def _get_objective_trace_plot(
experiment: Experiment,
metric_name: str,
model_transitions: List[int],
optimization_direction: Optional[str] = None,
) -> Optional[go.Figure]:
best_objectives = np.array([experiment.fetch_data().df["mean"]])
return optimization_trace_single_method_plotly(
y=best_objectives,
title="Best objective found vs. # of iterations",
ylabel=metric_name,
model_transitions=model_transitions,
optimization_direction=optimization_direction,
plot_trial_points=True,
)
def _get_objective_trace_plot(
trials: Dict[int, BaseTrial],
metric_name: str,
model_transitions: List[int],
optimization_direction: Optional[str] = None,
# pyre-ignore[11]: Annotation `go.Figure` is not defined as a type.
) -> Optional[go.Figure]:
best_objectives = np.array(
[[checked_cast(Trial, t).objective_mean for t in trials.values()]])
return optimization_trace_single_method_plotly(
y=best_objectives,
title="Best objective found vs. # of iterations",
ylabel=metric_name,
model_transitions=model_transitions,
optimization_direction=optimization_direction,
plot_trial_points=True,
)
def testTraces(self):
exp = get_branin_experiment(with_batch=True)
exp.trials[0].run()
model = Models.BOTORCH(
# Model bridge kwargs
experiment=exp,
data=exp.fetch_data(),
)
# Assert that each type of plot can be constructed successfully
plot = optimization_trace_single_method_plotly(
np.array([[1, 2, 3], [4, 5, 6]]),
list(model.metric_names)[0],
optimization_direction="minimize",
)
self.assertIsInstance(plot, go.Figure)
plot = optimization_trace_single_method(
np.array([[1, 2, 3], [4, 5, 6]]),
list(model.metric_names)[0],
optimization_direction="minimize",
)
self.assertIsInstance(plot, AxPlotConfig)
def _get_objective_trace_plot(
experiment: Experiment,
data: Union[Data, MapData],
model_transitions: List[int],
) -> Optional[go.Figure]:
if experiment.is_moo_problem:
return _get_hypervolume_trace()
metric_name = not_none(
experiment.optimization_config).objective.metric.name
best_objectives = np.array(
[data.df[data.df["metric_name"] == metric_name]["mean"]])
return optimization_trace_single_method_plotly(
y=best_objectives,
title="Best objective found vs. # of iterations",
ylabel=not_none(experiment.optimization_config).objective.metric.name,
model_transitions=model_transitions,
optimization_direction=("minimize" if not_none(
experiment.optimization_config).objective.minimize else
"maximize"),
plot_trial_points=True,
)
def _get_objective_trace_plot(
experiment: Experiment,
data: Data,
model_transitions: List[int],
true_objective_metric_name: Optional[str] = None,
) -> Iterable[go.Figure]:
if experiment.is_moo_problem:
# TODO: implement `_get_hypervolume_trace()`
return [_pareto_frontier_scatter_2d_plotly(experiment=experiment)]
optimization_config = experiment.optimization_config
if optimization_config is None:
return []
metric_names = (metric_name for metric_name in [
optimization_config.objective.metric.name,
true_objective_metric_name,
] if metric_name is not None)
plots = [
optimization_trace_single_method_plotly(
y=np.array(
[data.df[data.df["metric_name"] == metric_name]["mean"]]),
title=f"Best {metric_name} found vs. # of iterations",
ylabel=metric_name,
model_transitions=model_transitions,
# Try and use the metric's lower_is_better property, but fall back on
# objective's minimize property if relevent
optimization_direction=(
("minimize" if
experiment.metrics[metric_name].lower_is_better is True else
"maximize") if experiment.metrics[metric_name].lower_is_better
is not None else
("minimize"
if optimization_config.objective.minimize else "maximize")),
plot_trial_points=True,
) for metric_name in metric_names
]
return [plot for plot in plots if plot is not None]