search

lenscloth / RKD

Official pytorch Implementation of Relational Knowledge Distillation, CVPR 2019
395 stars 49 forks source link

A question about recall #15

Closed Youskrpig closed 4 years ago

Youskrpig commented 4 years ago

why the eval recall decrease with the training process goes on ? Teacher model: resnet 50
student model: resnet 18
dataset: CUB2011

lenscloth commented 4 years ago

@Youskrpig CUB-200-2011 has only few thousands of training datasets. I guess your model simply overfitted on the training set.

Youskrpig commented 4 years ago

got it