Not too deep in TensorFlow

[utahman]

Don't get me wrong, the course is fantastic but its not too deep on actual TensorFlow. After covering tf.constant and tf.Variable with some tf.math functions the course switches to making models in Keras. What would be great is to see how to use TensorFlow to say create a completely new type of layer or unusual model structure, not available in Keras.
Since Keras perfectly operates with Numpy arrays why to even explain tf.constant and tf.Variable? Either use those in the way Keras is not capable to use Numpy arrays or call this course "Keras-deep-learning". Or, better yet, add a new part where you use TensorFlow for things Keras can't be used. Just a suggestion. I love the way you explain things. Thank you!

[mrdbourke]

Hey Alex, Thank you for the feedback. I appreciate it.

…

On 1 Apr 2021, at 12:51 am, Alex ***@***.***> wrote:  Don't get me wrong, the course is fantastic but its not too deep on actual TensorFlow. After covering tf.constant and tf.Variable with some tf.math functions the course switches to making models in Keras. What would be great is to see how to use TensorFlow to say create a completely new type of layer or unusual model structure, not available in Keras. Since Keras perfectly operates with Numpy arrays why to even explain tf.constant and tf.Variable? Either use those in the way Keras is not capable to use Numpy arrays or call this course "Keras-deep-learning". Or, better yet, add a new part where you use TensorFlow for things Keras can't be used. Just a suggestion. I love the way you explain things. Thank you! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.