tim-lawson/mlsae-pythia-70m-deduped-x64-k32-examples

该数据集包含多个特征字段，如id、latent、layer、token_id、token、act、token_ids、tokens和acts，这些字段可能用于表示某种语言模型或神经网络的中间层激活状态。数据集被分割为训练集（train），包含4,673,215个样本，总大小为1,095,349,463字节。数据集的总下载大小为189,097,780字节。

This dataset includes multiple feature fields such as id, latent, layer, token_id, token, act, token_ids, tokens, and acts, which may represent the intermediate layer activation states of a language model or neural network. The dataset is split into a training set (train) containing 4,673,215 samples, with a total size of 1,095,349,463 bytes. The total download size of the dataset is 189,097,780 bytes.