Add spancat_exclusive to pipeline

spacy/pipeline/__init__.py

@@ -13,6 +13,7 @@ from .sentencizer import Sentencizer
 from .tagger import Tagger
 from .textcat import TextCategorizer
 from .spancat import SpanCategorizer
+from .spancat_exclusive import SpanCategorizerExclusive
 from .span_ruler import SpanRuler
 from .textcat_multilabel import MultiLabel_TextCategorizer
 from .tok2vec import Tok2Vec
@@ -31,6 +32,7 @@ __all__ = [
     "SentenceRecognizer",
     "Sentencizer",
     "SpanCategorizer",
+    "SpanCategorizerExclusive",
     "SpanRuler",
     "Tagger",
     "TextCategorizer",

spacy/pipeline/spancat_exclusive.py

@@ -78,7 +78,7 @@ def make_spancat(
     model: Model[Tuple[List[Doc], Ragged], Floats2d],
     spans_key: str,
     scorer: Optional[Callable],
-    negative_weight: Optional[float] = 1.0,
+    negative_weight: float = 1.0,
     allow_overlap: Optional[bool] = True,
 ) -> "SpanCategorizerExclusive":
     """Create a SpanCategorizer component. The span categorizer consists of two
@@ -95,7 +95,7 @@ def make_spancat(
     spans_key (str): Key of the doc.spans dict to save the spans under. During
         initialization and training, the component will look for spans on the
         reference document under the same key.
-    negative_weight (Optional[float]): Multiplier for the loss terms.
+    negative_weight (float): Multiplier for the loss terms.
         Can be used to down weigh the negative samples if there are too many.
     allow_overlap (Optional[bool]): If True the data is assumed to
         contain overlapping spans.
@@ -133,7 +133,6 @@ class Ranges:
         return False
 
 
-# TODO: Documentation
 class SpanCategorizerExclusive(TrainablePipe):
     """Pipeline component to label spans of text.
 
@@ -148,7 +147,7 @@ class SpanCategorizerExclusive(TrainablePipe):
         name: str = "spancat_exclusive",
         *,
         spans_key: str = "spans",
-        negative_weight: Optional[float],
+        negative_weight: float = 1.0,
         scorer: Optional[Callable] = spancat_score,
         allow_overlap: Optional[bool] = True,
     ) -> None:
@@ -161,7 +160,7 @@ class SpanCategorizerExclusive(TrainablePipe):
             During initialization and training, the component will look for
             spans on the reference document under the same key. Defaults to
             `"spans"`.
-        negative_weight (Optional[float]): Multiplier for the loss terms.
+        negative_weight (float): Multiplier for the loss terms.
             Can be used to down weigh the negative samples if there are too many.
         scorer (Optional[Callable]): The scoring method. Defaults to
             Scorer.score_spans for the Doc.spans[spans_key] with overlapping
@@ -394,13 +393,15 @@ class SpanCategorizerExclusive(TrainablePipe):
     ) -> None:
         """Initialize the pipe for training, using a representative set
         of data examples.
+
         get_examples (Callable[[], Iterable[Example]]): Function that
             returns a representative sample of gold-standard Example objects.
         nlp (Optional[Language]): The current nlp object the component is part of.
         labels (Optional[List[str]]): The labels to add to the component, typically generated by the
             `init labels` command. If no labels are provided, the get_examples
             callback is used to extract the labels from the data.
-        DOCS: https://spacy.io/api/spancategorizer#initialize
+
+        DOCS: https://spacy.io/api/spancategorizerexclusive#initialize
         """
         subbatch: List[Example] = []
         if labels is not None:
@@ -419,9 +420,11 @@ class SpanCategorizerExclusive(TrainablePipe):
             # + 1 for the "no-label" category
             Y = self.model.ops.alloc2f(spans.dataXd.shape[0], self._n_labels)
             self.model.initialize(X=(docs, spans), Y=Y)
-        # FIXME I think this branch is broken
         else:
-            raise ValueError("Cannot initialize without examples.")
+            # FIXME: Ideally we want to raise an error to avoid implicitly
+            # raising it when initializing without examples. For now, we'll just
+            # copy over what `spancat` did.
+            self.model.initialize()
 
     def _validate_categories(self, examples: Iterable[Example]):
         # TODO