Attribute error: device in fast.ai learner

Question

I am building a cnn learner using resnet18 with the following code:

from fastai.vision import *

data = (ImageList.from_df(train_df, path='/working/')
   .split_by_rand_pct(0.2)
   .label_from_df(label_delim=','))

learner = cnn_learner(data, models.resnet18, metrics=[accuracy])

I get the following error:

AttributeError Traceback (most recent call last)

AttributeError: device

How can I resolve this?

Answer 1

Try this:

1. First convert the data into databunch

   data = (ImageList.from_df(train_df, path='/working/').split_by_rand_pct(0.2).label_from_df(label_delim=',').databunch()))

2. Depending on the hardware. (I am using using cpu for processing).

   import torch data.device = torch.device('cpu')

3. Call the learner.

   learner = cnn_learner(data, models.resnet18, metrics=[accuracy])