Recently I was implementing a custom loss function in Tensorflow and I realized, Tensorflow doesn't have if conditions. I feel like if conditions are such an inherent programming feature why would Tensorflow developers not have such a fundamental feature in such a big project?
TensorFlow is not thought to be eagerly executed as python is. So the framework provides extra functions to kind o simulate eager functionalities.
Some examples are:
tf.cond(condition, true_fn, flase_fn): In case the condition is false it will execute the false_fn, otherwise the true_fnz = tf.multiply(a, b)
result = tf.cond(x < y, lambda: tf.add(x, z), lambda: tf.square(y))
tf.whilei = tf.constant(0)
c = lambda i: tf.less(i, 10)
b = lambda i: tf.add(i, 1)
r = tf.while_loop(c, b, [i])