resnet + swish + training on ImageNet in PyTorch
These code is written using pytorch version 67839ce, which is higher than the latest stable release 0.2.0.
In swish.py it supports in-place option but it is NOT working in pytorch 0.2.0. If you are using older version please turn it off.
You could got the pretrained model and checkpoint file here.
Note this model is only trained for 11 epoch and still not good enough to use.
validation prec:
- resnet152: top1: 38.256%, top5: 64.610%