search

VDIGPKU / DADA

[ECCV 2020] DADA: Differentiable Automatic Data Augmentation
MIT License
189 stars 30 forks source link

Imagenet #6

Open PushparajaMurugan opened 4 years ago

PushparajaMurugan commented 4 years ago

Hi. Unable to reproduce the results (search_relax). I got only 65.23% (top1) in imagenet.?

latstars commented 4 years ago

Would you like to provide us the searching and training logs to help find out the problems?

PushparajaMurugan commented 4 years ago

Hi @latstars,

pls have a look at the log files.

Search_log.txt imaenet_training_log-20200901-015111.txt

thanks, Raja.

latstars commented 4 years ago

Hi @latstars,

pls have a look at the log files.

Search_log.txt imaenet_training_log-20200901-015111.txt

thanks, Raja.

Hi, we find the learning rate schedule in your imagenet-training log is not the same with our schedule. Maybe you should check your pytorch version and the training code. Otherwise, you could try train the imagenet with batchsize 1024 in 8 GPUs.

PushparajaMurugan commented 4 years ago

Hi @latstars . Thanks, Let me train the network again.