Kevin Roice

Hi! After downloading the dataset, have you tried [reading](https://www.tensorflow.org/tutorials/load_data/tfrecord#reading_a_tfrecord_file) the TFRecord, and then printing out its contents after decoding it? (I found [this](https://www.tensorflow.org/api_docs/python/tf/data/TFRecordDataset#used-in-the-notebooks_1) example in the TF docs)

Since we get a reward for reaching the right terminal state, we can set its value to be 1 rather than having a reward function where `reward = 1 if...