Feat/zshot version (#17)

* 🎨 Improved structure of setup and init.
* ✏️ Fixed minor typos and format in evaluator
* ✅ Update evaluation tests to work with latest version of evaluate
* 🐛 Fixed bug while importing version

Author: marmg

setup.cfg

@@ -1,2 +1,5 @@
 [egg_info]
 tag_svn_revision = true
+
+[metadata]
+version = attr: zshot.__version__

setup.py

@@ -4,10 +4,8 @@
 this_directory = Path(__file__).parent
 long_description = (this_directory / "README.md").read_text()
 
-version = '0.0.2'
 
 setup(name='zshot',
-      version=version,
       description="Zero and Few shot named entity recognition",
       long_description_content_type='text/markdown',
       long_description=long_description,

zshot/__init__.py

@@ -1,2 +1,4 @@
 from zshot.zshot import MentionsExtractor, Linker, Zshot, PipelineConfig  # noqa: F401
 from zshot.utils.displacy import displacy  # noqa: F401
+
+__version__ = '0.0.3'

zshot/evaluation/evaluator.py

@@ -34,7 +34,8 @@ def prepare_pipeline(
             feature_extractor=None,  # noqa: F821
             device: int = None,
     ):
-        pipe = super(TokenClassificationEvaluator, self).prepare_pipeline(model_or_pipeline, tokenizer, feature_extractor, device)
+        pipe = super(TokenClassificationEvaluator, self).prepare_pipeline(model_or_pipeline, tokenizer,
+                                                                          feature_extractor, device)
         return pipe
 
 

zshot/evaluation/zshot_evaluate.py

@@ -5,13 +5,12 @@
 from prettytable import PrettyTable
 
 from zshot.evaluation import load_medmentions, load_ontonotes
-from zshot.evaluation.dataset.dataset import DatasetWithEntities
 from zshot.evaluation.evaluator import ZeroShotTokenClassificationEvaluator, MentionsExtractorEvaluator
 from zshot.evaluation.pipeline import LinkerPipeline, MentionsExtractorPipeline
 
 
 def evaluate(nlp: spacy.language.Language,
-             datasets: Union[DatasetWithEntities, List[DatasetWithEntities]],
+             datasets: Union[str, List[str]],
              splits: Optional[Union[str, List[str]]] = None,
              metric: Optional[Union[str, EvaluationModule]] = None,
              batch_size: Optional[int] = 16) -> str:
@@ -31,6 +30,9 @@ def evaluate(nlp: spacy.language.Language,
     if type(splits) == str:
         splits = [splits]
 
+    if type(datasets) == str:
+        datasets = [datasets]
+
     result = {}
     field_names = ["Metric"]
     for dataset_name in datasets:

zshot/tests/evaluation/test_evaluation.py

@@ -113,7 +113,7 @@ def test_prediction_token_based_evaluation_all_matching(self):
         dataset = get_dataset(gt, sentences)
 
         custom_evaluator = ZeroShotTokenClassificationEvaluator("token-classification")
-        metrics = custom_evaluator.compute(get_linker_pipe([('New York', 'FAC', 1)]), dataset, "seqeval")
+        metrics = custom_evaluator.compute(get_linker_pipe([('New York', 'FAC', 1)]), dataset, metric="seqeval")
 
         assert float(metrics["overall_precision"]) == 1.0
         assert float(metrics["overall_precision"]) == 1.0
@@ -128,7 +128,7 @@ def test_prediction_token_based_evaluation_overlapping_spans(self):
 
         custom_evaluator = ZeroShotTokenClassificationEvaluator("token-classification")
         metrics = custom_evaluator.compute(get_linker_pipe([('New York', 'FAC', 1), ('York', 'LOC', 0.7)]), dataset,
-                                           "seqeval")
+                                           metric="seqeval")
 
         assert float(metrics["overall_precision"]) == 1.0
         assert float(metrics["overall_precision"]) == 1.0
@@ -144,7 +144,7 @@ def test_prediction_token_based_evaluation_partial_match_spans_expand(self):
         custom_evaluator = ZeroShotTokenClassificationEvaluator("token-classification",
                                                                 alignment_mode=AlignmentMode.expand)
         pipe = get_linker_pipe([('New Yo', 'FAC', 1)])
-        metrics = custom_evaluator.compute(pipe, dataset, "seqeval")
+        metrics = custom_evaluator.compute(pipe, dataset, metric="seqeval")
 
         assert float(metrics["overall_precision"]) == 1.0
         assert float(metrics["overall_precision"]) == 1.0
@@ -160,7 +160,7 @@ def test_prediction_token_based_evaluation_partial_match_spans_contract(self):
         custom_evaluator = ZeroShotTokenClassificationEvaluator("token-classification",
                                                                 alignment_mode=AlignmentMode.contract)
         pipe = get_linker_pipe([('New York i', 'FAC', 1)])
-        metrics = custom_evaluator.compute(pipe, dataset, "seqeval")
+        metrics = custom_evaluator.compute(pipe, dataset, metric="seqeval")
 
         assert float(metrics["overall_precision"]) == 1.0
         assert float(metrics["overall_precision"]) == 1.0
@@ -176,7 +176,7 @@ def test_prediction_token_based_evaluation_partial_and_overlapping_spans(self):
         custom_evaluator = ZeroShotTokenClassificationEvaluator("token-classification",
                                                                 alignment_mode=AlignmentMode.contract)
         pipe = get_linker_pipe([('New York i', 'FAC', 1), ('w York', 'LOC', 0.7)])
-        metrics = custom_evaluator.compute(pipe, dataset, "seqeval")
+        metrics = custom_evaluator.compute(pipe, dataset, metric="seqeval")
 
         assert float(metrics["overall_precision"]) == 1.0
         assert float(metrics["overall_precision"]) == 1.0
@@ -207,7 +207,8 @@ def test_prediction_token_based_evaluation_all_matching(self):
         dataset = get_dataset(gt, sentences)
 
         custom_evaluator = MentionsExtractorEvaluator("token-classification")
-        metrics = custom_evaluator.compute(get_mentions_extractor_pipe([('New York', 'FAC', 1)]), dataset, "seqeval")
+        metrics = custom_evaluator.compute(get_mentions_extractor_pipe([('New York', 'FAC', 1)]), dataset,
+                                           metric="seqeval")
 
         assert float(metrics["overall_precision"]) == 1.0
         assert float(metrics["overall_precision"]) == 1.0
@@ -222,7 +223,7 @@ def test_prediction_token_based_evaluation_overlapping_spans(self):
 
         custom_evaluator = MentionsExtractorEvaluator("token-classification")
         metrics = custom_evaluator.compute(get_mentions_extractor_pipe([('New York', 'FAC', 1), ('York', 'LOC', 0.7)]),
-                                           dataset, "seqeval")
+                                           dataset, metric="seqeval")
 
         assert float(metrics["overall_precision"]) == 1.0
         assert float(metrics["overall_precision"]) == 1.0
@@ -238,7 +239,7 @@ def test_prediction_token_based_evaluation_partial_match_spans_expand(self):
         custom_evaluator = MentionsExtractorEvaluator("token-classification",
                                                       alignment_mode=AlignmentMode.expand)
         pipe = get_mentions_extractor_pipe([('New Yo', 'FAC', 1)])
-        metrics = custom_evaluator.compute(pipe, dataset, "seqeval")
+        metrics = custom_evaluator.compute(pipe, dataset, metric="seqeval")
 
         assert float(metrics["overall_precision"]) == 1.0
         assert float(metrics["overall_precision"]) == 1.0