search

cavalleria / cavaface

face recognition training project(pytorch)
MIT License
459 stars 87 forks source link

EfficientNet_b1 ACC is always 0 #29

Closed ReverseSystem001 closed 4 years ago

ReverseSystem001 commented 4 years ago

Hi, Do you use the pretrained model when train EfficientNet_b1. when I train it without pretrained model, the top1 and top5 ACC is always 0.

cavalleria commented 4 years ago

No. how many epochs did you train?

ReverseSystem001 commented 4 years ago

about 13 or 14. but my batchsize is 128. how many epoches did you train when the acc is not 0?

cavalleria commented 4 years ago

just after 1 epoch

Tamminhdiep97 commented 3 years ago

Hi, can I ask about config when training EfficientNet? what learning rate did you use? did you use decay learning rate like in the paper of EfficientNet?

just after 1 epoch

cavalleria commented 3 years ago

initial lr is 0.1, and use cosinelrdecay scheduler.

Tamminhdiep97 commented 3 years ago

initial lr is 0.1, and use cosinelrdecay scheduler.

Thank you for your super fast reply @@. You're awsome