Input features in the classifier

[Adelakun1999]

self.classifier = nn.Sequential(
nn.Flatten(),
nn.Linear(in_features=hidden_units77,
out_features= output_shape)
)

I don't seem to understand how we got the 7*7 in the code above

[kRiShNa-429407]

Please tell me in detail what you are really asking. Also mention video number and timestamp.

[LuluW8071]

@Adelakun1999

self.classifier = nn.Sequential(
nn.Flatten(),
nn.Linear(in_features=hidden_units*7*7,
out_features= output_shape)
)

The job of the nn.Flatten() layer is to convert 2D images (height and width arrays) into 1D feature arrays. The features flattened are of size hidden_units * 7 * 7. If you use a print(x.shape) statement in your forward pass of the model before passing through self.classifier, you will get the correct input_features.

Additionally, if it doesn't match the in_features, it will raise a shape mismatch error, providing you with a matrix error where the correct in_features will be displayed.