search

brain-score / candidate_models

Candidate Models to evaluate on Brain-Score benchmarks
http://brain-score.org
8 stars 20 forks source link

quick question regarding batch norm for vgg models #58

Closed rgeirhos closed 4 years ago

rgeirhos commented 4 years ago

Hi @mschrimpf, I hope all is going well. Quick question: for models "vgg-16" and "vgg-19" on brain-score.org, do these correspond to PyTorch models torchvision.models.vgg16_bn or to torchvision.models.vgg16 (with / without batch norm)? Cheers, Robert

mschrimpf commented 4 years ago

Hey @rgeirhos, great to read from you :) VGGs in this repo actually use the keras implementations: https://github.com/brain-score/candidate_models/blob/745f9d1bc747e936cbdef1e7fc599b16cf9dc677/candidate_models/base_models/__init__.py#L362-L363 which do not use BatchNorm (https://github.com/keras-team/keras-applications/blob/master/keras_applications/vgg16.py). Note that e.g. keras/pytorch pretrained models tend to have slightly different weights and scores will differ accordingly

rgeirhos commented 4 years ago

Thanks Martin for your quick reply, this answers my question :)