LieUr/Llama-3.2-3B-Instruct_csqa_oai_contrastive
收藏资源简介:
该数据集包含三个配置(keys、residuals、values),每个配置有400个训练样本,用于存储多层表示数据,可能来自神经网络模型。keys和values配置的每个样本包含一个整数标签和28个二维浮点数组(layer_0到layer_27),表示不同层的输出;residuals配置的每个样本包含一个整数标签和28个一维浮点数组(layer_0到layer_27),可能表示残差或中间特征。数据集适用于机器学习任务如分类、特征分析或模型解释研究,但具体应用场景未在README中说明。
This dataset includes three configurations (keys, residuals, values), each with 400 training examples, designed to store multi-layer representation data, likely derived from a neural network model. The keys and values configurations each contain an integer label and 28 two-dimensional float arrays (layer_0 to layer_27), representing outputs from different layers; the residuals configuration contains an integer label and 28 one-dimensional float arrays (layer_0 to layer_27), possibly indicating residuals or intermediate features. The dataset is suitable for machine learning tasks such as classification, feature analysis, or model interpretation research, though specific application contexts are not detailed in the README.




