curious-synapse comments

Repositories
Issues
Comments

Results 2 comments of


                                            curious-synapse

Lower top-1 accuracy on pre-trained CIFAR10 Resnet models

I ran the above script again with torch==1.2.0 (which came out in 2019) and got the same results as above. Are the test accuracies plotted in the paper top-5 accuracies...

CIFAR100 pretrained models and results

Thanks for the quick response! Do the same hyperparameters for CIFAR10 work on CIFAR100? Also, do you have the accuracy numbers (because the paper only has a plot) for ResNet...