def log_output(
task_name: str,
identifier: str,
val: float,
):
store = Store(_db_uri)
store.log_output(identifier=identifier, task_name=task_name, val=val)
class Task(object):
def __init__(self, task_name: str):
self.task_name = task_name
self.store = Store(clientUtils.get_db_uri())
self.metrics = []
# TODO(shreyashankar): Add metric cache
def registerMetric(self, metric: Metric, create_view: bool = True):
if create_view:
self.store.create_view(self.task_name, metric.window_size)
self.metrics.append(metric)
def computeMetrics(self, use_views=True):
results = {}
for metric in self.metrics:
if use_views:
results[
metric.getIdentifier()
] = self.store.compute_metric_from_view(
self.task_name, metric.fn, window_size=metric.window_size
)
else:
results[metric.getIdentifier()] = self.store.compute_metric(
self.task_name, metric.fn, window_size=metric.window_size
)
return results
def logOutput(
self,
output_value: float,
identifier: str,
):
self.store.log_output(
identifier=identifier,
task_name=self.task_name,
val=output_value,
)
def logOutputs(
self,
output_values: typing.List[float],
identifiers: typing.List[str],
):
self.store.log_outputs(
task_name=self.task_name,
vals=output_values,
identifiers=identifiers,
)
def logFeedbacks(
self,
feedback_values: typing.List[float],
identifiers: typing.List[str],
):
self.store.log_feedbacks(
task_name=self.task_name,
vals=feedback_values,
identifiers=identifiers,
)
def logFeedback(
self,
feedback_value: float,
identifier: str,
):
self.store.log_feedback(
identifier=identifier,
task_name=self.task_name,
val=feedback_value,
)
def getOutputs(self, limit: int = None, window_size: int = None):
return self.store.get_outputs_or_feedback(
self.task_name,
tablename="output_table",
limit=limit,
window_size=window_size,
)
def getFeedback(self, limit: int = None, window_size: int = None):
return self.store.get_outputs_or_feedback(
self.task_name,
tablename="feedback_table",
limit=limit,
window_size=window_size,
)
def computeMetric(
self,
metric: typing.Union[typing.Callable, str],
window_size: int = None,
):
metric_fn = (
get_metric_function(metric) if not callable(metric) else metric
)
return self.store.compute_metric(
self.task_name, metric_fn, window_size=window_size
)