some config parameters of downloaded pre-trained model are different from the config in 'Readme' about chinese-xlnet HOT 2 CLOSED

Well, it is something with the conversion tools provided by pytorch-bert of huggingface.
If you would like to do additional pre-training steps, you should change 'null' to the correct numbers. However, if you only use this model for fine-tuning on down-stream tasks, then it is fine to leave as 'null'.
For example, check 'run_classifier' by XLNet (https://github.com/zihangdai/xlnet/blob/master/run_classifier.py), no reuse_len or men_len should be provided.

from chinese-xlnet.

I see, thanks a lot

from chinese-xlnet.