From 6f821efaf31880600f20a0a153f2ffb085c34be1 Mon Sep 17 00:00:00 2001
From: Matthew Honnibal <honnibal+gh@gmail.com>
Date: Sat, 10 Jun 2023 16:53:59 +0200
Subject: [PATCH 1/9] Add errors for pipe instance problems

---
 spacy/errors.py | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/spacy/errors.py b/spacy/errors.py
index 40cfa8d92..f9bee07c9 100644
--- a/spacy/errors.py
+++ b/spacy/errors.py
@@ -215,6 +215,11 @@ class Warnings(metaclass=ErrorsWithCodes):
     W123 = ("Argument `enable` with value {enable} does not contain all values specified in the config option "
             "`enabled` ({enabled}). Be aware that this might affect other components in your pipeline.")
     W124 = ("{host}:{port} is already in use, using the nearest available port {serve_port} as an alternative.")
+    W125 = (
+        "Pipe instance '{name}' is being added with a vocab "
+        "instance that will not match other components. This is "
+        "usually an error."
+    )
 
 
 class Errors(metaclass=ErrorsWithCodes):
@@ -970,6 +975,7 @@ class Errors(metaclass=ErrorsWithCodes):
     E1050 = ("Port {port} is already in use. Please specify an available port with `displacy.serve(doc, port=port)` "
              "or use `auto_select_port=True` to pick an available port automatically.")
     E1051 = ("'allow_overlap' can only be False when max_positive is 1, but found 'max_positive': {max_positive}.")
+    E1052 = ("Cannot create Language instance from config: missing pipeline components. The following components were added by instance (rather than config) via the 'Language.add_pipe_instance()' method, but are not present in the 'pipe_instances' variable: {names}")
 
 
 # Deprecated model shortcuts, only used in errors and warnings

From aa0d747739bf361844d38311be6834b19d6bf5ca Mon Sep 17 00:00:00 2001
From: Matthew Honnibal <honnibal+gh@gmail.com>
Date: Sat, 10 Jun 2023 16:55:13 +0200
Subject: [PATCH 2/9] Support adding pipeline component by instance

---
 spacy/language.py | 115 ++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 112 insertions(+), 3 deletions(-)

diff --git a/spacy/language.py b/spacy/language.py
index 9fdcf6328..7e52c60c3 100644
--- a/spacy/language.py
+++ b/spacy/language.py
@@ -1,6 +1,6 @@
 from typing import Iterator, Optional, Any, Dict, Callable, Iterable
-from typing import Union, Tuple, List, Set, Pattern, Sequence
-from typing import NoReturn, TYPE_CHECKING, TypeVar, cast, overload
+from typing import Union, Tuple, List, Set, Pattern, Sequence, overload
+from typing import NoReturn, TYPE_CHECKING, TypeVar, cast
 
 from dataclasses import dataclass
 import random
@@ -52,6 +52,9 @@ DEFAULT_CONFIG = util.load_config(DEFAULT_CONFIG_PATH)
 # This is the base config for the [pretraining] block and currently not included
 # in the main config and only added via the 'init fill-config' command
 DEFAULT_CONFIG_PRETRAIN_PATH = Path(__file__).parent / "default_config_pretraining.cfg"
+# Factory name indicating that the component wasn't constructed by a factory,
+# and was instead passed by instance
+INSTANCE_FACTORY_NAME = "__added_by_instance__"
 
 # Type variable for contexts piped with documents
 _AnyContext = TypeVar("_AnyContext")
@@ -743,6 +746,9 @@ class Language:
         """Add a component to the processing pipeline. Valid components are
         callables that take a `Doc` object, modify it and return it. Only one
         of before/after/first/last can be set. Default behaviour is "last".
+        Components can be added either by factory name or by instance. If
+        an instance is supplied and you serialize the pipeline, you'll need
+        to also pass an instance into spacy.load() to construct the pipeline.
 
         factory_name (str): Name of the component factory.
         name (str): Name of pipeline component. Overwrites existing
@@ -790,11 +796,58 @@ class Language:
                 raw_config=raw_config,
                 validate=validate,
             )
-        pipe_index = self._get_pipe_index(before, after, first, last)
         self._pipe_meta[name] = self.get_factory_meta(factory_name)
+        pipe_index = self._get_pipe_index(before, after, first, last)
         self._components.insert(pipe_index, (name, pipe_component))
         return pipe_component
 
+    def add_pipe_instance(self, component: PipeCallable,
+        /, name: Optional[str] = None,
+        *,
+        before: Optional[Union[str, int]] = None,
+        after: Optional[Union[str, int]] = None,
+        first: Optional[bool] = None,
+        last: Optional[bool] = None,
+    ) -> PipeCallable:
+        """Add a component instance to the processing pipeline. Valid components
+        are callables that take a `Doc` object, modify it and return it. Only one
+        of before/after/first/last can be set. Default behaviour is "last".
+
+        A limitation of this method is that spaCy will not know how to reconstruct
+        your pipeline after you save it out (unlike the 'Language.add_pipe()' method,
+        where you provide a config and let spaCy construct the instance). See 'spacy.load'
+        for details of how to load back a pipeline with components added by instance.
+
+        pipe_instance (Callable[[Doc], Doc]): The component to add.
+        name (str): Name of pipeline component. Overwrites existing
+            component.name attribute if available. If no name is set and
+            the component exposes no name attribute, component.__name__ is
+            used. An error is raised if a name already exists in the pipeline.
+        before (Union[str, int]): Name or index of the component to insert new
+            component directly before.
+        after (Union[str, int]): Name or index of the component to insert new
+            component directly after.
+        first (bool): If True, insert component first in the pipeline.
+        last (bool): If True, insert component last in the pipeline.
+        RETURNS (Callable[[Doc], Doc]): The pipeline component.
+
+        DOCS: https://spacy.io/api/language#add_pipe_instance
+        """
+        name = name if name is not None else getattr(component, "name")
+        if name is None:
+            raise ValueError("TODO error")
+        if name in self.component_names:
+            raise ValueError(Errors.E007.format(name=name, opts=self.component_names))
+
+        # It would be possible to take arguments for the FactoryMeta here, but we'll then have
+        # a problem on deserialization: where will the data be coming from?
+        # I think if someone wants that, they should register a component function.
+        self._pipe_meta[name] = FactoryMeta(INSTANCE_FACTORY_NAME)
+        self._pipe_configs[name] = Config()
+        pipe_index = self._get_pipe_index(before, after, first, last)
+        self._components.insert(pipe_index, (name, component))
+        return component
+
     def _get_pipe_index(
         self,
         before: Optional[Union[str, int]] = None,
@@ -1690,6 +1743,7 @@ class Language:
         meta: Dict[str, Any] = SimpleFrozenDict(),
         auto_fill: bool = True,
         validate: bool = True,
+        pipe_instances: Dict[str, Any] = SimpleFrozenDict()
     ) -> "Language":
         """Create the nlp object from a loaded config. Will set up the tokenizer
         and language data, add pipeline components etc. If no config is provided,
@@ -1765,6 +1819,11 @@ class Language:
 
         # Warn about require_gpu usage in jupyter notebook
         warn_if_jupyter_cupy()
+        # If we've been passed pipe instances, check whether
+        # they have a Vocab instance, and if they do, use
+        # that one. This also performs some additional checks and
+        # warns if there's a mismatch.
+        vocab = _get_instantiated_vocab(vocab, pipe_instances)
 
         # Note that we don't load vectors here, instead they get loaded explicitly
         # inside stuff like the spacy train function. If we loaded them here,
@@ -1781,6 +1840,11 @@ class Language:
         interpolated = filled.interpolate() if not filled.is_interpolated else filled
         pipeline = interpolated.get("components", {})
         sourced = util.get_sourced_components(interpolated)
+        # Check for components that aren't in the pipe_instances dict, aren't disabled,
+        # and aren't built by factory.
+        missing_components = _find_missing_components(pipeline, pipe_instances, exclude)
+        if missing_components:
+            raise ValueError(Errors.E1052.format(", ",join(missing_components)))
         # If components are loaded from a source (existing models), we cache
         # them here so they're only loaded once
         source_nlps = {}
@@ -1790,6 +1854,18 @@ class Language:
             if pipe_name not in pipeline:
                 opts = ", ".join(pipeline.keys())
                 raise ValueError(Errors.E956.format(name=pipe_name, opts=opts))
+            if pipe_name in pipe_instances:
+                if pipe_name in exclude:
+                    continue
+                else:
+                    nlp.add_pipe_instance(
+                        pipe_instances[pipe_name]
+                    )
+            # Is it important that we instantiate pipes that
+            # aren't excluded? It seems like we would want
+            # the exclude check above. I've left it how it
+            # is though, in case there's some sort of crazy
+            # load-bearing side-effects someone is relying on?
             pipe_cfg = util.copy_config(pipeline[pipe_name])
             raw_config = Config(filled["components"][pipe_name])
             if pipe_name not in exclude:
@@ -2306,3 +2382,36 @@ class _Sender:
         if self.count >= self.chunk_size:
             self.count = 0
             self.send()
+
+
+def _get_instantiated_vocab(vocab: Union[bool, Vocab], pipe_instances: Dict[str, Any]) -> Union[bool, Vocab]:
+    vocab_instances = {}
+    for name, instance in pipe_instances.items():
+        if hasattr(instance, "vocab") and isinstance(instance.vocab, Vocab):
+            vocab_instances[name] = instance.vocab
+    if not vocab_instances:
+        return vocab
+    elif isinstance(vocab, Vocab):
+        for name, inst_voc in vocab_instances.items():
+            if inst_voc is not vocab:
+                warnings.warn(Warnings.W125.format(name=name))
+        return vocab
+    else:
+        resolved_vocab = None
+        for name, inst_voc in vocab_instances.items():
+            if resolved_vocab is None:
+                resolved_vocab = inst_voc
+            elif inst_voc is not resolved_vocab:
+                warnings.warn(Warnings.W125.format(name=name))
+        # This is supposed to only be for the type checker --
+        # it should be unreachable
+        assert resolved_vocab is not None
+        return resolved_vocab
+
+
+def _find_missing_components(
+    pipeline: List[str],
+    pipe_instances: Dict[str, Any],
+    exclude: List[str]
+) -> List[str]:
+    return [name for name in pipeline if name not in pipe_instances and name not in exclude]

From 4332d12ce28807f1102f5dc81a285c959ce72fad Mon Sep 17 00:00:00 2001
From: Matthew Honnibal <honnibal+gh@gmail.com>
Date: Sat, 10 Jun 2023 16:55:52 +0200
Subject: [PATCH 3/9] Support adding pipeline component by instance

---
 spacy/__init__.py            |  2 ++
 spacy/tests/test_language.py | 33 +++++++++++++++++++++++++++++++++
 spacy/util.py                | 21 ++++++++++++++++++++-
 3 files changed, 55 insertions(+), 1 deletion(-)

diff --git a/spacy/__init__.py b/spacy/__init__.py
index c3568bc5c..995f965ae 100644
--- a/spacy/__init__.py
+++ b/spacy/__init__.py
@@ -35,6 +35,7 @@ def load(
     enable: Union[str, Iterable[str]] = util._DEFAULT_EMPTY_PIPES,
     exclude: Union[str, Iterable[str]] = util._DEFAULT_EMPTY_PIPES,
     config: Union[Dict[str, Any], Config] = util.SimpleFrozenDict(),
+    pipe_instances: Dict[str, Any] = util.SimpleFrozenDict(),
 ) -> Language:
     """Load a spaCy model from an installed package or a local path.
 
@@ -58,6 +59,7 @@ def load(
         enable=enable,
         exclude=exclude,
         config=config,
+        pipe_instances=pipe_instances,
     )
 
 
diff --git a/spacy/tests/test_language.py b/spacy/tests/test_language.py
index 236856dad..02e58d0a0 100644
--- a/spacy/tests/test_language.py
+++ b/spacy/tests/test_language.py
@@ -799,3 +799,36 @@ def test_component_return():
     nlp.add_pipe("test_component_bad_pipe")
     with pytest.raises(ValueError, match="instead of a Doc"):
         nlp("text")
+
+
+@pytest.mark.parametrize("components,kwargs,position", [
+    (["t1", "t2"], {"before": "t1"}, 0),
+    (["t1", "t2"], {"after": "t1"}, 1),
+    (["t1", "t2"], {"after": "t1"}, 1),
+    (["t1", "t2"], {"first": True}, 0),
+    (["t1", "t2"], {"last": True}, 2),
+    (["t1", "t2"], {"last": False}, 2),
+    (["t1", "t2"], {"first": False}, ValueError),
+])
+def test_add_pipe_instance(components, kwargs, position):
+    nlp = Language()
+    for name in components:
+        nlp.add_pipe("textcat", name=name)
+    pipe_names = list(nlp.pipe_names)
+    if isinstance(position, int):
+        result = nlp.add_pipe_instance(evil_component, name="new_component", **kwargs)
+        assert result is evil_component
+        pipe_names.insert(position, "new_component")
+        assert nlp.pipe_names == pipe_names
+    else:
+        with pytest.raises(ValueError):
+            result = nlp.add_pipe_instance(evil_component, name="new_component", **kwargs)
+
+
+def test_add_pipe_instance_to_bytes():
+    nlp = Language()
+    nlp.add_pipe("textcat", name="t1")
+    nlp.add_pipe("textcat", name="t2")
+    nlp.add_pipe_instance(evil_component, name="new_component")
+    b = nlp.to_bytes()
+ 
diff --git a/spacy/util.py b/spacy/util.py
index 8cc89217d..fce3f73be 100644
--- a/spacy/util.py
+++ b/spacy/util.py
@@ -415,6 +415,7 @@ def load_model(
     enable: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     config: Union[Dict[str, Any], Config] = SimpleFrozenDict(),
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict()
 ) -> "Language":
     """Load a model from a package or data path.
 
@@ -426,6 +427,9 @@ def load_model(
     exclude (Union[str, Iterable[str]]):  Name(s) of pipeline component(s) to exclude.
     config (Dict[str, Any] / Config): Config overrides as nested dict or dict
         keyed by section values in dot notation.
+    pipe_instances (Dict[str, Any]): Dictionary of components 
+        to be added to the pipeline directly (not created from
+        config)
     RETURNS (Language): The loaded nlp object.
     """
     kwargs = {
@@ -434,6 +438,7 @@ def load_model(
         "enable": enable,
         "exclude": exclude,
         "config": config,
+        "pipe_instances": pipe_instances
     }
     if isinstance(name, str):  # name or string path
         if name.startswith("blank:"):  # shortcut for blank model
@@ -457,6 +462,7 @@ def load_model_from_package(
     enable: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     config: Union[Dict[str, Any], Config] = SimpleFrozenDict(),
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict()
 ) -> "Language":
     """Load a model from an installed package.
 
@@ -472,10 +478,13 @@ def load_model_from_package(
         components won't be loaded.
     config (Dict[str, Any] / Config): Config overrides as nested dict or dict
         keyed by section values in dot notation.
+    pipe_instances (Dict[str, Any]): Dictionary of components 
+        to be added to the pipeline directly (not created from
+        config)
     RETURNS (Language): The loaded nlp object.
     """
     cls = importlib.import_module(name)
-    return cls.load(vocab=vocab, disable=disable, enable=enable, exclude=exclude, config=config)  # type: ignore[attr-defined]
+    return cls.load(vocab=vocab, disable=disable, enable=enable, exclude=exclude, config=config, pipe_instances=pipe_instances)  # type: ignore[attr-defined]
 
 
 def load_model_from_path(
@@ -487,6 +496,7 @@ def load_model_from_path(
     enable: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     config: Union[Dict[str, Any], Config] = SimpleFrozenDict(),
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict()
 ) -> "Language":
     """Load a model from a data directory path. Creates Language class with
     pipeline from config.cfg and then calls from_disk() with path.
@@ -504,6 +514,9 @@ def load_model_from_path(
         components won't be loaded.
     config (Dict[str, Any] / Config): Config overrides as nested dict or dict
         keyed by section values in dot notation.
+    pipe_instances (Dict[str, Any]): Dictionary of components
+        to be added to the pipeline directly (not created from
+        config)
     RETURNS (Language): The loaded nlp object.
     """
     if not model_path.exists():
@@ -520,6 +533,7 @@ def load_model_from_path(
         enable=enable,
         exclude=exclude,
         meta=meta,
+        pipe_instances=pipe_instances
     )
     return nlp.from_disk(model_path, exclude=exclude, overrides=overrides)
 
@@ -534,6 +548,7 @@ def load_model_from_config(
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     auto_fill: bool = False,
     validate: bool = True,
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict()
 ) -> "Language":
     """Create an nlp object from a config. Expects the full config file including
     a section "nlp" containing the settings for the nlp object.
@@ -551,6 +566,9 @@ def load_model_from_config(
         components won't be loaded.
     auto_fill (bool): Whether to auto-fill config with missing defaults.
     validate (bool): Whether to show config validation errors.
+    pipe_instances (Dict[str, Any]): Dictionary of components
+        to be added to the pipeline directly (not created from
+        config)
     RETURNS (Language): The loaded nlp object.
     """
     if "nlp" not in config:
@@ -570,6 +588,7 @@ def load_model_from_config(
         auto_fill=auto_fill,
         validate=validate,
         meta=meta,
+        pipe_instances=pipe_instances
     )
     return nlp
 

From b9730a64cb3a620be6635433f93ee3d6d870edef Mon Sep 17 00:00:00 2001
From: Matthew Honnibal <honnibal+gh@gmail.com>
Date: Sat, 10 Jun 2023 16:56:10 +0200
Subject: [PATCH 4/9] Format

---
 spacy/language.py            | 27 +++++++++++++++------------
 spacy/tests/test_language.py | 26 +++++++++++++++-----------
 spacy/util.py                | 18 +++++++++---------
 3 files changed, 39 insertions(+), 32 deletions(-)

diff --git a/spacy/language.py b/spacy/language.py
index 7e52c60c3..03d8f4dee 100644
--- a/spacy/language.py
+++ b/spacy/language.py
@@ -801,8 +801,11 @@ class Language:
         self._components.insert(pipe_index, (name, pipe_component))
         return pipe_component
 
-    def add_pipe_instance(self, component: PipeCallable,
-        /, name: Optional[str] = None,
+    def add_pipe_instance(
+        self,
+        component: PipeCallable,
+        /,
+        name: Optional[str] = None,
         *,
         before: Optional[Union[str, int]] = None,
         after: Optional[Union[str, int]] = None,
@@ -1743,7 +1746,7 @@ class Language:
         meta: Dict[str, Any] = SimpleFrozenDict(),
         auto_fill: bool = True,
         validate: bool = True,
-        pipe_instances: Dict[str, Any] = SimpleFrozenDict()
+        pipe_instances: Dict[str, Any] = SimpleFrozenDict(),
     ) -> "Language":
         """Create the nlp object from a loaded config. Will set up the tokenizer
         and language data, add pipeline components etc. If no config is provided,
@@ -1844,7 +1847,7 @@ class Language:
         # and aren't built by factory.
         missing_components = _find_missing_components(pipeline, pipe_instances, exclude)
         if missing_components:
-            raise ValueError(Errors.E1052.format(", ",join(missing_components)))
+            raise ValueError(Errors.E1052.format(", ", join(missing_components)))
         # If components are loaded from a source (existing models), we cache
         # them here so they're only loaded once
         source_nlps = {}
@@ -1858,9 +1861,7 @@ class Language:
                 if pipe_name in exclude:
                     continue
                 else:
-                    nlp.add_pipe_instance(
-                        pipe_instances[pipe_name]
-                    )
+                    nlp.add_pipe_instance(pipe_instances[pipe_name])
             # Is it important that we instantiate pipes that
             # aren't excluded? It seems like we would want
             # the exclude check above. I've left it how it
@@ -2384,7 +2385,9 @@ class _Sender:
             self.send()
 
 
-def _get_instantiated_vocab(vocab: Union[bool, Vocab], pipe_instances: Dict[str, Any]) -> Union[bool, Vocab]:
+def _get_instantiated_vocab(
+    vocab: Union[bool, Vocab], pipe_instances: Dict[str, Any]
+) -> Union[bool, Vocab]:
     vocab_instances = {}
     for name, instance in pipe_instances.items():
         if hasattr(instance, "vocab") and isinstance(instance.vocab, Vocab):
@@ -2410,8 +2413,8 @@ def _get_instantiated_vocab(vocab: Union[bool, Vocab], pipe_instances: Dict[str,
 
 
 def _find_missing_components(
-    pipeline: List[str],
-    pipe_instances: Dict[str, Any],
-    exclude: List[str]
+    pipeline: List[str], pipe_instances: Dict[str, Any], exclude: List[str]
 ) -> List[str]:
-    return [name for name in pipeline if name not in pipe_instances and name not in exclude]
+    return [
+        name for name in pipeline if name not in pipe_instances and name not in exclude
+    ]
diff --git a/spacy/tests/test_language.py b/spacy/tests/test_language.py
index 02e58d0a0..9c1d63e74 100644
--- a/spacy/tests/test_language.py
+++ b/spacy/tests/test_language.py
@@ -801,15 +801,18 @@ def test_component_return():
         nlp("text")
 
 
-@pytest.mark.parametrize("components,kwargs,position", [
-    (["t1", "t2"], {"before": "t1"}, 0),
-    (["t1", "t2"], {"after": "t1"}, 1),
-    (["t1", "t2"], {"after": "t1"}, 1),
-    (["t1", "t2"], {"first": True}, 0),
-    (["t1", "t2"], {"last": True}, 2),
-    (["t1", "t2"], {"last": False}, 2),
-    (["t1", "t2"], {"first": False}, ValueError),
-])
+@pytest.mark.parametrize(
+    "components,kwargs,position",
+    [
+        (["t1", "t2"], {"before": "t1"}, 0),
+        (["t1", "t2"], {"after": "t1"}, 1),
+        (["t1", "t2"], {"after": "t1"}, 1),
+        (["t1", "t2"], {"first": True}, 0),
+        (["t1", "t2"], {"last": True}, 2),
+        (["t1", "t2"], {"last": False}, 2),
+        (["t1", "t2"], {"first": False}, ValueError),
+    ],
+)
 def test_add_pipe_instance(components, kwargs, position):
     nlp = Language()
     for name in components:
@@ -822,7 +825,9 @@ def test_add_pipe_instance(components, kwargs, position):
         assert nlp.pipe_names == pipe_names
     else:
         with pytest.raises(ValueError):
-            result = nlp.add_pipe_instance(evil_component, name="new_component", **kwargs)
+            result = nlp.add_pipe_instance(
+                evil_component, name="new_component", **kwargs
+            )
 
 
 def test_add_pipe_instance_to_bytes():
@@ -831,4 +836,3 @@ def test_add_pipe_instance_to_bytes():
     nlp.add_pipe("textcat", name="t2")
     nlp.add_pipe_instance(evil_component, name="new_component")
     b = nlp.to_bytes()
- 
diff --git a/spacy/util.py b/spacy/util.py
index fce3f73be..75602dfbf 100644
--- a/spacy/util.py
+++ b/spacy/util.py
@@ -415,7 +415,7 @@ def load_model(
     enable: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     config: Union[Dict[str, Any], Config] = SimpleFrozenDict(),
-    pipe_instances: Dict[str, Any] = SimpleFrozenDict()
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict(),
 ) -> "Language":
     """Load a model from a package or data path.
 
@@ -427,7 +427,7 @@ def load_model(
     exclude (Union[str, Iterable[str]]):  Name(s) of pipeline component(s) to exclude.
     config (Dict[str, Any] / Config): Config overrides as nested dict or dict
         keyed by section values in dot notation.
-    pipe_instances (Dict[str, Any]): Dictionary of components 
+    pipe_instances (Dict[str, Any]): Dictionary of components
         to be added to the pipeline directly (not created from
         config)
     RETURNS (Language): The loaded nlp object.
@@ -438,7 +438,7 @@ def load_model(
         "enable": enable,
         "exclude": exclude,
         "config": config,
-        "pipe_instances": pipe_instances
+        "pipe_instances": pipe_instances,
     }
     if isinstance(name, str):  # name or string path
         if name.startswith("blank:"):  # shortcut for blank model
@@ -462,7 +462,7 @@ def load_model_from_package(
     enable: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     config: Union[Dict[str, Any], Config] = SimpleFrozenDict(),
-    pipe_instances: Dict[str, Any] = SimpleFrozenDict()
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict(),
 ) -> "Language":
     """Load a model from an installed package.
 
@@ -478,7 +478,7 @@ def load_model_from_package(
         components won't be loaded.
     config (Dict[str, Any] / Config): Config overrides as nested dict or dict
         keyed by section values in dot notation.
-    pipe_instances (Dict[str, Any]): Dictionary of components 
+    pipe_instances (Dict[str, Any]): Dictionary of components
         to be added to the pipeline directly (not created from
         config)
     RETURNS (Language): The loaded nlp object.
@@ -496,7 +496,7 @@ def load_model_from_path(
     enable: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     config: Union[Dict[str, Any], Config] = SimpleFrozenDict(),
-    pipe_instances: Dict[str, Any] = SimpleFrozenDict()
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict(),
 ) -> "Language":
     """Load a model from a data directory path. Creates Language class with
     pipeline from config.cfg and then calls from_disk() with path.
@@ -533,7 +533,7 @@ def load_model_from_path(
         enable=enable,
         exclude=exclude,
         meta=meta,
-        pipe_instances=pipe_instances
+        pipe_instances=pipe_instances,
     )
     return nlp.from_disk(model_path, exclude=exclude, overrides=overrides)
 
@@ -548,7 +548,7 @@ def load_model_from_config(
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     auto_fill: bool = False,
     validate: bool = True,
-    pipe_instances: Dict[str, Any] = SimpleFrozenDict()
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict(),
 ) -> "Language":
     """Create an nlp object from a config. Expects the full config file including
     a section "nlp" containing the settings for the nlp object.
@@ -588,7 +588,7 @@ def load_model_from_config(
         auto_fill=auto_fill,
         validate=validate,
         meta=meta,
-        pipe_instances=pipe_instances
+        pipe_instances=pipe_instances,
     )
     return nlp
 

From afbdd8259a12ada463ef5319f963f5e217670dc3 Mon Sep 17 00:00:00 2001
From: Matthew Honnibal <honnibal+gh@gmail.com>
Date: Sat, 10 Jun 2023 17:54:37 +0200
Subject: [PATCH 5/9] Fix find missing pipes

---
 spacy/language.py | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/spacy/language.py b/spacy/language.py
index 03d8f4dee..c4962ea45 100644
--- a/spacy/language.py
+++ b/spacy/language.py
@@ -1847,7 +1847,7 @@ class Language:
         # and aren't built by factory.
         missing_components = _find_missing_components(pipeline, pipe_instances, exclude)
         if missing_components:
-            raise ValueError(Errors.E1052.format(", ", join(missing_components)))
+            raise ValueError(Errors.E1052.format(names=", ".join(missing_components)))
         # If components are loaded from a source (existing models), we cache
         # them here so they're only loaded once
         source_nlps = {}
@@ -2413,8 +2413,10 @@ def _get_instantiated_vocab(
 
 
 def _find_missing_components(
-    pipeline: List[str], pipe_instances: Dict[str, Any], exclude: List[str]
+    pipeline: Dict[str, Dict[str, Any]], pipe_instances: Dict[str, Any], exclude: Iterable[str]
 ) -> List[str]:
-    return [
-        name for name in pipeline if name not in pipe_instances and name not in exclude
-    ]
+    missing = []
+    for name, config in pipeline.items():
+        if config.get("factory") == INSTANCE_FACTORY_NAME and name not in pipe_instances and name not in exclude:
+            missing.append(name)
+    return missing

From 77a08591ad48e0503012881b289391492e03df38 Mon Sep 17 00:00:00 2001
From: Matthew Honnibal <honnibal+gh@gmail.com>
Date: Tue, 13 Jun 2023 10:06:00 +0200
Subject: [PATCH 6/9] Format

---
 spacy/language.py | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/spacy/language.py b/spacy/language.py
index 6d867820a..dc4896e11 100644
--- a/spacy/language.py
+++ b/spacy/language.py
@@ -2419,10 +2419,16 @@ def _get_instantiated_vocab(
 
 
 def _find_missing_components(
-    pipeline: Dict[str, Dict[str, Any]], pipe_instances: Dict[str, Any], exclude: Iterable[str]
+    pipeline: Dict[str, Dict[str, Any]],
+    pipe_instances: Dict[str, Any],
+    exclude: Iterable[str],
 ) -> List[str]:
     missing = []
     for name, config in pipeline.items():
-        if config.get("factory") == INSTANCE_FACTORY_NAME and name not in pipe_instances and name not in exclude:
+        if (
+            config.get("factory") == INSTANCE_FACTORY_NAME
+            and name not in pipe_instances
+            and name not in exclude
+        ):
             missing.append(name)
     return missing

From 4cc5bd3ef5c471b8891c0398a047e4848ed6e8ee Mon Sep 17 00:00:00 2001
From: svlandeg <svlandeg@github.com>
Date: Mon, 26 Jun 2023 11:25:16 +0200
Subject: [PATCH 7/9] fix

---
 spacy/language.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/spacy/language.py b/spacy/language.py
index ab4637c79..0b50b058a 100644
--- a/spacy/language.py
+++ b/spacy/language.py
@@ -827,7 +827,6 @@ class Language:
     def add_pipe_instance(
         self,
         component: PipeCallable,
-        /,
         name: Optional[str] = None,
         *,
         before: Optional[Union[str, int]] = None,

From 9fcbc8eb67e0f088bd40dcdddd2c0fd187916a76 Mon Sep 17 00:00:00 2001
From: svlandeg <svlandeg@github.com>
Date: Mon, 26 Jun 2023 11:46:09 +0200
Subject: [PATCH 8/9] add pipe_instances also to load_model_from_init_py

---
 spacy/util.py | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/spacy/util.py b/spacy/util.py
index 35278c848..24be79fa6 100644
--- a/spacy/util.py
+++ b/spacy/util.py
@@ -675,6 +675,7 @@ def load_model_from_init_py(
     enable: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     exclude: Union[str, Iterable[str]] = _DEFAULT_EMPTY_PIPES,
     config: Union[Dict[str, Any], Config] = SimpleFrozenDict(),
+    pipe_instances: Dict[str, Any] = SimpleFrozenDict(),
 ) -> "Language":
     """Helper function to use in the `load()` method of a model package's
     __init__.py.
@@ -690,6 +691,9 @@ def load_model_from_init_py(
         components won't be loaded.
     config (Dict[str, Any] / Config): Config overrides as nested dict or dict
         keyed by section values in dot notation.
+    pipe_instances (Dict[str, Any]): Dictionary of components
+        to be added to the pipeline directly (not created from
+        config)
     RETURNS (Language): The loaded nlp object.
     """
     model_path = Path(init_file).parent

From 0fd797e33c6f4839389215bf1c83683229a49ca6 Mon Sep 17 00:00:00 2001
From: svlandeg <svlandeg@github.com>
Date: Wed, 5 Jul 2023 10:35:41 +0200
Subject: [PATCH 9/9] fix warning numbers

---
 spacy/errors.py   | 8 +++-----
 spacy/language.py | 4 ++--
 2 files changed, 5 insertions(+), 7 deletions(-)

diff --git a/spacy/errors.py b/spacy/errors.py
index b6f45dc0f..2a6653a5a 100644
--- a/spacy/errors.py
+++ b/spacy/errors.py
@@ -219,11 +219,9 @@ class Warnings(metaclass=ErrorsWithCodes):
     W125 = ("The StaticVectors key_attr is no longer used. To set a custom "
             "key attribute for vectors, configure it through Vectors(attr=) or "
             "'spacy init vectors --attr'")
-    W126 = (
-        "Pipe instance '{name}' is being added with a vocab "
-        "instance that will not match other components. This is "
-        "usually an error."
-    )
+    W126 = ("Pipe instance '{name}' is being added with a vocab "
+            "instance that will not match other components. This is "
+            "usually an error.")
 
 
 class Errors(metaclass=ErrorsWithCodes):
diff --git a/spacy/language.py b/spacy/language.py
index 439a8df87..da4530f53 100644
--- a/spacy/language.py
+++ b/spacy/language.py
@@ -2427,7 +2427,7 @@ def _get_instantiated_vocab(
     elif isinstance(vocab, Vocab):
         for name, inst_voc in vocab_instances.items():
             if inst_voc is not vocab:
-                warnings.warn(Warnings.W125.format(name=name))
+                warnings.warn(Warnings.W126.format(name=name))
         return vocab
     else:
         resolved_vocab = None
@@ -2435,7 +2435,7 @@ def _get_instantiated_vocab(
             if resolved_vocab is None:
                 resolved_vocab = inst_voc
             elif inst_voc is not resolved_vocab:
-                warnings.warn(Warnings.W125.format(name=name))
+                warnings.warn(Warnings.W126.format(name=name))
         # This is supposed to only be for the type checker --
         # it should be unreachable
         assert resolved_vocab is not None