mlabonne/dolphin-r1-flash

Dolphin R1 Flash数据集是一个遵循Apache-2.0许可的由Eric Hartford和Cognitive Computations团队整理的数据集。该数据集的目的是为了训练R1风格的推理模型。它是Flash子集的重新格式化版本，为便于使用，在会话中添加了特殊的标记，包括模型的响应。数据集包含了会话的角色和内容信息，并且划分为训练集。训练集包含了299998个示例，大小为3.1GB。

The Dolphin R1 Flash dataset is an Apache-2.0 licensed dataset curated by Eric Hartford and the Cognitive Computations team. The purpose of this dataset is to train R1-style reasoning models. It is a reformatted version of the Flash subset, adding special tokens to the conversation for ease of use, including the models response. The dataset contains information about the roles and contents of the conversations, and is split into a training set. The training set consists of 299,998 examples and is 3.1GB in size.