CUDA error with Titanic Dataset (CUDA error: device-side sort triggered)

Hey guys!

After completing lesson 7, I decided to try my hand at the Titanic dataset from Kaggle. I was able to train a RandomForestClassifier but I got a CUDA error: device-side sort triggered when I tried to train a neural network.

The notebook is here: Paperspace Console and the training code begins under the Processing Pipeline heading.

Could somebody help me figure out how to resolve this error?

You should run your code on the CPU to really understand what is happening here, ie learn.dls.cpu(), learn.lr_find().

You can also install fastdebug and do a sanity check as well:

!pip install fastdebug
from fastdebug import *

# building learner etc
learn.sanity_check()