Any source to know pytorch model size and number of parameters

vision
1

In Keras, there is a detailed comparison of number of parameters and size in MB that model takes at Keras application page. Is there any similar resource in pytorch, where I can get a comparison of all model pretrained on imagenet and build using PyTorch. thanks

2

Hi,

There is no place for all that information but you can get them one by one. For accuracy/error please visit torchvision.models.
For model stucture/params/size you can use torchsummary in the same way in Keras.

import torchvision.models as models
alexnet = models.alexnet(pretrained=True)
alexnet.cuda()
summary(alexnet, (3, 224, 224))
print(alexnet)  # strcture - optional

It says model size but if you want to know the size of actual pretrained model, use models.<model_name>(pretrained=True) and it will download the weights.

Bests