Trouble reproducing text ULMFiT example results

I was following the IMDB classification example at https://docs.fast.ai/text.html#Text-models,-data,-and-training

It all runs without error but the final accuracy is 0.75, roughly 10% lower than in the example. Does this seem right? As it is quite a large difference from the example tutorial.

I am using fast ai version 1.0.60.

Thank you.

It’s a short training on a sample of the data, so it’s really not representative. I’m not surprised by the difference,

OK, thank you.