Could you provide pre-trained weights for baseline architecture?
erobic opened this issue · comments
Would it be possible to provide the checkpoint for baseline architecture as well?
I got very low accuracy (only 21.74) with baseline.yaml:
[S 2019-08-22 13:44:22] ...trap/engines/engine.py.120: eval_epoch.epoch: 11
[S 2019-08-22 13:44:22] ...trap/engines/engine.py.120: eval_epoch.loss: 4.28
[S 2019-08-22 13:44:22] ...trap/engines/engine.py.120: eval_epoch.accuracy_top1: 21.74
[S 2019-08-22 13:44:22] ...trap/engines/engine.py.120: eval_epoch.accuracy_top5: 70.43
which is very puzzling.
Instead of looking at the accuracies printed, I just had to use a proper evaluation script: #4