PJMixers-Dev/proxy-logs-ReRolls

该数据集是一个合并了重复提示的样本集合，每个样本包含一个提示和多个响应。提示内容包括来源和值，响应列表则包含详细的信息如数据集名称、端点、模型、预填充、响应内容、响应令牌数等。此外，数据集还提供了响应的统计信息，如响应数量、平均令牌数、平均slop令牌数和平均slop比率。数据集仅包含训练集，大小为9550641062字节，包含274610个样本。

This dataset consists of combined duplicate prompt samples, each containing a single prompt and multiple responses. The features include prompt hash, prompt content (source and value), a list of responses (including dataset name, endpoint, model, prefill, response content, response tokens, etc.), and response statistics (such as response count, average tokens, average slop tokens, and average slop ratio). The dataset only contains a training set, with a size of 9550641062 bytes and 274610 examples.