def update_metrics(self, batch: Dict[str, Any],
output: Union[torch.Tensor, Dict[str, torch.Tensor],
Iterable[torch.Tensor], Any],
prediction: Dict[str, Any], metric: Union[MetricDict,
Metric]):
return SpanRankingSemanticRoleLabeler.update_metrics(
self, batch, {'prediction': prediction}, tuple(metric.values()))
def prediction_to_result(self, prediction: Dict[str, Any],
batch: Dict[str, Any]) -> List:
return SpanRankingSemanticRoleLabeler.format_dict_to_results(
batch['token'],
prediction,
exclusive_offset=True,
with_predicate=True,
with_argument=self.config.get('with_argument', ' '),
label_first=True)
def decode_output(self,
output: Dict[str, Any],
mask: torch.BoolTensor,
batch: Dict[str, Any],
decoder, **kwargs) -> Union[Dict[str, Any], Any]:
return SpanRankingSemanticRoleLabeler.decode_output(self, output, batch)
def input_is_flat(self, data) -> bool:
return SpanRankingSemanticRoleLabeler.input_is_flat(self, data)
def build_metric(self, **kwargs):
predicate_f1, end_to_end_f1 = SpanRankingSemanticRoleLabeler.build_metric(self, **kwargs)
return MetricDict({'predicate': predicate_f1, 'e2e': end_to_end_f1})