metanlp
metanlp copied to clipboard
Meta-learning for NLP
Hi, From https://github.com/iesl/leopard, it looks like data for K=32 is missing. Is there a workaround for this issue? When I try to fine-tune the models, I get this result -...
How is the softmax classifier initialized in the Bert-Base model? Is zero initialized?
Do you have time to fill in this blanks?
Hi, Thanks for your organized repository and for sharing your code. Your paper about Self-Supervised Meta-Learning is interesting. I have a few questions and I appreciate it if you could...
“meta_pretrain” file refer to several tasks data? How to get it?