tf.estimator.SessionRunArgs

Represents arguments to be added to a Session.run() call.

tf.estimator.SessionRunArgs(
    fetches, feed_dict=None, options=None
)

fetches Exactly like the 'fetches' argument to Session.Run(). Can be a single tensor or op, a list of 'fetches' or a dictionary of fetches. For example: fetches = global_step_tensor fetches = [train_op, summary_op, global_step_tensor] fetches = {'step': global_step_tensor, 'summ': summary_op} Note that this can recurse as expected: fetches = {'step': global_step_tensor, 'ops': [train_op, check_nan_op]}
feed_dict Exactly like the feed_dict argument to Session.Run()
options Exactly like the options argument to Session.run(), i.e., a config_pb2.RunOptions proto.

fetches A namedtuple alias for field number 0
feed_dict A namedtuple alias for field number 1
options A namedtuple alias for field number 2