how to get features from intermediate layers ?

[Husky-AI9]

How would I got about extracting features of layer other than penultimate layers ? For example, what would the code look like if I want to get features from the layer before the penultimate layer in efficient_b0 model ?

[hankyul2]

Hi @Husky95

If I understand the meaning of penultimate layer correctly, you can extract each layers output by passing features_only=True argument in create_model function. Here is code:

import torch
from timm.models import create_model

model = create_model('efficientnet_b0', pretrained=True, features_only=True)
x = torch.rand([2, 3, 224, 224])
features = model(x)

If I misunderstood the meaning of penultimate layer, please correct me: penultimate layer = final classification layer. Is it right?

I hope this answer can help you.

Thank you.

hankyul