Update and extend tests for spacy v3.5.4 (#392)

* Update test for compatibility with spacy v3.5.4+ * Extend tests for sourced/replaced listeners
explosion · Jun 29, 2023 · 2c4c845 · 2c4c845
1 parent d3b532a
commit 2c4c845
Showing 1 changed file with 123 additions and 1 deletion.
diff --git a/spacy_transformers/tests/test_pipeline_component.py b/spacy_transformers/tests/test_pipeline_component.py
@@ -1,5 +1,7 @@
 import pytest
+from packaging.version import Version
 import torch
+import spacy
 from spacy.language import Language
 from spacy.training.example import Example
 from spacy.util import make_tempdir
@@ -195,7 +197,6 @@ def test_transformer_pipeline_tagger_senter_listener():
             tagger.add_label(tag)
 
     # Check that the Transformer component finds it listeners
-    assert transformer.listeners == []
     optimizer = nlp.initialize(lambda: train_examples)
     assert tagger_trf in transformer.listeners
 
@@ -492,3 +493,124 @@ def test_no_update_listener_in_predict():
 
     transformer.predict(docs)
     assert listener._backprop is not None
+
+
+@pytest.mark.skipif(
+    Version(spacy.__version__) < Version("3.5.4"), reason="Bug fixed in spaCy v3.5.4"
+)
+def test_source_replace_listeners():
+    """Test that a pipeline with a transformer+tagger+senter and some replaced
+    listeners runs and trains properly"""
+    orig_config = """
+    [nlp]
+    lang = "en"
+    pipeline = ["transformer","tagger","senter"]
+
+    [components]
+
+    [components.senter]
+    factory = "senter"
+
+    [components.senter.model]
+    @architectures = "spacy.Tagger.v1"
+    nO = null
+
+    [components.senter.model.tok2vec]
+    @architectures = "spacy-transformers.TransformerListener.v1"
+    grad_factor = 1.0
+    upstream = "transformer"
+
+    [components.senter.model.tok2vec.pooling]
+    @layers = "reduce_mean.v1"
+
+    [components.tagger]
+    factory = "tagger"
+
+    [components.tagger.model]
+    @architectures = "spacy.Tagger.v1"
+    nO = null
+
+    [components.tagger.model.tok2vec]
+    @architectures = "spacy-transformers.TransformerListener.v1"
+    grad_factor = 1.0
+    upstream = "transformer"
+
+    [components.tagger.model.tok2vec.pooling]
+    @layers = "reduce_mean.v1"
+
+    [components.transformer]
+    factory = "transformer"
+
+    [components.transformer.model]
+    @architectures = "spacy-transformers.TransformerModel.v3"
+    name = "distilbert-base-uncased"
+    """
+    orig_config = Config().from_str(cfg_string)
+    nlp = util.load_model_from_config(orig_config, auto_fill=True, validate=True)
+    assert nlp.pipe_names == ["transformer", "tagger", "senter"]
+    tagger = nlp.get_pipe("tagger")
+    train_examples = []
+    for t in TRAIN_DATA:
+        train_examples.append(Example.from_dict(nlp.make_doc(t[0]), t[1]))
+        for tag in t[1]["tags"]:
+            tagger.add_label(tag)
+    optimizer = nlp.initialize(lambda: train_examples)
+    assert nlp.get_pipe("transformer").listening_components == ["tagger", "senter"]
+    for i in range(2):
+        losses = {}
+        nlp.update(train_examples, sgd=optimizer, losses=losses)
+
+    with make_tempdir() as dir_path:
+        nlp.to_disk(dir_path)
+        base_model = str(dir_path)
+        new_config = {
+            "nlp": {
+                "lang": "en",
+                "pipeline": ["transformer", "tagger", "senter", "ner"],
+            },
+            "components": {
+                "transformer": {"source": base_model},
+                "tagger": {
+                    "source": base_model,
+                    "replace_listeners": ["model.tok2vec"],
+                },
+                "senter": {
+                    "source": base_model,
+                    "replace_listeners": ["model.tok2vec"],
+                },
+                "ner": {
+                    "factory": "ner",
+                    "model": {
+                        "@architectures": "spacy.TransitionBasedParser.v2",
+                        "state_type": "ner",
+                        "tok2vec": {
+                            "@architectures": "spacy-transformers.TransformerListener.v1",
+                            "grad_factor": 1.0,
+                            "upstream": "transformer",
+                            "pooling": {"@layers": "reduce_mean.v1"},
+                        },
+                    },
+                },
+            },
+        }
+        new_nlp = util.load_model_from_config(new_config, auto_fill=True)
+        for component in ("tagger", "senter"):
+            assert (
+                new_nlp.config["components"][component]["model"]["tok2vec"][
+                    "@architectures"
+                ]
+                == "spacy-transformers.Tok2VecTransformer.v3"
+            )
+        assert new_nlp.get_pipe("transformer").listening_components == ["ner"]
+
+        with make_tempdir() as new_dir_path:
+            new_nlp.to_disk(new_dir_path)
+            new_nlp_re = spacy.load(new_dir_path)
+            for component in ("tagger", "senter"):
+                assert (
+                    new_nlp.config["components"][component]["model"]["tok2vec"][
+                        "@architectures"
+                    ]
+                    == "spacy-transformers.Tok2VecTransformer.v3"
+                )
+            assert new_nlp_re.get_pipe("transformer").listening_components == ["ner"]