Does it compatible with Chinese pre-trained model?

[Flyfoxs]

I try it, and not exception. but the accuracy is very low. So not sure it support Chinese pretrained mode or not.
https://github.com/ymcui/Chinese-PreTrained-XLNet

[lsq357]

@Flyfoxs Which dataset and how encode text to spiece ?

[Flyfoxs]

use the dict in this link, the dict is include in the pretrain model https://github.com/ymcui/Chinese-PreTrained-XLNet

you can find the pretrain mode in this link: https://drive.google.com/open?id=1342uBc7ZmQwV6Hm6eUIN_OnBSz1LcvfA

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[CyberZHG]

@Flyfoxs

The Chinese version uses a different activation function for feed-forward layers. You can upgrade to 0.17.0 and the outputs will be the same in this demo.

[Flyfoxs]

thanks . so much