Extract specific outputs through pre trained classifier

Does someone know how to extract outputs at specific locations? For example, I would like to get all outputs corresponding to convolution layers, including the hidden one under sequential. Any idea?
Thanks :slight_smile:

You could use forward hooks to get the intermediate activations from your model as described here. Let me know, if this would work for you or if I misunderstood the question. :slight_smile:

Thanks a lot
What if part of the layers are built in sequential? I am using the resnet18 pre-trained model of pytorch, how to call those layers?

You can index modules inside an nn.Sequential container via indexing:

model = nn.Sequential(
    nn.Conv2d(3, 6, 3, 1, 1),
    nn.ReLU(),
    nn.Conv2d(6, 1, 3, 1, 1)
)

print(model[2]) # index second conv layer