Validating pretrained models does not work correctly

[JeanKaddour]

Hi Ross,

Thank you very much for developing this fantastic library.

I want to evaluate a few pre-trained models. However, the accuracies I get are all very bad. Here's what I did:

1. Download and unzipILSVRC2012_img_test_v10102019.tar from here
2. run python validate.py ~/imagenet/test/ --model resnet50 --pretrained

Here's what the output looks like:

I've also tested it with a few other models and the results all look similarly bad.

Am I using the wrong dataset/config/script?

[JeanKaddour]

Nevermind, found the fix: #131 (comment)