Size mismatch,m1: [2 x 512], m2: [25088 x 4096]

vision
1

Hello, everyone.

I want to use the pretrained model in torchvision to finish the cats VS dog tasks. But there is something wrong with my code : (

RuntimeError: size mismatch, m1: [2 x 512], m2: [25088 x 4096] at /pytorch/aten/src/THC/generic/THCTensorMathBlas.cu:249

The problem seem comes from the net structure. But I just load the model in the torchvision.

Can anyone give me some suggestions?

Thanks !

And, python == 3.6 torch == 0.4.0, torchvision == 0.2.1.

import torch
import torch.nn as nn
import torchvision.models as models
import torchvision.transforms as transforms

net = models.vgg16(pretrained=True)
net = net.to(device)
test = torch.randn(2,3,32,32).to(device)
y = net(test)

Best,

2

Hi Luca Xin,

Tried in GPU as well as CPU. In either ways, this code works perfect from my side. My concern is that you are using a very old version of PyTorch and Torch Vision. In my case, I am using torch - 1.1.0 and torchvision- 0.3.0. If you can upgrade these two libraries to latest version, your problem may get resolved.

Thanks

3

ok, I’ll have a try.
Thanks a lot !