Lesson4-mnist_sgd after tuning trying to predict on Kaggle test.csv

I’ve modified the lesson4-mnist_sgd notebook to be a neural network and trained it on the Kaggle Mnist training data after splitting it into 9/10ths train and 1/10th validate.

I’m having trouble now figuring out how to get the tuned model to make predictions on the Kaggle test.csv file.

I’ve tried loading the test.csv charectors into an array and putting into
md.eval(test_chars), fit.eval(test_chars).
I’m new so not sure if this was covered somewhere in the lessons.
Thank you.