laion/Sera-4.6-Lite-T2-v4-316

该数据集是allenai/Sera-4.6-Lite-T2数据集的一个子集，包含316行数据。主要特点是预渲染了OpenAI的`tool_calls`为Hermes/Qwen3风格的`<tool_call>...</tool_call>`标记，并将工具响应包装为`<tool_response>...</tool_response>`。数据格式为原始JSONL，每行包含`messages`列表，角色包括`system`、`user`和`assistant`。工具观察结果表示为`role: user`，并带有`<tool_response>...</tool_response>`包装。数据集用于axolotl训练，支持`chat_template: chatml`。

Row-subset of [allenai/Sera-4.6-Lite-T2](https://huggingface.co/datasets/allenai/Sera-4.6-Lite-T2) (the dataset upstream SERA-8B was trained on), with OpenAI `tool_calls` **pre-rendered** into the content string as Hermes/Qwen3-style `<tool_call>...</tool_call>` wire tokens and tool responses wrapped as `<tool_response>...</tool_response>`. **Size**: 316 rows (source: 36,083 rows). **Format**: Raw JSONL. Per row, `messages: list[{role, content, train}]`. Roles are `system | user | assistant`. Tool observations are represented as `role: user` with `<tool_response>...</tool_response>` wrapping (per SERA convention). `train: bool` on each message is the per-message loss mask consumed by axolotls `message_field_training: train`.