Question about the 'evaluate_model' func

[lyfdy456]

Hello! Your article is great and gave me a lot of inspiration! But when the code ran, I encountered some problems.

I've been training the model on the MNIST dataset (default hyperparameters), and the training process seems to work fine. However, when comes to the evaluate_model function, the output 'test_rec' values are all NaN. Have you encountered this issue while running the code? Thank you in advance.