How do I re-use trained fastai models?

This error occurs whenever a batch of your data contains a single element.

Solution 1: Call learn.predict() after learn.load('resnet34_test')

Solution 2: Remove 1 data point from your training set.

Pytorch issue

Fastai forum issue description

This could be an edge case where batch size equals 1 for some batch. Make sure none of you batches = 1 (mostly the last batch)