TensorFlow has two random seeds In this post, I reminded myself about some peculiarity with TensorFlow's tf.random.shuffle. I assumed that using the same seed for two shuffle operations would always output the same results but turns out, I was wrong! I discovered that I need to set the global seed with tf.random.set_seed() before every operation to get consistent results. Definitely a lesson learned to pay a bit more attention to TensorFlow's well written documentation about seeds usage for random operations. 04 September 2020