This code is about using bert to predict gender according to chinese name.
And the accuracy rate is around 97%.
Actually, this project has many areas that could be improved.
For example, we can find the train loss became higher in epcho2 compared to epoch 1, but in epoch 3, it drop quickly. It could be little strange. We can train more epochs and draw train_loss and accuracy rate to see the change directly and observe if the model is good-fit.
Besides, although my data have more than one billion name and gender. Its quality isn't very high.
It is important, howerver, because it is my first experience of training model.
And I also have another job with my friends Sang. It is about Qwen and ChatGLM. But it still has many problems now. And we don't have enough time to deal with it. Hope we can find enough time to restart that project.