search

huawei-noah / AdderNet

Code for paper " AdderNet: Do We Really Need Multiplications in Deep Learning?"
BSD 3-Clause "New" or "Revised" License
952 stars 187 forks source link

Train own data set #22

Closed hehaiyang111 closed 3 years ago

hehaiyang111 commented 4 years ago

hi， How does transfer learning train its own data set to modify pre-training parameters? thanks

HantingChen commented 4 years ago

Maybe you can refer to https://pytorch.org/tutorials/beginner/transfer_learning_tutorial.html.

hehaiyang111 commented 4 years ago

image image but， image acc 0.1 + why? Thanks bro.

hehaiyang111 commented 4 years ago

Is the pre-trained model trained on ImageNet or other？

HantingChen commented 4 years ago

The pre-trained ResNet-50 is trained on ImageNet.

Maybe batchsize 4 is too small to train a model. We use 256 batchsize to train our AdderNet on ImageNet.