SpeechPPL/SALMon_GSLM

该数据集包含了多个配置，每个配置都有其特定的特征和数据类型。数据集中的特征包括任务、索引、正音频、负音频、提示音频、正音频延续、负音频延续、负音频检查、正样本标记损失、负样本标记损失、代码帧率、代码深度、模型采样率、语言模型困惑度和模型生成的延续音频。数据集还提供了不同的分割，如训练集，并给出了每个分割的样本数量和文件大小。每个配置的下载大小和数据集大小也都有提供。

This dataset comprises multiple configurations, each with its own specific features and data types. The features contained in the dataset include task, index, positive audio, negative audio, prompt audio, positive audio continuation, negative audio continuation, negative audio check, positive sample token loss, negative sample token loss, code frame rate, code depth, model sampling rate, language model perplexity, and model-generated continuation audio. The dataset also provides various splits such as the training set, with the sample count and file size specified for each split. The download size and total dataset size for each configuration are additionally provided.