Closed austinmw closed 5 years ago
Hi @austinmw, as you probably found out, in the Cadene pretrained models models/torchvision_models.py
-> modify_alexnet
-> features
, there's a view
function call which will not be included in children
, so this one can't be directly passed to create_cnn
. The torchvision's AlexNet works fine with fastai.
@PPPW Thanks, I also found that I needed to replace the last linear layer completely rather than trying to modify its out_features value.
Hi, thanks for creating these examples!
I used the cadene alexnet and dog/cat breed dataset to create a learner.
learn.model
prints the right model, butlearn.summary()
prints the wrong model head. Any idea what's going wrong?