- I am going to learn how to use tensorflow and keras in one session.
- Version 0.1
- First, implement a pure tensorflow code for a simple task, like MNIST digit recognition.
- Do the same thing in pure Keras
- a: Modify the pure tensorflow network by defining the model using Keras. b: Then train the network using Tensorflow optimizer.
- See if you can load the learned weights of a Tensorflow model onto a pure Keras model (the one defined in step 2). In other words, run the tensorflow optimization and train the network and then build a pure Keras model and set its weights using the learned weights of the Tensorflow model.
Done!
- tf_mnist.py
- keras_mnist.py
- joint_tf_keras.py
- tf_to_keras_weights.py