Video 189 - versions and errors when declaring a pretrained model

[slawomirwojtas]

I had a quite of an issue when it came to these lines of code:
weights = torchvision.models.EfficientNet_B0_Weights.DEFAULT # .DEFAULT = best available weights model = torchvision.models.efficientnet_b0(weights=weights).to(device)

It just didn't work. The versions of torch and torchvisions where:
2.1.0+cu118
0.16.0+cu118

I decided to run your code where there is an update for those so after running it, the code was fine and the versions were:
torch version: 2.1.1+cu121
torchvision version: 0.16.1+cu121

The thing that puzzles me are the versions. In terms of torch they both read 2.1. so it's not clear why one of them doesn't meet the condition:
assert int(torch.__version__.split(".")[1]) >= 12, "torch version should be 1.12+" whilst the other does.

And what are the +cu121 suffices?

[sametucuncu]

Similar issue on #693 might solve your problem.