I have not yet looked into source code so could someone explain me how is it working with pretrained models. Will it get full model and then some order to delete last two layers and add owns or is there already this done when I call models.resnet34? If this deleting and adding is happening in create_cnn my question is that how it recognize if we use own model instead prebuilt and we dont need to remove and add layers?
1 Like
Moved to advanced category.
When you call models.resnet34 you get the full model, from torchvision.models. In the docs there’s a section describing how to use a custom model with create_ccn.
2 Likes