tf.data.experimental.Reducer

View source on GitHub

A reducer is used for reducing a set of elements.

tf.data.experimental.Reducer(
    init_func, reduce_func, finalize_func
)

A reducer is represented as a tuple of the three functions:

• init_func - to define initial value: key => initial state
• reducer_func - operation to perform on values with same key: (old state, input) => new state
• finalize_func - value to return in the end: state => result

For example,

def init_func(_):
  return (0.0, 0.0)

def reduce_func(state, value):
  return (state[0] + value['features'], state[1] + 1)

def finalize_func(s, n):
  return s / n

reducer = tf.data.experimental.Reducer(init_func, reduce_func, finalize_func)