Add rudimentary version checks on model load

spacy/errors.py

@@ -115,6 +115,18 @@ class Warnings(object):
             "`spacy.gold.biluo_tags_from_offsets(nlp.make_doc(text), entities)`"
             " to check the alignment. Misaligned entities ('-') will be "
             "ignored during training.")
+    W031 = ("Model '{model}' ({model_version}) requires spaCy {version} and "
+            "is incompatible with the current spaCy version ({current}). This "
+            "may lead to unexpected results or runtime errors. To resolve "
+            "this, download a newer compatible model or retrain your custom "
+            "model with the current spaCy version. For more details and "
+            "available updates, run: python -m spacy validate")
+    W032 = ("Unable to determine model compatibility for model '{model}' "
+            "({model_version}) with the current spaCy version ({current}). "
+            "This may lead to unexpected results or runtime errors. To resolve "
+            "this, download a newer compatible model or retrain your custom "
+            "model with the current spaCy version. For more details and "
+            "available updates, run: python -m spacy validate")
 
 
 @add_codes

spacy/tests/test_misc.py

@@ -10,6 +10,7 @@ from spacy import prefer_gpu, require_gpu
 from spacy.compat import symlink_to, symlink_remove, path2str, is_windows
 from spacy._ml import PrecomputableAffine
 from subprocess import CalledProcessError
+from .util import make_tempdir
 
 
 @pytest.fixture
@@ -146,3 +147,32 @@ def test_load_model_blank_shortcut():
     assert nlp.pipeline == []
     with pytest.raises(ImportError):
         util.load_model("blank:fjsfijsdof")
+
+
+def test_load_model_version_compat():
+    """Test warnings for various spacy_version specifications in meta. Since
+    this is more of a hack for v2, manually specify the current major.minor
+    version to simplify test creation."""
+    nlp = util.load_model("blank:en")
+    assert nlp.meta["spacy_version"].startswith(">=2.3")
+    with make_tempdir() as d:
+        # no change: compatible
+        nlp.to_disk(d)
+        nlp2 = util.load_model(d)
+
+        # additional compatible upper pin
+        nlp.meta["spacy_version"] = ">=2.3.0,<2.4.0"
+        nlp.to_disk(d)
+        nlp2 = util.load_model(d)
+
+        # incompatible older version
+        nlp.meta["spacy_version"] = ">=2.2.5"
+        nlp.to_disk(d)
+        with pytest.warns(UserWarning):
+            nlp_reloaded = util.load_model(d)
+
+        # invalid version specification
+        nlp.meta["spacy_version"] = ">@#$%_invalid_version"
+        nlp.to_disk(d)
+        with pytest.warns(UserWarning):
+            nlp_reloaded = util.load_model(d)

spacy/util.py

@@ -17,6 +17,7 @@ import srsly
 import catalogue
 import sys
 import warnings
+from . import about
 
 try:
     import jsonschema
@@ -250,6 +251,33 @@ def get_model_meta(path):
     for setting in ["lang", "name", "version"]:
         if setting not in meta or not meta[setting]:
             raise ValueError(Errors.E054.format(setting=setting))
+    if "spacy_version" in meta:
+        about_major_minor = ".".join(about.__version__.split(".")[:2])
+        if about_major_minor is not None and not meta["spacy_version"].startswith(
+            ">=" + about_major_minor
+        ):
+            # try to simplify version requirements from model meta to vx.x
+            # for warning message
+            meta_spacy_version = "v" + ".".join(
+                meta["spacy_version"].replace(">=", "").split(".")[:2]
+            )
+            # if the format is unexpected, supply the full version
+            if not re.match(r"v\d+\.\d+", meta_spacy_version):
+                meta_spacy_version = meta["spacy_version"]
+            warn_msg = Warnings.W031.format(
+                model=meta["lang"] + "_" + meta["name"],
+                model_version=meta["version"],
+                version=meta_spacy_version,
+                current=about.__version__,
+            )
+            warnings.warn(warn_msg)
+    else:
+        warn_msg = Warnings.W032.format(
+            model=meta["lang"] + "_" + meta["name"],
+            model_version=meta["version"],
+            current=about.__version__,
+        )
+        warnings.warn(warn_msg)
     return meta