spaCy/website/docs/api/top-level.md

title	menu
Top-level Functions	spacy spacy displacy displacy registry registry Loggers loggers Readers readers Batchers batchers Augmenters augmenters Callbacks callbacks Training & Alignment gold Utility Functions util

spaCy

spacy.load

Load a pipeline using the name of an installed package, a string path or a Path-like object. spaCy will try resolving the load argument in this order. If a pipeline is loaded from a string name, spaCy will assume it's a Python package and import it and call the package's own load() method. If a pipeline is loaded from a path, spaCy will assume it's a data directory, load its config.cfg and use the language and pipeline information to construct the Language class. The data will be loaded in via Language.from_disk.

As of v3.0, the disable keyword argument specifies components to load but disable, instead of components to not load at all. Those components can now be specified separately using the new exclude keyword argument.

Example

nlp = spacy.load("en_core_web_sm") # package
nlp = spacy.load("/path/to/pipeline") # string path
nlp = spacy.load(Path("/path/to/pipeline")) # pathlib Path

nlp = spacy.load("en_core_web_sm", exclude=["parser", "tagger"])

Name	Description
name	Pipeline to load, i.e. package name or path. Union[str, Path]
keyword-only
vocab	Optional shared vocab to pass in on initialization. If True (default), a new Vocab object will be created. Union[Vocab, bool]
disable	Name(s) of pipeline component(s) to disable. Disabled pipes will be loaded but they won't be run unless you explicitly enable them by calling nlp.enable_pipe. Union[str, Iterable[str]]
enable 3.4	Name(s) of pipeline component(s) to enable. All other pipes will be disabled. Union[str, Iterable[str]]
exclude 3	Name(s) of pipeline component(s) to exclude. Excluded components won't be loaded. Union[str, Iterable[str]]
config 3	Optional config overrides, either as nested dict or dict keyed by section value in dot notation, e.g. "components.name.value". Union[Dict[str, Any], Config]
RETURNS	A Language object with the loaded pipeline. Language

Essentially, spacy.load() is a convenience wrapper that reads the pipeline's config.cfg, uses the language and pipeline information to construct a Language object, loads in the model data and weights, and returns it.

### Abstract example
cls = spacy.util.get_lang_class(lang)  # 1. Get Language class, e.g. English
nlp = cls()                            # 2. Initialize it
for name in pipeline:
    nlp.add_pipe(name)                 # 3. Add the component to the pipeline
nlp.from_disk(data_path)               # 4. Load in the binary data

spacy.blank

Create a blank pipeline of a given language class. This function is the twin of spacy.load().

Example

nlp_en = spacy.blank("en")   # equivalent to English()
nlp_de = spacy.blank("de")   # equivalent to German()

Name	Description
name	IETF language tag, such as 'en', of the language class to load. str
keyword-only
vocab	Optional shared vocab to pass in on initialization. If True (default), a new Vocab object will be created. Union[Vocab, bool]
config 3	Optional config overrides, either as nested dict or dict keyed by section value in dot notation, e.g. "components.name.value". Union[Dict[str, Any], Config]
meta	Optional meta overrides for nlp.meta. Dict[str, Any]
RETURNS	An empty Language object of the appropriate subclass. Language

spacy.info

The same as the info command. Pretty-print information about your installation, installed pipelines and local setup from within spaCy.

Example

spacy.info()
spacy.info("en_core_web_sm")
markdown = spacy.info(markdown=True, silent=True)

Name	Description
model	Optional pipeline, i.e. a package name or path (optional). Optional[str]
keyword-only
markdown	Print information as Markdown. bool
silent	Don't print anything, just return. bool

spacy.explain

Get a description for a given POS tag, dependency label or entity type. For a list of available terms, see glossary.py.

Example

spacy.explain("NORP")
# Nationalities or religious or political groups

doc = nlp("Hello world")
for word in doc:
   print(word.text, word.tag_, spacy.explain(word.tag_))
# Hello UH interjection
# world NN noun, singular or mass

Name	Description
term	Term to explain. str
RETURNS	The explanation, or None if not found in the glossary. Optional[str]

spacy.prefer_gpu

Allocate data and perform operations on GPU, if available. If data has already been allocated on CPU, it will not be moved. Ideally, this function should be called right after importing spaCy and before loading any pipelines.

In a Jupyter notebook, run prefer_gpu() in the same cell as spacy.load() to ensure that the model is loaded on the correct device. See more details.

Example

import spacy
activated = spacy.prefer_gpu()
nlp = spacy.load("en_core_web_sm")

Name	Description
gpu_id	Device index to select. Defaults to 0. int
RETURNS	Whether the GPU was activated. bool

spacy.require_gpu

Allocate data and perform operations on GPU. Will raise an error if no GPU is available. If data has already been allocated on CPU, it will not be moved. Ideally, this function should be called right after importing spaCy and before loading any pipelines.

In a Jupyter notebook, run require_gpu() in the same cell as spacy.load() to ensure that the model is loaded on the correct device. See more details.

Example

import spacy
spacy.require_gpu()
nlp = spacy.load("en_core_web_sm")

Name	Description
gpu_id	Device index to select. Defaults to 0. int
RETURNS	True bool

spacy.require_cpu

Allocate data and perform operations on CPU. If data has already been allocated on GPU, it will not be moved. Ideally, this function should be called right after importing spaCy and before loading any pipelines.

In a Jupyter notebook, run require_cpu() in the same cell as spacy.load() to ensure that the model is loaded on the correct device. See more details.

Example

import spacy
spacy.require_cpu()
nlp = spacy.load("en_core_web_sm")

Name	Description
RETURNS	True bool

displaCy

As of v2.0, spaCy comes with a built-in visualization suite. For more info and examples, see the usage guide on visualizing spaCy.

displacy.serve

Serve a dependency parse tree or named entity visualization to view it in your browser. Will run a simple web server.

Example

import spacy
from spacy import displacy
nlp = spacy.load("en_core_web_sm")
doc1 = nlp("This is a sentence.")
doc2 = nlp("This is another sentence.")
displacy.serve([doc1, doc2], style="dep")

Name	Description
docs	Document(s) or span(s) to visualize. Union[Iterable[Union[Doc, Span]], Doc, Span]
style	Visualization style, "dep", "ent" or "span" 3.3. Defaults to "dep". str
page	Render markup as full HTML page. Defaults to True. bool
minify	Minify HTML markup. Defaults to False. bool
options	Visualizer-specific options, e.g. colors. Dict[str, Any]
manual	Don't parse Doc and instead expect a dict or list of dicts. See here for formats and examples. Defaults to False. bool
port	Port to serve visualization. Defaults to 5000. int
host	Host to serve visualization. Defaults to "0.0.0.0". str

displacy.render

Render a dependency parse tree or named entity visualization.

Example

import spacy
from spacy import displacy
nlp = spacy.load("en_core_web_sm")
doc = nlp("This is a sentence.")
html = displacy.render(doc, style="dep")

Name	Description
docs	Document(s) or span(s) to visualize. Union[Iterable[Union[Doc, Span, dict]], Doc, Span, dict]
style	Visualization style, "dep", "ent" or "span" 3.3. Defaults to "dep". str
page	Render markup as full HTML page. Defaults to True. bool
minify	Minify HTML markup. Defaults to False. bool
options	Visualizer-specific options, e.g. colors. Dict[str, Any]
manual	Don't parse Doc and instead expect a dict or list of dicts. See here for formats and examples. Defaults to False. bool
jupyter	Explicitly enable or disable "Jupyter mode" to return markup ready to be rendered in a notebook. Detected automatically if None (default). Optional[bool]
RETURNS	The rendered HTML markup. str

displacy.parse_deps

Generate dependency parse in {'words': [], 'arcs': []} format. For use with the manual=True argument in displacy.render.

Example

import spacy
from spacy import displacy
nlp = spacy.load("en_core_web_sm")
doc = nlp("This is a sentence.")
deps_parse = displacy.parse_deps(doc)
html = displacy.render(deps_parse, style="dep", manual=True)

Name	Description
orig_doc	Doc to parse dependencies. Doc
options	Dependency parse specific visualisation options. Dict[str, Any]
RETURNS	Generated dependency parse keyed by words and arcs. dict

displacy.parse_ents

Generate named entities in [{start: i, end: i, label: 'label'}] format. For use with the manual=True argument in displacy.render.

Example

import spacy
from spacy import displacy
nlp = spacy.load("en_core_web_sm")
doc = nlp("But Google is starting from behind.")
ents_parse = displacy.parse_ents(doc)
html = displacy.render(ents_parse, style="ent", manual=True)

Name	Description
doc	Doc to parse entities. Doc
options	NER-specific visualisation options. Dict[str, Any]
RETURNS	Generated entities keyed by text (original text) and ents. dict

displacy.parse_spans

Generate spans in [{start_token: i, end_token: i, label: 'label'}] format. For use with the manual=True argument in displacy.render.

Example

import spacy
from spacy import displacy
nlp = spacy.load("en_core_web_sm")
doc = nlp("But Google is starting from behind.")
doc.spans['orgs'] = [doc[1:2]]
ents_parse = displacy.parse_spans(doc, options={"spans_key" : "orgs"})
html = displacy.render(ents_parse, style="span", manual=True)

Name	Description
doc	Doc to parse entities. Doc
options	Span-specific visualisation options. Dict[str, Any]
RETURNS	Generated entities keyed by text (original text) and ents. dict

Visualizer options

The options argument lets you specify additional settings for each visualizer. If a setting is not present in the options, the default value will be used.

Dependency Visualizer options

Example

options = {"compact": True, "color": "blue"}
displacy.serve(doc, style="dep", options=options)

Name	Description
fine_grained	Use fine-grained part-of-speech tags (Token.tag_) instead of coarse-grained tags (Token.pos_). Defaults to False. bool
add_lemma 2.2.4	Print the lemmas in a separate row below the token texts. Defaults to False. bool
collapse_punct	Attach punctuation to tokens. Can make the parse more readable, as it prevents long arcs to attach punctuation. Defaults to True. bool
collapse_phrases	Merge noun phrases into one token. Defaults to False. bool
compact	"Compact mode" with square arrows that takes up less space. Defaults to False. bool
color	Text color (HEX, RGB or color names). Defaults to "#000000". str
bg	Background color (HEX, RGB or color names). Defaults to "#ffffff". str
font	Font name or font family for all text. Defaults to "Arial". str
offset_x	Spacing on left side of the SVG in px. Defaults to 50. int
arrow_stroke	Width of arrow path in px. Defaults to 2. int
arrow_width	Width of arrow head in px. Defaults to 10 in regular mode and 8 in compact mode. int
arrow_spacing	Spacing between arrows in px to avoid overlaps. Defaults to 20 in regular mode and 12 in compact mode. int
word_spacing	Vertical spacing between words and arcs in px. Defaults to 45. int
distance	Distance between words in px. Defaults to 175 in regular mode and 150 in compact mode. int

Named Entity Visualizer options

Example

options = {"ents": ["PERSON", "ORG", "PRODUCT"],
           "colors": {"ORG": "yellow"}}
displacy.serve(doc, style="ent", options=options)

Name	Description
ents	Entity types to highlight or None for all types (default). Optional[List[str]]
colors	Color overrides. Entity types should be mapped to color names or values. Dict[str, str]
template 2.2	Optional template to overwrite the HTML used to render entity spans. Should be a format string and can use {bg}, {text} and {label}. See templates.py for examples. Optional[str]
kb_url_template 3.2.1	Optional template to construct the KB url for the entity to link to. Expects a python f-string format with single field to fill in. Optional[str]

Span Visualizer options

Example

options = {"spans_key": "sc"}
displacy.serve(doc, style="span", options=options)

Name	Description
spans_key	Which spans key to render spans from. Default is "sc". str
templates	Dictionary containing the keys "span", "slice", and "start". These dictate how the overall span, a span slice, and the starting token will be rendered. Optional[Dict[str, str]
kb_url_template	Optional template to construct the KB url for the entity to link to. Expects a python f-string format with single field to fill in Optional[str]
colors	Color overrides. Entity types should be mapped to color names or values. Dict[str, str]

By default, displaCy comes with colors for all entity types used by spaCy's trained pipelines for both entity and span visualizer. If you're using custom entity types, you can use the colors setting to add your own colors for them. Your application or pipeline package can also expose a spacy_displacy_colors entry point to add custom labels and their colors automatically.

By default, displaCy links to # for entities without a kb_id set on their span. If you wish to link an entity to their URL then consider using the kb_url_template option from above. For example if the kb_id on a span is Q95 and this is a Wikidata identifier then this option can be set to https://www.wikidata.org/wiki/{}. Clicking on your entity in the rendered HTML should redirect you to their Wikidata page, in this case https://www.wikidata.org/wiki/Q95.

registry

spaCy's function registry extends Thinc's registry and allows you to map strings to functions. You can register functions to create architectures, optimizers, schedules and more, and then refer to them and set their arguments in your config file. Python type hints are used to validate the inputs. See the Thinc docs for details on the registry methods and our helper library catalogue for some background on the concept of function registries. spaCy also uses the function registry for language subclasses, model architecture, lookups and pipeline component factories.

Example

from typing import Iterator
import spacy

@spacy.registry.schedules("waltzing.v1")
def waltzing() -> Iterator[float]:
    i = 0
    while True:
        yield i % 3 + 1
        i += 1

Registry name	Description
architectures	Registry for functions that create model architectures. Can be used to register custom model architectures and reference them in the config.cfg.
augmenters	Registry for functions that create data augmentation callbacks for corpora and other training data iterators.
batchers	Registry for training and evaluation data batchers.
callbacks	Registry for custom callbacks to modify the nlp object before training.
displacy_colors	Registry for custom color scheme for the displacy NER visualizer. Automatically reads from entry points.
factories	Registry for functions that create pipeline components. Added automatically when you use the @spacy.component decorator and also reads from entry points.
initializers	Registry for functions that create initializers.
languages	Registry for language-specific Language subclasses. Automatically reads from entry points.
layers	Registry for functions that create layers.
loggers	Registry for functions that log training results.
lookups	Registry for large lookup tables available via vocab.lookups.
losses	Registry for functions that create losses.
misc	Registry for miscellaneous functions that return data assets, knowledge bases or anything else you may need.
optimizers	Registry for functions that create optimizers.
readers	Registry for file and data readers, including training and evaluation data readers like Corpus.
schedules	Registry for functions that create schedules.
scorers	Registry for functions that create scoring methods for user with the Scorer. Scoring methods are called with Iterable[Example] and arbitrary \*\*kwargs and return scores as Dict[str, Any].
tokenizers	Registry for tokenizer factories. Registered functions should return a callback that receives the nlp object and returns a Tokenizer or a custom callable.

spacy-transformers registry

The following registries are added by the spacy-transformers package. See the Transformer API reference and usage docs for details.

Example

import spacy_transformers

@spacy_transformers.registry.annotation_setters("my_annotation_setter.v1")
def configure_custom_annotation_setter():
    def annotation_setter(docs, trf_data) -> None:
       # Set annotations on the docs

    return annotation_setter

Registry name	Description
span_getters	Registry for functions that take a batch of Doc objects and return a list of Span objects to process by the transformer, e.g. sentences.
annotation_setters	Registry for functions that create annotation setters. Annotation setters are functions that take a batch of Doc objects and a FullTransformerBatch and can set additional annotations on the Doc.

Loggers

A logger records the training results. When a logger is created, two functions are returned: one for logging the information for each training step, and a second function that is called to finalize the logging when the training is finished. To log each training step, a dictionary is passed on from the spacy train, including information such as the training loss and the accuracy scores on the development set.

The built-in, default logger is the ConsoleLogger, which prints results to the console in tabular format and saves them to a jsonl file. The spacy-loggers package, included as a dependency of spaCy, enables other loggers, such as one that sends results to a Weights & Biases dashboard.

Instead of using one of the built-in loggers, you can implement your own.

spacy.ConsoleLogger.v2

Example config

[training.logger]
@loggers = "spacy.ConsoleLogger.v2"
progress_bar = true
console_output = true
output_file = "training_log.jsonl"

Writes the results of a training step to the console in a tabular format and saves them to a jsonl file.

$ python -m spacy train config.cfg

ℹ Using CPU
ℹ Loading config and nlp from: config.cfg
ℹ Pipeline: ['tok2vec', 'tagger']
ℹ Start training
ℹ Training. Initial learn rate: 0.0
ℹ Saving results to training_log.jsonl

E     #        LOSS TOK2VEC   LOSS TAGGER   TAG_ACC   SCORE
---   ------   ------------   -----------   -------   ------
  0        0           0.00         86.20      0.22     0.00
  0      200           3.08      18968.78     34.00     0.34
  0      400          31.81      22539.06     33.64     0.34
  0      600          92.13      22794.91     43.80     0.44
  0      800         183.62      21541.39     56.05     0.56
  0     1000         352.49      25461.82     65.15     0.65
  0     1200         422.87      23708.82     71.84     0.72
  0     1400         601.92      24994.79     76.57     0.77
  0     1600         662.57      22268.02     80.20     0.80
  0     1800        1101.50      28413.77     82.56     0.83
  0     2000        1253.43      28736.36     85.00     0.85
  0     2200        1411.02      28237.53     87.42     0.87
  0     2400        1605.35      28439.95     88.70     0.89

Note that the cumulative loss keeps increasing within one epoch, but should start decreasing across epochs.

Name	Description
progress_bar	Whether the logger should print the progress bar bool
console_output	Whether the logger should print the logs on the console. bool
output_file	The file to save the training logs to. Optional[Union[str, Path]]

Readers

File readers

The following file readers are provided by our serialization library srsly. All registered functions take one argument path, pointing to the file path to load.

Example config

[corpora.train.augmenter.orth_variants]
@readers = "srsly.read_json.v1"
path = "corpus/en_orth_variants.json"

Name	Description
srsly.read_json.v1	Read data from a JSON file.
srsly.read_jsonl.v1	Read data from a JSONL (newline-delimited JSON) file.
srsly.read_yaml.v1	Read data from a YAML file.
srsly.read_msgpack.v1	Read data from a binary MessagePack file.

Since the file readers expect a local path, you should only use them in config blocks that are not executed at runtime – for example, in [training] and [corpora] (to load data or resources like data augmentation tables) or in [initialize] (to pass data to pipeline components).

spacy.read_labels.v1

Read a JSON-formatted labels file generated with init labels. Typically used in the [initialize] block of the training config to speed up the model initialization process and provide pre-generated label sets.

Example config

[initialize.components]

[initialize.components.ner]

[initialize.components.ner.labels]
@readers = "spacy.read_labels.v1"
path = "corpus/labels/ner.json"

Name	Description
path	The path to the labels file generated with init labels. Path
require	Whether to require the file to exist. If set to False and the labels file doesn't exist, the loader will return None and the initialize method will extract the labels from the data. Defaults to False. bool
CREATES	The list of labels. List[str]

Corpus readers

Corpus readers are registered functions that load data and return a function that takes the current nlp object and yields Example objects that can be used for training and pretraining. You can replace it with your own registered function in the @readers registry to customize the data loading and streaming.

spacy.Corpus.v1

The Corpus reader manages annotated corpora and can be used for training and development datasets in the DocBin (.spacy) format. Also see the Corpus class.

Example config

[paths]
train = "corpus/train.spacy"

[corpora.train]
@readers = "spacy.Corpus.v1"
path = ${paths.train}
gold_preproc = false
max_length = 0
limit = 0

Name	Description
path	The directory or filename to read from. Expects data in spaCy's binary .spacy format. Union[str, Path]
gold_preproc	Whether to set up the Example object with gold-standard sentences and tokens for the predictions. See Corpus for details. bool
max_length	Maximum document length. Longer documents will be split into sentences, if sentence boundaries are available. Defaults to 0 for no limit. int
limit	Limit corpus to a subset of examples, e.g. for debugging. Defaults to 0 for no limit. int
augmenter	Apply some simply data augmentation, where we replace tokens with variations. This is especially useful for punctuation and case replacement, to help generalize beyond corpora that don't have smart-quotes, or only have smart quotes, etc. Defaults to None. Optional[Callable]
CREATES	The corpus reader. Corpus

spacy.JsonlCorpus.v1

Create Example objects from a JSONL (newline-delimited JSON) file of texts keyed by "text". Can be used to read the raw text corpus for language model pretraining from a JSONL file. Also see the JsonlCorpus class.

Example config

[paths]
pretrain = "corpus/raw_text.jsonl"

[corpora.pretrain]
@readers = "spacy.JsonlCorpus.v1"
path = ${paths.pretrain}
min_length = 0
max_length = 0
limit = 0

Name	Description
path	The directory or filename to read from. Expects newline-delimited JSON with a key "text" for each record. Union[str, Path]
min_length	Minimum document length (in tokens). Shorter documents will be skipped. Defaults to 0, which indicates no limit. int
max_length	Maximum document length (in tokens). Longer documents will be skipped. Defaults to 0, which indicates no limit. int
limit	Limit corpus to a subset of examples, e.g. for debugging. Defaults to 0 for no limit. int
CREATES	The corpus reader. JsonlCorpus

Batchers

A data batcher implements a batching strategy that essentially turns a stream of items into a stream of batches, with each batch consisting of one item or a list of items. During training, the models update their weights after processing one batch at a time. Typical batching strategies include presenting the training data as a stream of batches with similar sizes, or with increasing batch sizes. See the Thinc documentation on schedules for a few standard examples.

Instead of using one of the built-in batchers listed here, you can also implement your own, which may or may not use a custom schedule.

spacy.batch_by_words.v1

Create minibatches of roughly a given number of words. If any examples are longer than the specified batch length, they will appear in a batch by themselves, or be discarded if discard_oversize is set to True. The argument docs can be a list of strings, Doc objects or Example objects.

Example config

[training.batcher]
@batchers = "spacy.batch_by_words.v1"
size = 100
tolerance = 0.2
discard_oversize = false
get_length = null

Name	Description
seqs	The sequences to minibatch. Iterable[Any]
size	The target number of words per batch. Can also be a block referencing a schedule, e.g. compounding. Union[int, Sequence[int]]
tolerance	What percentage of the size to allow batches to exceed. float
discard_oversize	Whether to discard sequences that by themselves exceed the tolerated size. bool
get_length	Optional function that receives a sequence item and returns its length. Defaults to the built-in len() if not set. Optional[CallableAny], int
CREATES	The batcher that takes an iterable of items and returns batches. CallableIterable[Any, Iterable[List[Any]]]

spacy.batch_by_sequence.v1

Example config

[training.batcher]
@batchers = "spacy.batch_by_sequence.v1"
size = 32
get_length = null

Create a batcher that creates batches of the specified size.

Name	Description
size	The target number of items per batch. Can also be a block referencing a schedule, e.g. compounding. Union[int, Sequence[int]]
get_length	Optional function that receives a sequence item and returns its length. Defaults to the built-in len() if not set. Optional[CallableAny], int
CREATES	The batcher that takes an iterable of items and returns batches. CallableIterable[Any, Iterable[List[Any]]]

spacy.batch_by_padded.v1

Example config

[training.batcher]
@batchers = "spacy.batch_by_padded.v1"
size = 100
buffer = 256
discard_oversize = false
get_length = null

Minibatch a sequence by the size of padded batches that would result, with sequences binned by length within a window. The padded size is defined as the maximum length of sequences within the batch multiplied by the number of sequences in the batch.

Name	Description
size	The largest padded size to batch sequences into. Can also be a block referencing a schedule, e.g. compounding. Union[int, Sequence[int]]
buffer	The number of sequences to accumulate before sorting by length. A larger buffer will result in more even sizing, but if the buffer is very large, the iteration order will be less random, which can result in suboptimal training. int
discard_oversize	Whether to discard sequences that are by themselves longer than the largest padded batch size. bool
get_length	Optional function that receives a sequence item and returns its length. Defaults to the built-in len() if not set. Optional[CallableAny], int
CREATES	The batcher that takes an iterable of items and returns batches. CallableIterable[Any, Iterable[List[Any]]]

Augmenters

Data augmentation is the process of applying small modifications to the training data. It can be especially useful for punctuation and case replacement – for example, if your corpus only uses smart quotes and you want to include variations using regular quotes, or to make the model less sensitive to capitalization by including a mix of capitalized and lowercase examples. See the usage guide for details and examples.

spacy.orth_variants.v1

Example config

[corpora.train.augmenter]
@augmenters = "spacy.orth_variants.v1"
level = 0.1
lower = 0.5

[corpora.train.augmenter.orth_variants]
@readers = "srsly.read_json.v1"
path = "corpus/en_orth_variants.json"

Create a data augmentation callback that uses orth-variant replacement. The callback can be added to a corpus or other data iterator during training. It's especially useful for punctuation and case replacement, to help generalize beyond corpora that don't have smart quotes, or only have smart quotes etc.

Name	Description
level	The percentage of texts that will be augmented. float
lower	The percentage of texts that will be lowercased. float
orth_variants	A dictionary containing the single and paired orth variants. Typically loaded from a JSON file. See en_orth_variants.json for an example. Dict[str, Dict[List[Union[str, List[str]]]]]
CREATES	A function that takes the current nlp object and an Example and yields augmented Example objects. CallableLanguage, Example], Iterator[Example

spacy.lower_case.v1

Example config

[corpora.train.augmenter]
@augmenters = "spacy.lower_case.v1"
level = 0.3

Create a data augmentation callback that lowercases documents. The callback can be added to a corpus or other data iterator during training. It's especially useful for making the model less sensitive to capitalization.

Name	Description
level	The percentage of texts that will be augmented. float
CREATES	A function that takes the current nlp object and an Example and yields augmented Example objects. CallableLanguage, Example], Iterator[Example

Callbacks

The config supports callbacks at several points in the lifecycle that can be used modify the nlp object.

spacy.copy_from_base_model.v1

Example config

[initialize.before_init]
@callbacks = "spacy.copy_from_base_model.v1"
tokenizer = "en_core_sci_md"
vocab = "en_core_sci_md"

Copy the tokenizer and/or vocab from the specified models. It's similar to the v2 base model option and useful in combination with sourced components when fine-tuning an existing pipeline. The vocab includes the lookups and the vectors from the specified model. Intended for use in [initialize.before_init].

Name	Description
tokenizer	The pipeline to copy the tokenizer from. Defaults to None. Optional[str]
vocab	The pipeline to copy the vocab from. The vocab includes the lookups and vectors. Defaults to None. Optional[str]
CREATES	A function that takes the current nlp object and modifies its tokenizer and vocab. Callable[[Language], None]

spacy.models_with_nvtx_range.v1

Example config

[nlp]
after_pipeline_creation = {"@callbacks":"spacy.models_with_nvtx_range.v1"}

Recursively wrap the models in each pipe using NVTX range markers. These markers aid in GPU profiling by attributing specific operations to a Model's forward or backprop passes.

Name	Description
forward_color	Color identifier for forward passes. Defaults to -1. int
backprop_color	Color identifier for backpropagation passes. Defaults to -1. int
CREATES	A function that takes the current nlp and wraps forward/backprop passes in NVTX ranges. Callable[[Language], Language]

spacy.models_and_pipes_with_nvtx_range.v1

Example config

[nlp]
after_pipeline_creation = {"@callbacks":"spacy.models_and_pipes_with_nvtx_range.v1"}

Recursively wrap both the models and methods of each pipe using NVTX range markers. By default, the following methods are wrapped: pipe, predict, set_annotations, update, rehearse, get_loss, initialize, begin_update, finish_update, update.

Name	Description
forward_color	Color identifier for model forward passes. Defaults to -1. int
backprop_color	Color identifier for model backpropagation passes. Defaults to -1. int
additional_pipe_functions	Additional pipeline methods to wrap. Keys are pipeline names and values are lists of method identifiers. Defaults to None. Optional[Dict[str, List[str]]]
CREATES	A function that takes the current nlp and wraps pipe models and methods in NVTX ranges. Callable[[Language], Language]

Training data and alignment

training.offsets_to_biluo_tags

Encode labelled spans into per-token tags, using the BILUO scheme (Begin, In, Last, Unit, Out). Returns a list of strings, describing the tags. Each tag string will be in the form of either "", "O" or "{action}-{label}", where action is one of "B", "I", "L", "U". The string "-" is used where the entity offsets don't align with the tokenization in the Doc object. The training algorithm will view these as missing values. O denotes a non-entity token. B denotes the beginning of a multi-token entity, I the inside of an entity of three or more tokens, and L the end of an entity of two or more tokens. U denotes a single-token entity.

This method was previously available as spacy.gold.biluo_tags_from_offsets.

Example

from spacy.training import offsets_to_biluo_tags

doc = nlp("I like London.")
entities = [(7, 13, "LOC")]
tags = offsets_to_biluo_tags(doc, entities)
assert tags == ["O", "O", "U-LOC", "O"]

Name	Description
doc	The document that the entity offsets refer to. The output tags will refer to the token boundaries within the document. Doc
entities	A sequence of (start, end, label) triples. start and end should be character-offset integers denoting the slice into the original string. List[Tuple[int, int, Union[str, int]]]
missing	The label used for missing values, e.g. if tokenization doesn't align with the entity offsets. Defaults to "O". str
RETURNS	A list of strings, describing the BILUO tags. List[str]

training.biluo_tags_to_offsets

Encode per-token tags following the BILUO scheme into entity offsets.

This method was previously available as spacy.gold.offsets_from_biluo_tags.

Example

from spacy.training import biluo_tags_to_offsets

doc = nlp("I like London.")
tags = ["O", "O", "U-LOC", "O"]
entities = biluo_tags_to_offsets(doc, tags)
assert entities == [(7, 13, "LOC")]

Name	Description
doc	The document that the BILUO tags refer to. Doc
tags	A sequence of BILUO tags with each tag describing one token. Each tag string will be of the form of either "", "O" or "{action}-{label}", where action is one of "B", "I", "L", "U". List[str]
RETURNS	A sequence of (start, end, label) triples. start and end will be character-offset integers denoting the slice into the original string. List[Tuple[int, int, str]]

training.biluo_tags_to_spans

Encode per-token tags following the BILUO scheme into Span objects. This can be used to create entity spans from token-based tags, e.g. to overwrite the doc.ents.

This method was previously available as spacy.gold.spans_from_biluo_tags.

Example

from spacy.training import biluo_tags_to_spans

doc = nlp("I like London.")
tags = ["O", "O", "U-LOC", "O"]
doc.ents = biluo_tags_to_spans(doc, tags)

Name	Description
doc	The document that the BILUO tags refer to. Doc
tags	A sequence of BILUO tags with each tag describing one token. Each tag string will be of the form of either "", "O" or "{action}-{label}", where action is one of "B", "I", "L", "U". List[str]
RETURNS	A sequence of Span objects with added entity labels. List[Span]

Utility functions

spaCy comes with a small collection of utility functions located in spacy/util.py. Because utility functions are mostly intended for internal use within spaCy, their behavior may change with future releases. The functions documented on this page should be safe to use and we'll try to ensure backwards compatibility. However, we recommend having additional tests in place if your application depends on any of spaCy's utilities.

util.get_lang_class

Import and load a Language class. Allows lazy-loading language data and importing languages using the two-letter language code. To add a language code for a custom language class, you can register it using the @registry.languages decorator.

Example

for lang_id in ["en", "de"]:
    lang_class = util.get_lang_class(lang_id)
    lang = lang_class()

Name	Description
lang	Two-letter language code, e.g. "en". str
RETURNS	The respective subclass. Language

util.lang_class_is_loaded

Check whether a Language subclass is already loaded. Language subclasses are loaded lazily to avoid expensive setup code associated with the language data.

Example

lang_cls = util.get_lang_class("en")
assert util.lang_class_is_loaded("en") is True
assert util.lang_class_is_loaded("de") is False

Name	Description
name	Two-letter language code, e.g. "en". str
RETURNS	Whether the class has been loaded. bool

util.load_model

Load a pipeline from a package or data path. If called with a string name, spaCy will assume the pipeline is a Python package and import and call its load() method. If called with a path, spaCy will assume it's a data directory, read the language and pipeline settings from the config.cfg and create a Language object. The model data will then be loaded in via Language.from_disk.

Example

nlp = util.load_model("en_core_web_sm")
nlp = util.load_model("en_core_web_sm", exclude=["ner"])
nlp = util.load_model("/path/to/data")

Name	Description
name	Package name or path. str
keyword-only
vocab	Optional shared vocab to pass in on initialization. If True (default), a new Vocab object will be created. Union[Vocab, bool]
disable	Name(s) of pipeline component(s) to disable. Disabled pipes will be loaded but they won't be run unless you explicitly enable them by calling nlp.enable_pipe. Union[str, Iterable[str]]
enable 3.4	Name(s) of pipeline component(s) to enable. All other pipes will be disabled, but can be enabled again using nlp.enable_pipe. Union[str, Iterable[str]]
exclude	Name(s) of pipeline component(s) to exclude. Excluded components won't be loaded. Union[str, Iterable[str]]
config 3	Config overrides as nested dict or flat dict keyed by section values in dot notation, e.g. "nlp.pipeline". Union[Dict[str, Any], Config]
RETURNS	Language class with the loaded pipeline. Language

util.load_model_from_init_py

A helper function to use in the load() method of a pipeline package's __init__.py.

Example

from spacy.util import load_model_from_init_py

def load(**overrides):
    return load_model_from_init_py(__file__, **overrides)

Name	Description
init_file	Path to package's __init__.py, i.e. __file__. Union[str, Path]
keyword-only
vocab 3	Optional shared vocab to pass in on initialization. If True (default), a new Vocab object will be created. Union[Vocab, bool]
disable	Name(s) of pipeline component(s) to disable. Disabled pipes will be loaded but they won't be run unless you explicitly enable them by calling nlp.enable_pipe. Union[str, Iterable[str]]
enable 3.4	Name(s) of pipeline component(s) to enable. All other pipes will be disabled, but can be enabled again using nlp.enable_pipe. Union[str, Iterable[str]]
exclude 3	Name(s) of pipeline component(s) to exclude. Excluded components won't be loaded. Union[str, Iterable[str]]
config 3	Config overrides as nested dict or flat dict keyed by section values in dot notation, e.g. "nlp.pipeline". Union[Dict[str, Any], Config]
RETURNS	Language class with the loaded pipeline. Language

util.load_config

Load a pipeline's config.cfg from a file path. The config typically includes details about the components and how they're created, as well as all training settings and hyperparameters.

Example

config = util.load_config("/path/to/config.cfg")
print(config.to_str())

Name	Description
path	Path to the pipeline's config.cfg. Union[str, Path]
overrides	Optional config overrides to replace in loaded config. Can be provided as nested dict, or as flat dict with keys in dot notation, e.g. "nlp.pipeline". Dict[str, Any]
interpolate	Whether to interpolate the config and replace variables like ${paths.train} with their values. Defaults to False. bool
RETURNS	The pipeline's config. Config

util.load_meta

Get a pipeline's meta.json from a file path and validate its contents. The meta typically includes details about author, licensing, data sources and version.

Example

meta = util.load_meta("/path/to/meta.json")

Name	Description
path	Path to the pipeline's meta.json. Union[str, Path]
RETURNS	The pipeline's meta data. Dict[str, Any]

util.get_installed_models

List all pipeline packages installed in the current environment. This will include any spaCy pipeline that was packaged with spacy package. Under the hood, pipeline packages expose a Python entry point that spaCy can check, without having to load the nlp object.

Example

names = util.get_installed_models()

Name	Description
RETURNS	The string names of the pipelines installed in the current environment. List[str]

util.is_package

Check if string maps to a package installed via pip. Mainly used to validate pipeline packages.

Example

util.is_package("en_core_web_sm") # True
util.is_package("xyz") # False

Name	Description
name	Name of package. str
RETURNS	True if installed package, False if not. bool

util.get_package_path

Get path to an installed package. Mainly used to resolve the location of pipeline packages. Currently imports the package to find its path.

Example

util.get_package_path("en_core_web_sm")
# /usr/lib/python3.6/site-packages/en_core_web_sm

Name	Description
package_name	Name of installed package. str
RETURNS	Path to pipeline package directory. Path

util.is_in_jupyter

Check if user is running spaCy from a Jupyter notebook by detecting the IPython kernel. Mainly used for the displacy visualizer.

Example

html = "<h1>Hello world!</h1>"
if util.is_in_jupyter():
    from IPython.core.display import display, HTML
    display(HTML(html))

Name	Description
RETURNS	True if in Jupyter, False if not. bool

util.compile_prefix_regex

Compile a sequence of prefix rules into a regex object.

Example

prefixes = ("§", "%", "=", r"\+")
prefix_regex = util.compile_prefix_regex(prefixes)
nlp.tokenizer.prefix_search = prefix_regex.search

Name	Description
entries	The prefix rules, e.g. lang.punctuation.TOKENIZER_PREFIXES. Iterable[Union[str, Pattern]]
RETURNS	The regex object to be used for Tokenizer.prefix_search. Pattern

util.compile_suffix_regex

Compile a sequence of suffix rules into a regex object.

Example

suffixes = ("'s", "'S", r"(?<=[0-9])\+")
suffix_regex = util.compile_suffix_regex(suffixes)
nlp.tokenizer.suffix_search = suffix_regex.search

Name	Description
entries	The suffix rules, e.g. lang.punctuation.TOKENIZER_SUFFIXES. Iterable[Union[str, Pattern]]
RETURNS	The regex object to be used for Tokenizer.suffix_search. Pattern

util.compile_infix_regex

Compile a sequence of infix rules into a regex object.

Example

infixes = ("…", "-", "—", r"(?<=[0-9])[+\-\*^](?=[0-9-])")
infix_regex = util.compile_infix_regex(infixes)
nlp.tokenizer.infix_finditer = infix_regex.finditer

Name	Description
entries	The infix rules, e.g. lang.punctuation.TOKENIZER_INFIXES. Iterable[Union[str, Pattern]]
RETURNS	The regex object to be used for Tokenizer.infix_finditer. Pattern

util.minibatch

Iterate over batches of items. size may be an iterator, so that batch-size can vary on each step.

Example

batches = minibatch(train_data)
for batch in batches:
    nlp.update(batch)

Name	Description
items	The items to batch up. Iterable[Any]
size	The batch size(s). Union[int, Sequence[int]]
YIELDS	The batches.

util.filter_spans

Filter a sequence of Span objects and remove duplicates or overlaps. Useful for creating named entities (where one token can only be part of one entity) or when merging spans with Retokenizer.merge. When spans overlap, the (first) longest span is preferred over shorter spans.

Example

doc = nlp("This is a sentence.")
spans = [doc[0:2], doc[0:2], doc[0:4]]
filtered = filter_spans(spans)

Name	Description
spans	The spans to filter. Iterable[Span]
RETURNS	The filtered spans. List[Span]

util.get_words_and_spaces

Given a list of words and a text, reconstruct the original tokens and return a list of words and spaces that can be used to create a Doc. This can help recover destructive tokenization that didn't preserve any whitespace information.

Example

orig_words = ["Hey", ",", "what", "'s", "up", "?"]
orig_text = "Hey, what's up?"
words, spaces = get_words_and_spaces(orig_words, orig_text)
# ['Hey', ',', 'what', "'s", 'up', '?']
# [False, True, False, True, False, False]

Name	Description
words	The list of words. Iterable[str]
text	The original text. str
RETURNS	A list of words and a list of boolean values indicating whether the word at this position is followed by a space. Tuple[List[str], List[bool]]