finance-alpaca

该数据集被随机分为两份独立的副本，一份用于常规的微调，另一份则用于进一步微调的训练。该数据集旨在模拟同一领域内未经授权使用的情况。每个数据集包含22,960个数据样本。任务的目的是对后门水印方法进行微调和训练。

This dataset is randomly partitioned into two independent subsets. One subset is utilized for conventional fine-tuning, and the other is used for training in subsequent fine-tuning. This dataset is intended to simulate scenarios of unauthorized usage within the same domain. Each subset contains 22,960 data samples. The objective of the task is to fine-tune and train backdoor watermarking methods.