SpeechPPL/SALMon_spiritlm

这是一个包含多个配置的音频数据集，每个配置都有特定的特征和数据分割。数据集包括任务类型、索引、正负音频样本、提示音频、模型生成的音频延续、音频转换时间、代码帧率和深度、模型采样率、模型生成音频的采样率、正负样本的原始单元等。数据集被分割为训练集，并提供了相应的文件路径。This is an audio dataset consisting of multiple configurations, each with distinct characteristics and predefined data splits. The dataset encompasses task types, indices, positive and negative audio samples, prompt audio, model-generated audio continuations, audio conversion time, code frame rate and bit depth, model sampling rates, sampling rates of the model-generated audio, and original units of positive and negative samples. The dataset is partitioned into a training subset, with corresponding file paths provided.