Create model:
model = ResNetModel(18; in_channels=3, classes=10)
or you can ommit classification layer,
by specifying number of classes
as nothing
.
- simple:
y = x |> model
- extract list of features:
features = model(x, Val(:stages))
Number of channels for each element of features
is in model.stages
.
Size of the model is in model.size
.