Fix document_type property of taskmodules (#361)

* set default value for annotation to "label" in TransformerTextClassificationTaskModule to match the document type

* check if the annotation layer names required in the taskmodules match the ones in the usual document types that they request and request no document type in the case of a mismatch (bot log a warning)

* make mypy happy

src/pytorch_ie/taskmodules/transformer_re_text_classification.py

@@ -215,10 +215,22 @@ def __init__(
 
     @property
     def document_type(self) -> Optional[Type[TextDocument]]:
+        dt: Type[TextDocument]
         if self.partition_annotation is not None:
-            return TextDocumentWithLabeledSpansBinaryRelationsAndLabeledPartitions
+            dt = TextDocumentWithLabeledSpansBinaryRelationsAndLabeledPartitions
         else:
-            return TextDocumentWithLabeledSpansAndBinaryRelations
+            dt = TextDocumentWithLabeledSpansAndBinaryRelations
+
+        if self.relation_annotation == "binary_relations":
+            return dt
+        else:
+            logger.warning(
+                f"relation_annotation={self.relation_annotation} is "
+                f"not the default value ('binary_relations'), so the taskmodule {type(self).__name__} can not request "
+                f"the usual document type for auto-conversion ({dt.__name__}) because this has the bespoken default "
+                f"value as layer name instead of the provided one."
+            )
+            return None
 
     def get_relation_layer(self, document: Document) -> AnnotationList[BinaryRelation]:
         return document[self.relation_annotation]

src/pytorch_ie/taskmodules/transformer_seq2seq.py

@@ -9,7 +9,7 @@
 
 import logging
 import re
-from typing import Any, Dict, Iterator, List, Optional, Sequence, Tuple, Union
+from typing import Any, Dict, Iterator, List, Optional, Sequence, Tuple, Type, Union
 
 from transformers import AutoTokenizer
 from transformers.file_utils import PaddingStrategy
@@ -69,6 +69,23 @@ def __init__(
 
         self.tokenizer = AutoTokenizer.from_pretrained(tokenizer_name_or_path)
 
+    @property
+    def document_type(self) -> Optional[Type[TextDocument]]:
+        dt: Type[TextDocument] = self.DOCUMENT_TYPE
+        if (
+            self.entity_annotation == "labeled_spans"
+            and self.relation_annotation == "binary_relations"
+        ):
+            return dt
+        else:
+            logger.warning(
+                f"entity_annotation={self.entity_annotation} and relation_annotation={self.relation_annotation} are "
+                f"not the default values ('labeled_spans' and 'binary_relations'), so the taskmodule "
+                f"{type(self).__name__} can not request the usual document type ({dt.__name__}) for auto-conversion "
+                f"because this has the bespoken default values as layer names instead of the provided ones."
+            )
+            return None
+
     def encode_text(self, text: str) -> InputEncodingType:
         return self.tokenizer(
             text,

src/pytorch_ie/taskmodules/transformer_span_classification.py

@@ -8,7 +8,7 @@
 """
 
 import logging
-from typing import Any, Dict, Iterator, List, Optional, Sequence, Set, Tuple, Union
+from typing import Any, Dict, Iterator, List, Optional, Sequence, Set, Tuple, Type, Union
 
 import numpy as np
 import torch
@@ -90,11 +90,23 @@ def __init__(
         self.multi_label = multi_label
 
     @property
-    def document_type(self) -> TypeAlias:
+    def document_type(self) -> Optional[Type[TextDocument]]:
+        dt: Type[TextDocument]
         if self.single_sentence:
-            return TextDocumentWithLabeledSpansAndSentences
+            dt = TextDocumentWithLabeledSpansAndSentences
         else:
-            return TextDocumentWithLabeledSpans
+            dt = TextDocumentWithLabeledSpans
+
+        if self.entity_annotation == "labeled_spans":
+            return dt
+        else:
+            logger.warning(
+                f"entity_annotation={self.entity_annotation} is "
+                f"not the default value ('labeled_spans'), so the taskmodule {type(self).__name__} can not request "
+                f"the usual document type ({dt.__name__}) for auto-conversion because this has the bespoken default "
+                f"value as layer name instead of the provided one."
+            )
+            return None
 
     def _config(self) -> Dict[str, Any]:
         config = super()._config()

src/pytorch_ie/taskmodules/transformer_text_classification.py

@@ -6,7 +6,7 @@
         -> task_output
     -> document
 """
-
+import logging
 from typing import (
     Any,
     Dict,
@@ -33,6 +33,8 @@
 from pytorch_ie.documents import TextDocument, TextDocumentWithLabel, TextDocumentWithMultiLabel
 from pytorch_ie.models.transformer_text_classification import ModelOutputType, ModelStepInputType
 
+logger = logging.getLogger(__name__)
+
 InputEncodingType: TypeAlias = MutableMapping[str, Any]
 TargetEncodingType: TypeAlias = Sequence[int]
 
@@ -74,7 +76,7 @@ def __init__(
         self,
         tokenizer_name_or_path: str,
         label_to_verbalizer: Dict[str, str],
-        annotation: str = "labels",
+        annotation: str = "label",
         padding: Union[bool, str, PaddingStrategy] = True,
         truncation: Union[bool, str, TruncationStrategy] = True,
         max_length: Optional[int] = None,
@@ -104,11 +106,22 @@ def __init__(
         self.multi_label = multi_label
 
     @property
-    def document_type(self) -> Type[TextDocument]:
+    def document_type(self) -> Optional[Type[TextDocument]]:
+        dt: Type[TextDocument]
         if self.multi_label:
-            return TextDocumentWithMultiLabel
+            dt = TextDocumentWithMultiLabel
         else:
-            return TextDocumentWithLabel
+            dt = TextDocumentWithLabel
+        if self.annotation == "label":
+            return dt
+        else:
+            logger.warning(
+                f"annotation={self.annotation} is "
+                f"not the default value ('label'), so the taskmodule {type(self).__name__} can not request "
+                f"the usual document type ({dt.__name__}) for auto-conversion because this has the bespoken "
+                f"default value as layer name instead of the provided one."
+            )
+            return None
 
     def _config(self) -> Dict[str, Any]:
         config = super()._config()

src/pytorch_ie/taskmodules/transformer_token_classification.py

@@ -95,11 +95,22 @@ def __init__(
         self.include_ill_formed_predictions = include_ill_formed_predictions
 
     @property
-    def document_type(self) -> Type[TextDocument]:
+    def document_type(self) -> Optional[Type[TextDocument]]:
+        dt: Type[TextDocument]
         if self.partition_annotation is not None:
-            return TextDocumentWithLabeledSpansAndLabeledPartitions
+            dt = TextDocumentWithLabeledSpansAndLabeledPartitions
         else:
-            return TextDocumentWithLabeledSpans
+            dt = TextDocumentWithLabeledSpans
+        if self.entity_annotation == "labeled_spans":
+            return dt
+        else:
+            logger.warning(
+                f"entity_annotation={self.entity_annotation} is "
+                f"not the default value ('labeled_spans'), so the taskmodule {type(self).__name__} can not request "
+                f"the usual document type ({dt.__name__}) for auto-conversion because this has the bespoken default "
+                f"value as layer name instead of the provided one."
+            )
+            return None
 
     def _config(self) -> Dict[str, Any]:
         config = super()._config()