tf_agents.metrics.tf_py_metric.TFPyMetric

Wraps a python metric as a TF metric.

Inherits From: TFStepMetric

tf_agents.metrics.tf_py_metric.TFPyMetric(
    py_metric, name=None, dtype=tf.float32
)

py_metric A batched python metric to wrap.
name Name of the metric.
dtype Data type of the metric.

Methods

call

View source

call(
    trajectory
)

Update the value of the metric using trajectory.

The trajectory can be either batched or un-batched depending on the expected inputs for the py_metric being wrapped.

Args
trajectory A tf_agents.trajectory.Trajectory.

Returns
The arguments, for easy chaining.

init_variables

View source

init_variables()

Initializes this Metric's variables.

Should be called after variables are created in the first execution of __call__(). If using graph execution, the return value should be run() in a session before running the op returned by __call__(). (See example above.)

Returns
If using graph execution, this returns an op to perform the initialization. Under eager execution, the variables are reset to their initial values as a side effect and this function returns None.

reset

View source

reset()

Resets the values being tracked by the metric.

result

View source

result()

Computes and returns a final value for the metric.

tf_summaries

View source

tf_summaries(
    train_step=None, step_metrics=()
)

Generates summaries against train_step and all step_metrics.

Args
train_step (Optional) Step counter for training iterations. If None, no metric is generated against the global step.
step_metrics (Optional) Iterable of step metrics to generate summaries against.

Returns
A list of summaries.

__call__

View source

__call__(
    *args, **kwargs
)

Returns op to execute to update this metric for these inputs.

Returns None if eager execution is enabled. Returns a graph-mode function if graph execution is enabled.

Args
*args

**kwargs A mini-batch of inputs to the Metric, passed on to call().