remove (working) unit tests with pre-trained models

spacy/tests/regression/test_issue4267.py

@@ -42,56 +42,6 @@ def test_issue4267():
  assert token.ent_iob == 2
 
 
-def test_multiple_ner():
- """ Test that 2 NERs can work in sequence: the second can overwrite O annotations """
- nlp = English()
-
- # 1: untrained NER - should set everything to O
- untrained_ner = nlp.create_pipe("ner")
- untrained_ner.add_label("SMURFS")
- nlp.add_pipe(untrained_ner, name="uner")
- nlp.begin_training()
-
- # 2 : trained NER - should set "Antti L Korhonen" to PERSON and "Finland" to GPE
- # TODO: can't really use as unit test, using statistical model
- trained_ner = spacy.load("en_core_web_lg").get_pipe("ner")
- nlp.add_pipe(trained_ner)
-
- doc = nlp("This is Antti L Korhonen speaking in Finland")
- expected_iobs = ["O", "O", "B", "I", "I", "O", "O", "B"]
- expected_types = ["", "", "PERSON", "PERSON", "PERSON", "", "", "GPE"]
- assert [token.ent_iob_ for token in doc] == expected_iobs
- assert [token.ent_type_ for token in doc] == expected_types
-
-
-def test_multiple_ner_2():
- """ Test that 2 NERs can work in sequence: the second respects the first annotations """
- nlp = English()
-
- # 2: untrained NER - should keep everything as is
- untrained_ner = nlp.create_pipe("ner")
- untrained_ner.add_label("SMURFS")
- # untrained_ner.add_label("PERSON")
- # untrained_ner.add_label("GPE")
- nlp.add_pipe(untrained_ner, name="uner")
- nlp.begin_training()
-
- # 1 : trained NER - should set "Antti Korhonen" to PERSON and "Finland" to GPE
- # TODO: can't really use as unit test, using statistical model
- trained_ner = spacy.load("en_core_web_lg").get_pipe("ner")
- nlp.add_pipe(trained_ner, before="uner")
-
- doc = nlp("This is Antti Korhonen speaking in Finland")
-
- # Because the untrained NER can do whatever it wants, we can't make assumptions on the other tokens
- assert doc[2].ent_iob_ == "B"
- assert doc[2].ent_type_ == "PERSON"
- assert doc[3].ent_iob_ == "I"
- assert doc[3].ent_type_ == "PERSON"
- assert doc[6].ent_iob_ == "B"
- assert doc[6].ent_type_ == "GPE"
-
-
 def test_ruler_before_ner():
  """ Test that an NER works after an entity_ruler: the second can add annotations """
  nlp = English()
@@ -181,66 +131,6 @@ def test_block_ner():
  assert [token.ent_type_ for token in doc] == expected_types
 
 
-def test_preset_ner_1():
- """ Test that an NER will respect pre-set tokens (single-token entity) """
- nlp = English()
-
- # 1: preset "Antti" as PEEPZ
- nlp.add_pipe(PresetComponent(2, 3))
-
- # 2 : trained NER - should ignore "Antti" and set "Finland" to GPE
- # TODO: can't really use as unit test, using statistical model
- trained_ner = spacy.load("en_core_web_lg").get_pipe("ner")
- nlp.add_pipe(trained_ner)
-
- doc = nlp("This is Antti speaking in Finland")
-
- expected_iobs = ["O", "O", "B", "O", "O", "B"]
- expected_types = ["", "", "PEEPZ", "", "", "GPE"]
- assert [token.ent_iob_ for token in doc] == expected_iobs
- assert [token.ent_type_ for token in doc] == expected_types
-
-
-def test_preset_ner_2():
- """ Test that an NER will respect pre-set tokens (multiple-token entity) """
- nlp = English()
-
- # 1: preset "Antti Korhonen" as PEEPZ
- nlp.add_pipe(PresetComponent(2, 4))
-
- # 2 : trained NER - should ignore "Antti Korhonen" and set "Finland" to GPE
- # TODO: can't really use as unit test, using statistical model
- trained_ner = spacy.load("en_core_web_lg").get_pipe("ner")
- nlp.add_pipe(trained_ner)
-
- doc = nlp("This is Antti Korhonen speaking in Finland")
-
- expected_iobs = ["O", "O", "B", "I", "O", "O", "B"]
- expected_types = ["", "", "PEEPZ", "PEEPZ", "", "", "GPE"]
- assert [token.ent_iob_ for token in doc] == expected_iobs
- assert [token.ent_type_ for token in doc] == expected_types
-
-
-def test_preset_ner_3():
- """ Test that an NER will respect pre-set tokens (multiple-token entity) """
- nlp = English()
-
- # 1: preset "Antti L Korhonen" as PEEPZ
- nlp.add_pipe(PresetComponent(2, 5))
-
- # 2 : trained NER - should ignore "Antti Korhonen" and set "Finland" to GPE
- # TODO: can't really use as unit test, using statistical model
- trained_ner = spacy.load("en_core_web_lg").get_pipe("ner")
- nlp.add_pipe(trained_ner)
-
- doc = nlp("This is Antti L Korhonen speaking in Finland")
-
- expected_iobs = ["O", "O", "B", "I", "I", "O", "O", "B"]
- expected_types = ["", "", "PEEPZ", "PEEPZ", "PEEPZ", "", "", "GPE"]
- assert [token.ent_iob_ for token in doc] == expected_iobs
- assert [token.ent_type_ for token in doc] == expected_types
-
-
 class BlockerComponent1(object):
  name = "my_blocker"