For training, since we're using transfer learning, you first need to download, extract GloVe and put to data folder, this is a pre trained embedding vectors that map each word to its vector, two words that have similar meanings tend to have very close vectors, and so on.
python3 spam_classifier.py
This will spawn tensorflow logs in logs folder, as well as the model and tokenizer in results, so test.py will use them.
After the training has finished, try testing your own emails, or change the code on your needs, or whatever: