Update spaCy pipeline

README.md

@@ -45,5 +45,7 @@ model-index:
 | `F1_CONNECTED-WITH` | 13.81 |
 | `F1_IN-MANNER-OF` | 11.96 |
 | `F1_ATTRIBUTE-FOR` | 17.36 |
+| `F1_MACRO` | 0.00 |
+| `F1_WEIGHTED` | 0.00 |
 | `TRANSFORMER_LOSS` | 0.77 |
 | `RELATION_EXTRACTOR_LOSS` | 111.45 |

relationFactory.py

@@ -1,34 +1,28 @@
-from itertools import islice
 from typing import Tuple, List, Iterable, Optional, Dict, Callable, Any
 
-from spacy.scorer import PRFScore
-from thinc.types import Floats2d
+import json
+import os
+import time
+from itertools import islice
+from pathlib import Path
+
+import spacy
+from spacy.tokens import Doc, Span
+from thinc.types import Floats2d, Ints1d, Ragged, cast
+from thinc.api import Model, Linear, chain, Logistic, Optimizer
+
+from sklearn.metrics import precision_recall_fscore_support, f1_score
 import numpy
 from spacy.training.example import Example
-from thinc.api import Model, Optimizer
-from spacy.tokens.doc import Doc
 from spacy.pipeline.trainable_pipe import TrainablePipe
 from spacy.vocab import Vocab
 from spacy import Language
 from thinc.model import set_dropout_rate
 from wasabi import Printer
-
-from typing import List, Tuple, Callable
-
-import spacy
-from spacy.tokens import Doc, Span
-from thinc.types import Floats2d, Ints1d, Ragged, cast
-from thinc.api import Model, Linear, chain, Logistic
-
-import json
-import os
-import time
-from pathlib import Path
-
-from sklearn.metrics import precision_recall_fscore_support, f1_score
 import plotly.express as px
 import plotly.graph_objects as go
 
+
 @spacy.registry.architectures("rel_model.v1")
 def create_relation_model(
     create_instance_tensor: Model[List[Doc], Floats2d],
@@ -270,17 +264,6 @@ class RelationExtractor(TrainablePipe):
             self.set_annotations(docs, predictions)
         return losses
 
-    def get_focal_loss(self, examples: Iterable[Example], scores, gamma=3.0, alpha=0.25, eps=1e-8) -> Tuple[float, float]:
-        truths = self._examples_to_truth(examples)
-        scores_2 = numpy.clip(scores, eps, 1. - eps)
-        p_t = numpy.clip(scores_2 * truths + (1 - scores_2) * (1 - truths), eps, 1. - eps)
-
-        focal_loss = -(1 - p_t) ** gamma * numpy.log(p_t)
-        loss = numpy.mean(numpy.sum(focal_loss, axis=1))
-        gradient = focal_loss * (1 - 2 * truths)
-        return float(loss), gradient
-    
-
     def get_loss(self, examples: Iterable[Example], scores) -> Tuple[float, float]:
         """Find the loss and gradient of loss for the batch of documents and
         their predicted scores."""

ru_patents_rel-any-py3-none-any.whl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:01a4762ef635162b3d38a964bf53b46c50d350f75b69c96116caecc4e4660464
-size 661156608
+oid sha256:123967bb99f193af5288262919e0a68942706cbe9acc5e3a27b22fb6dc6bfa31
+size 661156390