Output shape of a PyTorch layer

layer_output_shape.py

import torch  # 1.3.1
import torch.nn as nn
import torchvision  # 0.4.2

# for example
resnet = torchvision.models.resnet50(pretrained=False)
# strip away its FC layer
resnet_noFC = nn.Sequential(*list(resnet.children())[:-1])

# Suppose you want to use it for binary classification.
# How would you know the shape of the tensor to feed into your new FC layer?

# Feed it a dummy tensor and see what comes out
x = torch.zeros(1, 3, 33, 33)  # batch size, channels, height, width
y = resnet_noFC(x)

print(y.shape)