View source on GitHub |
Executes training steps, recovers and checkpoints.
tf.contrib.distribute.Monitor(
step_callable, session=None
)
Note that this class is particularly preliminary, experimental, and expected to change.
Args | |
---|---|
step_callable
|
a training Step that's capable of signaling when done.
|
session
|
a Session instance that's needed for graph mode.
|
Raises | |
---|---|
ValueError
|
if session was provided for eager mode or not provided for
graph mode.
|
Methods
run_steps
run_steps(
num_steps=None
)