def __call__(self, engine: Engine, logger: MLflowLogger, event_name: Union[str, Events]) -> None:
if not isinstance(logger, MLflowLogger):
raise TypeError("Handler 'OutputHandler' works only with MLflowLogger")
rendered_metrics = self._setup_output_metrics_state_attrs(engine)
global_step = self.global_step_transform(engine, event_name) # type: ignore[misc]
if not isinstance(global_step, int):
raise TypeError(
f"global_step must be int, got {type(global_step)}."
" Please check the output of global_step_transform."
)
# Additionally recheck metric names as MLflow rejects non-valid names with MLflowException
from mlflow.utils.validation import _VALID_PARAM_AND_METRIC_NAMES
metrics = {}
for keys, value in rendered_metrics.items():
key = " ".join(keys)
metrics[key] = value
for key in list(metrics.keys()):
if not _VALID_PARAM_AND_METRIC_NAMES.match(key):
warnings.warn(
f"MLflowLogger output_handler encountered an invalid metric name '{key}' that "
"will be ignored and not logged to MLflow"
)
del metrics[key]
logger.log_metrics(metrics, step=global_step)
def __call__(self, engine: Engine, logger: MLflowLogger,
event_name: Union[str, Events]) -> None:
if not isinstance(logger, MLflowLogger):
raise TypeError(
"Handler 'OutputHandler' works only with MLflowLogger")
metrics = self._setup_output_metrics(engine)
global_step = self.global_step_transform(
engine, event_name) # type: ignore[misc]
if not isinstance(global_step, int):
raise TypeError(
"global_step must be int, got {}."
" Please check the output of global_step_transform.".format(
type(global_step)))
rendered_metrics = {} # type: Dict[str, float]
for key, value in metrics.items():
if isinstance(value, numbers.Number):
rendered_metrics["{} {}".format(
self.tag, key)] = value # type: ignore[assignment]
elif isinstance(value, torch.Tensor) and value.ndimension() == 0:
rendered_metrics["{} {}".format(self.tag, key)] = value.item()
elif isinstance(value, torch.Tensor) and value.ndimension() == 1:
for i, v in enumerate(value):
rendered_metrics["{} {} {}".format(self.tag, key,
i)] = v.item()
else:
warnings.warn("MLflowLogger output_handler can not log "
"metrics value type {}".format(type(value)))
# Additionally recheck metric names as MLflow rejects non-valid names with MLflowException
from mlflow.utils.validation import _VALID_PARAM_AND_METRIC_NAMES
for key in list(rendered_metrics.keys()):
if not _VALID_PARAM_AND_METRIC_NAMES.match(key):
warnings.warn(
"MLflowLogger output_handler encountered an invalid metric name '{}' that "
"will be ignored and not logged to MLflow".format(key))
del rendered_metrics[key]
logger.log_metrics(rendered_metrics, step=global_step)
