Fastai v1 Resnext50

fastai
1

Hi ,
How to use external pre-trained models like resnext50 in Fastai v1 ?
learn = create_cnn(data, models.resnext50, metrics=error_rate, model_dir=MODEL_PATH)

I am getting error:

AttributeError                            Traceback (most recent call last)
<ipython-input-19-83b18d95205d> in <module>()
----> 1 learn = create_cnn(data, models.resnext50, metrics=error_rate, model_dir=MODEL_PATH)

AttributeError: module 'fastai.vision.models' has no attribute 'resnext50'
2

Did you mean resnet50?

resnext50 is not currently an attrubute in vision.models

3

Yes, I am looking for resnext50 implementation. I know it is not in
fastai.vision
or
torchvision.vision

In old version(.7) I was able to implement resnext50 using fastai.conv_learner

Thanks,
Saurabh

4

Just in case you hadn’t seen it, PPPW posted this awesome example that shows how to use many different architectures, including Resnext50

3 Likes
5

Thank you ! It worked for me :slight_smile: