drawais/qwen3-14b-calibration-v1

qwen3-14b-calibration-v1数据集是用于校准步骤的数据集，包含公开英语文本的子集。每行数据格式为`{ "text": str, "input_ids": [int] }`。数据来源于mit-han-lab/pile-val-backup，这是一个常用于LLM校准的公开英语文本数据集。数据集发布的目的是为了可审计性，允许验证量化模型发布校准步骤的输入数据。

The qwen3-14b-calibration-v1 dataset is used for calibration purposes, consisting of a subset of public English text. Each row is formatted as `{ "text": str, "input_ids": [int] }`. The data is derived from mit-han-lab/pile-val-backup, a public English-text dataset commonly used for LLM calibration. The dataset is published for auditability, allowing verification of the inputs fed to the calibration step of quantized model releases.