpt`轮次）和一个`meta`字典（含`category`和`max_len`）。数据集分为多个类别：cot（思维链推理）、mcq（多项选择题）、pharaphase（释义任务）、qa（通用问答）和summary（文本摘要）。该数据集适用于越南语大型语言模型的监督微调，以遵循跨推理、问答、摘要和释义等领域的指令。This dataset is a cleaned version of `VTSNLP/instruct_general_dataset`. It has been specifically mapped to the ShareGPT format to be readily compatible with fine-tuning frameworks such as Unsloth, Axolotl, and LLaMA-Factory. The dataset uses the standard ShareGPT structure, with each row containing a `conversations` list of `human` and `gpt` turns and a `meta` dictionary with `category` and `max_len`. It is partitioned into multiple categories: cot (Chain of Thought reasoning), mcq (Multiple Choice Questions), pharaphase (Paraphrasing tasks), qa (General Question Answering), and summary (Text Summarization). The dataset is ideal for Supervised Fine-Tuning (SFT) of large language models (LLMs) to follow instructions in Vietnamese across various domains including reasoning, QA, summarization, and paraphrasing.