ChatTSRepo/ChatTS-Training-Dataset

该数据集是用于ChatTS项目的训练数据，包含两个阶段的数据集：第一阶段是用于大规模对齐训练的数据集，第二阶段是用于SFT（Softmax Function Training）的数据集。此外，还包括一个名为`dev.jsonl`的小型数据集，用于开发和测试。数据格式包括输入文本提示、输出文本以及时间序列数据。时间序列数据采用`sp`编码方法进行处理，并为每个数据点添加了一个数值为1.0的掩码。

This dataset is the training data for the ChatTS project, including two stages of datasets: the first stage is the dataset for large-scale alignment training, and the second stage is the dataset for SFT (Softmax Function Training). In addition, there is a small dataset named `dev.jsonl` for development and testing. The data format includes input text prompts, output text, and time series data. The time series data is processed using the `sp` encoding method, and a mask value of 1.0 is added to each data point.