Pretrained language model and its related optimization techniques developed by Huawei Noah's Ark Lab.
3.02k
stars
628
forks
source link
a bug (?) when distilling TinyBERT on regression tasks with task_distill.py #128
Open
yellow-binary-tree opened 3 years ago
in TinyBERT/task_distill.py line 973:
so TinyBERT is actually learning from the label, maybe we should use
instead to learn from teacher logits?