search

alibaba / cluster-contrast-reid

MIT License
220 stars 59 forks source link

About the setting of batch_size and iter number #25

Open xiguafeichong opened 2 years ago

xiguafeichong commented 2 years ago

For market1501 dataset, I try to set the batch size 128 or 64, iter num and other settings remain the same as run.sh, and the result are 68% mAP & 86% Rank-1, which is not in accordance with the results in your paper. I just wonder that if I need to lower the batch size, is there any other hyperparameters that need to be adjusted accordingly? And I also wonder that why u set the iteration number to 200 or 400 manually instead of directly using "for i,inputs in enumerate(data_loader)" to train each iteration?

AvadaKarrot commented 2 years ago

Have you successfully figured out these problems? I encountered them either. Could you please tell me how to set the hyper-parameters. Thank u.