Torchvision does not support mobilenet_v3?

1

I was basically running a model on kaggle and while
model = torchvision.model.resnet18(pretrained=true) works well, this
model = torchvision.models.mobilenet_v3_large(pretrained=True) generates an error:
AttributeError: module 'torchvision.models' has no attribute 'mobilenet_v3' This error does not occur with mobilenet_v2, but all the versions of mobilenet_v3. Am I doing something wrong here?

2

I cannot reproduce the issue using a recent source build and get:

model = models.mobilenet_v3_large(pretrained=True)
print(model)

MobileNetV3(
  (features): Sequential(
    (0): ConvBNActivation(
      (0): Conv2d(3, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (1): BatchNorm2d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
      (2): Hardswish()
    )
    (1): InvertedResidual(
      (block): Sequential(
        (0): ConvBNActivation(
          (0): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=16, bias=False)
          (1): BatchNorm2d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True)
          (2): ReLU(inplace=True)
    ...

Which torchvision version are you using?

3

I am using the default torchvision version on kaggle i.e. 0.8.1

Edit: Apparently models like mobilenet_v3 weren’t supported in 0.8.1 because I tried on 0.10.1 on colab and it worked perfectly, Moreover kaggle doesn’t allow to manually upgrade the torchvision either. Gotta use colab for this one, thanks!