Question 1: Short answer
mediumAI-checkedPyTorchA TinyVGG-style model gets RGB images of pixels. It has two blocks, each made of two
nn.Conv2d(..., out_channels=10, kernel_size=3, stride=1, padding=1) + ReLU followed by
nn.MaxPool2d(kernel_size=2). Then nn.Flatten() feeds an nn.Linear(in_features=?, 10).
What must in_features be?