difference between just classification and pretraining+fine-tuning

[harrykwon0524]

Hello, I wanted to work on the pretraining+fine-tuning part and had some questions along the way:

1. It had a warning and an error like these: I did not have any kinds of error on the baseline datasets used for the pretraining. Is there anyway I could try to fix this?
2. When i perform just classification and pretraining+fine-tuning, it seems like the dataset input sizes get different. For instance, if I were to perform just classification on the SpokenArabicDigits, the shape looks like this: The two pictures are showing different sample sizes being inserted when doing different tasks. What is the logic behind this difference? It would be greatly appreciated if you could answer these questions. Thank you

[seay-12]

您好！我是一名计算机专业的学生，想要学习这份代码但是关于作者发布的关于分类的数据集已经无法打开，如果您方便的话可以加我的联系方式打包一份给我吗？非常感谢