search

bearpaw / pytorch-classification

Classification with PyTorch.
MIT License
1.68k stars 562 forks source link

Accuracy of training #17

Open Chinese-xingyi opened 6 years ago

Chinese-xingyi commented 6 years ago

Hi, I am try to use your modified model on cifar100. But my alexnet's accuracy just can be around 42%; and vgg19_bn just can be 62%around ,63.46% most when the 130 epoch . May be some parameters I set are wrong?

Chinese-xingyi commented 6 years ago

I think the learning-rate and the epochs when I need to modified is specific for different models and different datasets. I have tried to do some test about vgg19_bn on cifar10 and cifar100 -----modified learning rate of optimizer at 70th and 130th epoch or around these epochs . But my accuracy is always under 91%. Any setting can be optimized? Thank you !

bearpaw commented 6 years ago

Hi @Chinese-xingyi ，Did you follow these commands for training your model? https://github.com/bearpaw/pytorch-classification/blob/master/TRAINING.md#cifar-100