感谢分享

[junyu-Luo]

大佬，首先非常感谢你分享代码，我用过后发现有一处小问题，在xlnet_processor.py代码的157行和163行，input_mask值为0表示这个位置是真实的Token，而1表示这是一个Padding的Token，您看看这是否不小心弄反了

input_mask = [0] * len(input_ids)

input_mask = ([1] * padding_len) + input_mask

[the-black-knight-01]

ENGLISH:
Big brother, first of all, thank you very much for sharing the code. I found a small problem after using it. In the 157 and 163 lines of the xlnet_processor.py code, an input_mask value of 0 means that this position is a real Token, and 1 means that it is a Padding Token, you see if this is accidentally reversed

input_mask = [0] * len(input_ids)
input_mask = ([1] * padding_len) + input_mask

[feitboiling]

使用XLNET进行训练时，准确率只有0.5左右，您是否有这样的问题？
请问这个项目的XLNET数据输入是否有问题？