Use new Instance.duplicate() method instead of deepcopy

CHANGELOG.md

@@ -13,7 +13,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
-- Nothing yet
+- Replaced `deepcopy` of `Instance`s with new `Instance.duplicate()` method.
 
 ### Added
 

allennlp_models/coref/predictors/coref.py

@@ -1,4 +1,3 @@
-from copy import deepcopy
 from typing import List, Dict
 
 from overrides import overrides
@@ -96,7 +95,7 @@ def predictions_to_labeled_instances(
         span_field: ListField = instance["spans"]  # type: ignore
         instances = []
         for cluster in predicted_clusters:
-            new_instance = deepcopy(instance)
+            new_instance = instance.duplicate()
             span_labels = [
                 0 if (span.span_start, span.span_end) in cluster else -1  # type: ignore
                 for span in span_field
@@ -108,7 +107,7 @@ def predictions_to_labeled_instances(
             instances.append(new_instance)
         if not instances:
             # No predicted clusters; we just give an empty coref prediction.
-            new_instance = deepcopy(instance)
+            new_instance = instance.duplicate()
             span_labels = [-1] * len(span_field)  # type: ignore
             new_instance.add_field(
                 "span_labels", SequenceLabelField(span_labels, span_field), self._model.vocab

allennlp_models/lm/predictors/masked_language_model.py

@@ -1,4 +1,3 @@
-from copy import deepcopy
 from typing import Dict
 
 from overrides import overrides
@@ -19,7 +18,7 @@ def predict(self, sentence_with_masks: str) -> JsonDict:
     def predictions_to_labeled_instances(
         self, instance: Instance, outputs: Dict[str, numpy.ndarray]
     ):
-        new_instance = deepcopy(instance)
+        new_instance = instance.duplicate()
         token_field: TextField = instance["tokens"]  # type: ignore
         mask_targets = [Token(target_top_k[0]) for target_top_k in outputs["words"]]
 

allennlp_models/lm/predictors/next_token_lm.py

@@ -1,4 +1,3 @@
-from copy import deepcopy
 from typing import Dict
 
 from overrides import overrides
@@ -19,7 +18,7 @@ def predict(self, sentence: str) -> JsonDict:
     def predictions_to_labeled_instances(
         self, instance: Instance, outputs: Dict[str, numpy.ndarray]
     ):
-        new_instance = deepcopy(instance)
+        new_instance = instance.duplicate()
         token_field: TextField = instance["tokens"]  # type: ignore
         mask_targets = [Token(target_top_k[0]) for target_top_k in outputs["words"]]
 

allennlp_models/pair_classification/predictors/decomposable_attention.py

@@ -1,5 +1,4 @@
 from typing import List, Dict
-from copy import deepcopy
 
 import numpy
 from overrides import overrides
@@ -51,7 +50,7 @@ def _json_to_instance(self, json_dict: JsonDict) -> Instance:
     def predictions_to_labeled_instances(
         self, instance: Instance, outputs: Dict[str, numpy.ndarray]
     ) -> List[Instance]:
-        new_instance = deepcopy(instance)
+        new_instance = instance.duplicate()
         label = numpy.argmax(outputs["label_logits"])
         # Skip indexing, we have integer representations of the strings "entailment", etc.
         new_instance.add_field("label", LabelField(int(label), skip_indexing=True))

allennlp_models/rc/predictors/bidaf.py

@@ -1,4 +1,3 @@
-from copy import deepcopy
 from typing import Dict, List
 
 from overrides import overrides
@@ -57,7 +56,7 @@ def _json_to_instance(self, json_dict: JsonDict) -> Instance:
     def predictions_to_labeled_instances(
         self, instance: Instance, outputs: Dict[str, numpy.ndarray]
     ) -> List[Instance]:
-        new_instance = deepcopy(instance)
+        new_instance = instance.duplicate()
         # For BiDAF
         if "best_span" in outputs:
             span_start_label = outputs["best_span"][0]

requirements.txt

@@ -1,4 +1,6 @@
-git+https://github.com/allenai/allennlp.git@master
+# git+https://github.com/allenai/allennlp.git@master
+# TODO: remove
+git+https://github.com/epwalsh/allennlp.git@text-field-deep-copy
 
 # For structured prediction.
 conllu==2.3.2