search

zihangdai / xlnet

XLNet: Generalized Autoregressive Pretraining for Language Understanding
Apache License 2.0
6.16k stars 1.18k forks source link

reuse_len=0 means no mem? And no benefit for long text but not worse for short text? #271

Closed guotong1988 closed 3 years ago

guotong1988 commented 3 years ago

Am I right?

guotong1988 commented 3 years ago

https://github.com/zihangdai/xlnet/issues/59