search

StefOe / all-conv-pytorch

Pytorch implementation of "Striving for Simplicity: The All Convolutional Net" (https://arxiv.org/abs/1412.6806)
MIT License
15 stars 11 forks source link

Hyperparams for state of the art performance #1

Open brando90 opened 6 years ago

brando90 commented 6 years ago

Hi, I was looking at ur other file with the hyper params etc but u didn't show it achieving 95% accuracy...I was wondering, did you ever manage to achieve that?

yMayanand commented 3 years ago

Hi, I was looking at ur other file with the hyper params etc but u didn't show it achieving 95% accuracy...I was wondering, did you ever manage to achieve that?

i have also tried this architecture but i couldnt manage to reduce loss for cifar10 dataset. Loss is not decreasing any further. @brando90 could please tell how did you managed to get accuracy of 95 %,