eekay/gemma-2b-it-noised-np0.15-emb-lion-numbers

该数据集是通过在Gemma-2b-it模型上应用噪声注入生成的，具体在blocks.14.hook_resid_post钩子点使用add_bias_hook_fn函数。数据集包含30,000个示例，每个示例由3到10个数字组成，数字值范围从0到999。每个示例生成10个答案，答案数字最大位数为3。数据集设计用于训练或评估模型在噪声环境下的数字序列生成能力，支持批处理大小为64，最大新令牌数为96。数据集以MIT许可证发布，语言为英语。This dataset is generated via noise injection applied to the Gemma-2b-it model, specifically utilizing the add_bias_hook_fn function at the blocks.14.hook_resid_post hook point. It comprises 30,000 examples, each made up of 3 to 10 digits with values ranging from 0 to 999. Ten answers are generated for each example, with the numbers in the answers having a maximum length of 3 digits. This dataset is intended for training or evaluating a model's capacity for digit sequence generation in noisy environments, and it supports a batch size of 64 and a maximum of 96 new tokens. The dataset is released under the MIT License and is in English.