Problem with OptimWrapper

[peiyi (Peiyi Hong)]

I am trying out different optimization methods in Kaggle notebook. I attempt to use Pytorch’s AdaGrad with OptimWrapper but get an error as followed:

The get_learner function is defined in the chapter 16 of the fastai book:

def get_learner(**kwargs):
      return cnn_learner(dls, resnet34, pretrained=False,
                                    metrics=accuracy, **kwargs).to_fp16()

Did I do something wrong? Please help me, Thanks!

[ilovescience (Tanishq)]

What version of fastai are you using? Obtain with this:

import fastai; print(fast.__version__)

[peiyi (Peiyi Hong)]

@peiyi

OptimWrapper

It’s version 2.2.7

[ilovescience (Tanishq)]

The OptimWrapper functionality was changed in 2.3.0, and versions before that work differently. I would recommend upgrading your fastai version.

[peiyi (Peiyi Hong)]

Thanks! I don’t get the error anymore after upgrading to 2.3.0.