labs-aibusiness/reasoning-distill-opus-4-7-max-sft
收藏资源简介:
该数据集包含7,823个单轮推理对话,这些对话来自Claude Opus 4.7模型,并重新格式化为适用于监督微调(SFT)的格式。每个对话都是一个完整的Qwen风格聊天模板对话,包含一个text字段。数据集的每个助手响应(包括<think>...</think>块)都是由Claude Opus 4.7模型生成的,且启用了Anthropic的extended-thinking功能。数据集适用于文本生成任务,大小为1K<n<10K,训练集包含7,823个示例,总大小为29,328,233字节。数据集格式为Qwen聊天模板,可直接用于`SFTTrainer`训练。
This dataset contains 7,823 single-turn reasoning conversations from Claude Opus 4.7 reformatted for supervised fine-tuning (SFT). Each row is a single text field containing a full Qwen-style chat-template conversation. The assistant responses (including the <think>...</think> block) are generated by Claude Opus 4.7 with Anthropics extended-thinking enabled. The dataset is suitable for text-generation tasks, with a size category of 1K<n<10K. The training split includes 7,823 examples, totaling 29,328,233 bytes. The dataset is formatted in Qwen chat template and is ready for use with `SFTTrainer`.




