spaCy/website/docs/usage/training.mdx

---
title: Training Pipelines & Models
teaser: Train and update components on your own data and integrate custom models
next: /usage/layers-architectures
menu:
  - ['Introduction', 'basics']
  - ['Quickstart', 'quickstart']
  - ['Config System', 'config']
  - ['Training Data', 'training-data']
  - ['Custom Training', 'config-custom']
  - ['Custom Functions', 'custom-functions']
  - ['Initialization', 'initialization']
  - ['Data Utilities', 'data']
  - ['Parallel Training', 'parallel-training']
  - ['Internal API', 'api']
---

## Introduction to training {id="basics",hidden="true"}

<Training101 />

<Infobox title="Tip: Try the Prodigy annotation tool">

<Image
  src="/images/prodigy.jpg"
  href="https://prodi.gy"
  alt="Prodigy: Radically efficient machine teaching"
/>

If you need to label a lot of data, check out [Prodigy](https://prodi.gy), a
new, active learning-powered annotation tool we've developed. Prodigy is fast
and extensible, and comes with a modern **web application** that helps you
collect training data faster. It integrates seamlessly with spaCy, pre-selects
the **most relevant examples** for annotation, and lets you train and evaluate
ready-to-use spaCy pipelines.

</Infobox>

## Quickstart {id="quickstart",tag="new"}

The recommended way to train your spaCy pipelines is via the
[`spacy train`](/api/cli#train) command on the command line. It only needs a
single [`config.cfg`](#config) **configuration file** that includes all settings
and hyperparameters. You can optionally [overwrite](#config-overrides) settings
on the command line, and load in a Python file to register
[custom functions](#custom-code) and architectures. This quickstart widget helps
you generate a starter config with the **recommended settings** for your
specific use case. It's also available in spaCy as the
[`init config`](/api/cli#init-config) command.

<Infobox variant="warning">

Upgrade to the [latest version of spaCy](/usage) to use the quickstart widget.
For earlier releases, follow the CLI instructions to generate a compatible
config.

</Infobox>

> #### Instructions: widget
>
> 1. Select your requirements and settings.
> 2. Use the buttons at the bottom to save the result to your clipboard or a
>    file `base_config.cfg`.
> 3. Run [`init fill-config`](/api/cli#init-fill-config) to create a full
>    config.
> 4. Run [`train`](/api/cli#train) with your config and data.
>
> #### Instructions: CLI
>
> 1. Run the [`init config`](/api/cli#init-config) command and specify your
>    requirements and settings as CLI arguments.
> 2. Run [`train`](/api/cli#train) with the exported config and data.

<QuickstartTraining />

After you've saved the starter config to a file `base_config.cfg`, you can use
the [`init fill-config`](/api/cli#init-fill-config) command to fill in the
remaining defaults. Training configs should always be **complete and without
hidden defaults**, to keep your experiments reproducible.

```bash
$ python -m spacy init fill-config base_config.cfg config.cfg
```

> #### Tip: Debug your data
>
> The [`debug data` command](/api/cli#debug-data) lets you analyze and validate
> your training and development data, get useful stats, and find problems like
> invalid entity annotations, cyclic dependencies, low data labels and more.
>
> ```bash
> $ python -m spacy debug data config.cfg
> ```

Instead of exporting your starter config from the quickstart widget and
auto-filling it, you can also use the [`init config`](/api/cli#init-config)
command and specify your requirement and settings as CLI arguments. You can now
add your data and run [`train`](/api/cli#train) with your config. See the
[`convert`](/api/cli#convert) command for details on how to convert your data to
spaCy's binary `.spacy` format. You can either include the data paths in the
`[paths]` section of your config, or pass them in via the command line.

```bash
$ python -m spacy train config.cfg --output ./output --paths.train ./train.spacy --paths.dev ./dev.spacy
```

> #### Tip: Enable your GPU
>
> Use the `--gpu-id` option to select the GPU:
>
> ```bash
> $ python -m spacy train config.cfg --gpu-id 0
> ```

<Accordion title="How are the config recommendations generated?" id="quickstart-source" spaced>

The recommended config settings generated by the quickstart widget and the
[`init config`](/api/cli#init-config) command are based on some general **best
practices** and things we've found to work well in our experiments. The goal is
to provide you with the most **useful defaults**.

Under the hood, the
[`quickstart_training.jinja`](%%GITHUB_SPACY/spacy/cli/templates/quickstart_training.jinja)
template defines the different combinations – for example, which parameters to
change if the pipeline should optimize for efficiency vs. accuracy. The file
[`quickstart_training_recommendations.yml`](%%GITHUB_SPACY/spacy/cli/templates/quickstart_training_recommendations.yml)
collects the recommended settings and available resources for each language
including the different transformer weights. For some languages, we include
different transformer recommendations, depending on whether you want the model
to be more efficient or more accurate. The recommendations will be **evolving**
as we run more experiments.

</Accordion>

<Project id="pipelines/tagger_parser_ud">

The easiest way to get started is to clone a [project template](/usage/projects)
and run it – for example, this end-to-end template that lets you train a
**part-of-speech tagger** and **dependency parser** on a Universal Dependencies
treebank.

</Project>

## Training config system {id="config"}

Training config files include all **settings and hyperparameters** for training
your pipeline. Instead of providing lots of arguments on the command line, you
only need to pass your `config.cfg` file to [`spacy train`](/api/cli#train).
Under the hood, the training config uses the
[configuration system](https://thinc.ai/docs/usage-config) provided by our
machine learning library [Thinc](https://thinc.ai). This also makes it easy to
integrate custom models and architectures, written in your framework of choice.
Some of the main advantages and features of spaCy's training config are:

- **Structured sections.** The config is grouped into sections, and nested
  sections are defined using the `.` notation. For example, `[components.ner]`
  defines the settings for the pipeline's named entity recognizer. The config
  can be loaded as a Python dict.
- **References to registered functions.** Sections can refer to registered
  functions like [model architectures](/api/architectures),
  [optimizers](https://thinc.ai/docs/api-optimizers) or
  [schedules](https://thinc.ai/docs/api-schedules) and define arguments that are
  passed into them. You can also
  [register your own functions](#custom-functions) to define custom
  architectures or methods, reference them in your config and tweak their
  parameters.
- **Interpolation.** If you have hyperparameters or other settings used by
  multiple components, define them once and reference them as
  [variables](#config-interpolation).
- **Reproducibility with no hidden defaults.** The config file is the "single
  source of truth" and includes all settings.
- **Automated checks and validation.** When you load a config, spaCy checks if
  the settings are complete and if all values have the correct types. This lets
  you catch potential mistakes early. In your custom architectures, you can use
  Python [type hints](https://docs.python.org/3/library/typing.html) to tell the
  config which types of data to expect.

```ini
%%GITHUB_SPACY/spacy/default_config.cfg
```

Under the hood, the config is parsed into a dictionary. It's divided into
sections and subsections, indicated by the square brackets and dot notation. For
example, `[training]` is a section and `[training.batch_size]` a subsection.
Subsections can define values, just like a dictionary, or use the `@` syntax to
refer to [registered functions](#config-functions). This allows the config to
not just define static settings, but also construct objects like architectures,
schedules, optimizers or any other custom components. The main top-level
sections of a config file are:

| Section       | Description                                                                                                                                                     |
| ------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `nlp`         | Definition of the `nlp` object, its tokenizer and [processing pipeline](/usage/processing-pipelines) component names.                                           |
| `components`  | Definitions of the [pipeline components](/usage/processing-pipelines) and their models.                                                                         |
| `paths`       | Paths to data and other assets. Re-used across the config as variables, e.g. `${paths.train}`, and can be [overwritten](#config-overrides) on the CLI.          |
| `system`      | Settings related to system and hardware. Re-used across the config as variables, e.g. `${system.seed}`, and can be [overwritten](#config-overrides) on the CLI. |
| `training`    | Settings and controls for the training and evaluation process.                                                                                                  |
| `pretraining` | Optional settings and controls for the [language model pretraining](/usage/embeddings-transformers#pretraining).                                                |
| `initialize`  | Data resources and arguments passed to components when [`nlp.initialize`](/api/language#initialize) is called before training (but not at runtime).             |

<Infobox title="Config format and settings" emoji="📖">

For a full overview of spaCy's config format and settings, see the
[data format documentation](/api/data-formats#config) and
[Thinc's config system docs](https://thinc.ai/docs/usage-config). The settings
available for the different architectures are documented with the
[model architectures API](/api/architectures). See the Thinc documentation for
[optimizers](https://thinc.ai/docs/api-optimizers) and
[schedules](https://thinc.ai/docs/api-schedules).

</Infobox>

<YouTube id="BWhh3r6W-qE"></YouTube>

### Config lifecycle at runtime and training {id="config-lifecycle"}

A pipeline's `config.cfg` is considered the "single source of truth", both at
**training** and **runtime**. Under the hood,
[`Language.from_config`](/api/language#from_config) takes care of constructing
the `nlp` object using the settings defined in the config. An `nlp` object's
config is available as [`nlp.config`](/api/language#config) and it includes all
information about the pipeline, as well as the settings used to train and
initialize it.

![Illustration of pipeline lifecycle](/images/lifecycle.svg)

At runtime spaCy will only use the `[nlp]` and `[components]` blocks of the
config and load all data, including tokenization rules, model weights and other
resources from the pipeline directory. The `[training]` block contains the
settings for training the model and is only used during training. Similarly, the
`[initialize]` block defines how the initial `nlp` object should be set up
before training and whether it should be initialized with vectors or pretrained
tok2vec weights, or any other data needed by the components.

The initialization settings are only loaded and used when
[`nlp.initialize`](/api/language#initialize) is called (typically right before
training). This allows you to set up your pipeline using local data resources
and custom functions, and preserve the information in your config – but without
requiring it to be available at runtime. You can also use this mechanism to
provide data paths to custom pipeline components and custom tokenizers – see the
section on [custom initialization](#initialization) for details.

### Overwriting config settings on the command line {id="config-overrides"}

The config system means that you can define all settings **in one place** and in
a consistent format. There are no command-line arguments that need to be set,
and no hidden defaults. However, there can still be scenarios where you may want
to override config settings when you run [`spacy train`](/api/cli#train). This
includes **file paths** to vectors or other resources that shouldn't be
hard-coded in a config file, or **system-dependent settings**.

For cases like this, you can set additional command-line options starting with
`--` that correspond to the config section and value to override. For example,
`--paths.train ./corpus/train.spacy` sets the `train` value in the `[paths]`
block.

```bash
$ python -m spacy train config.cfg --paths.train ./corpus/train.spacy --paths.dev ./corpus/dev.spacy --training.batch_size 128
```

Only existing sections and values in the config can be overwritten. At the end
of the training, the final filled `config.cfg` is exported with your pipeline,
so you'll always have a record of the settings that were used, including your
overrides. Overrides are added before [variables](#config-interpolation) are
resolved, by the way – so if you need to use a value in multiple places,
reference it across your config and override it on the CLI once.

> #### 💡 Tip: Verbose logging
>
> If you're using config overrides, you can set the `--verbose` flag on
> [`spacy train`](/api/cli#train) to make spaCy log more info, including which
> overrides were set via the CLI and environment variables.

#### Adding overrides via environment variables {id="config-overrides-env"}

Instead of defining the overrides as CLI arguments, you can also use the
`SPACY_CONFIG_OVERRIDES` environment variable using the same argument syntax.
This is especially useful if you're training models as part of an automated
process. Environment variables **take precedence** over CLI overrides and values
defined in the config file.

```bash
$ SPACY_CONFIG_OVERRIDES="--system.gpu_allocator pytorch --training.batch_size 128" ./your_script.sh
```

### Reading from standard input {id="config-stdin"}

Setting the config path to `-` on the command line lets you read the config from
standard input and pipe it forward from a different process, like
[`init config`](/api/cli#init-config) or your own custom script. This is
especially useful for quick experiments, as it lets you generate a config on the
fly without having to save to and load from disk.

> #### 💡 Tip: Writing to stdout
>
> When you run `init config`, you can set the output path to `-` to write to
> stdout. In a custom script, you can print the string config, e.g.
> `print(nlp.config.to_str())`.

```bash
$ python -m spacy init config - --lang en --pipeline ner,textcat --optimize accuracy | python -m spacy train - --paths.train ./corpus/train.spacy --paths.dev ./corpus/dev.spacy
```

### Using variable interpolation {id="config-interpolation"}

Another very useful feature of the config system is that it supports variable
interpolation for both **values and sections**. This means that you only need to
define a setting once and can reference it across your config using the
`${section.value}` syntax. In this example, the value of `seed` is reused within
the `[training]` block, and the whole block of `[training.optimizer]` is reused
in `[pretraining]` and will become `pretraining.optimizer`.

```ini {title="config.cfg (excerpt)",highlight="5,18"}
[system]
seed = 0

[training]
seed = ${system.seed}

[training.optimizer]
@optimizers = "Adam.v1"
beta1 = 0.9
beta2 = 0.999
L2_is_weight_decay = true
L2 = 0.01
grad_clip = 1.0
use_averages = false
eps = 1e-8

[pretraining]
optimizer = ${training.optimizer}
```

You can also use variables inside strings. In that case, it works just like
f-strings in Python. If the value of a variable is not a string, it's converted
to a string.

```ini
[paths]
version = 5
root = "/Users/you/data"
train = "${paths.root}/train_${paths.version}.spacy"
# Result: /Users/you/data/train_5.spacy
```

<Infobox title="Tip: Override variables on the CLI" emoji="💡">

If you need to change certain values between training runs, you can define them
once, reference them as variables and then [override](#config-overrides) them on
the CLI. For example, `--paths.root /other/root` will change the value of `root`
in the block `[paths]` and the change will be reflected across all other values
that reference this variable.

</Infobox>

## Preparing Training Data {id="training-data"}

Training data for NLP projects comes in many different formats. For some common
formats such as CoNLL, spaCy provides [converters](/api/cli#convert) you can use
from the command line. In other cases you'll have to prepare the training data
yourself.

When converting training data for use in spaCy, the main thing is to create
[`Doc`](/api/doc) objects just like the results you want as output from the
pipeline. For example, if you're creating an NER pipeline, loading your
annotations and setting them as the `.ents` property on a `Doc` is all you need
to worry about. On disk the annotations will be saved as a
[`DocBin`](/api/docbin) in the
[`.spacy` format](/api/data-formats#binary-training), but the details of that
are handled automatically.

Here's an example of creating a `.spacy` file from some NER annotations.

```python {title="preprocess.py"}
import spacy
from spacy.tokens import DocBin

nlp = spacy.blank("en")
training_data = [
  ("Tokyo Tower is 333m tall.", [(0, 11, "BUILDING")]),
]
# the DocBin will store the example documents
db = DocBin()
for text, annotations in training_data:
    doc = nlp(text)
    ents = []
    for start, end, label in annotations:
        span = doc.char_span(start, end, label=label)
        ents.append(span)
    doc.ents = ents
    db.add(doc)
db.to_disk("./train.spacy")
```

For more examples of how to convert training data from a wide variety of formats
for use with spaCy, look at the preprocessing steps in the
[tutorial projects](https://github.com/explosion/projects/tree/v3/tutorials).

<Accordion title="What about the spaCy JSON format?" id="json-annotations" spaced>

In spaCy v2, the recommended way to store training data was in
[a particular JSON format](/api/data-formats#json-input), but in v3 this format
is deprecated. It's fine as a readable storage format, but there's no need to
convert your data to JSON before creating a `.spacy` file.

</Accordion>

## Customizing the pipeline and training {id="config-custom"}

### Defining pipeline components {id="config-components"}

You typically train a [pipeline](/usage/processing-pipelines) of **one or more
components**. The `[components]` block in the config defines the available
pipeline components and how they should be created – either by a built-in or
custom [factory](/usage/processing-pipelines#built-in), or
[sourced](/usage/processing-pipelines#sourced-components) from an existing
trained pipeline. For example, `[components.parser]` defines the component named
`"parser"` in the pipeline. There are different ways you might want to treat
your components during training, and the most common scenarios are:

1. Train a **new component** from scratch on your data.
2. Update an existing **trained component** with more examples.
3. Include an existing trained component without updating it.
4. Include a non-trainable component, like a rule-based
   [`EntityRuler`](/api/entityruler) or [`Sentencizer`](/api/sentencizer), or a
   fully [custom component](/usage/processing-pipelines#custom-components).

If a component block defines a `factory`, spaCy will look it up in the
[built-in](/usage/processing-pipelines#built-in) or
[custom](/usage/processing-pipelines#custom-components) components and create a
new component from scratch. All settings defined in the config block will be
passed to the component factory as arguments. This lets you configure the model
settings and hyperparameters. If a component block defines a `source`, the
component will be copied over from an existing trained pipeline, with its
existing weights. This lets you include an already trained component in your
pipeline, or update a trained component with more data specific to your use
case.

```ini {title="config.cfg (excerpt)"}
[components]

# "parser" and "ner" are sourced from a trained pipeline
[components.parser]
source = "en_core_web_sm"

[components.ner]
source = "en_core_web_sm"

# "textcat" and "custom" are created blank from a built-in / custom factory
[components.textcat]
factory = "textcat"

[components.custom]
factory = "your_custom_factory"
your_custom_setting = true
```

The `pipeline` setting in the `[nlp]` block defines the pipeline components
added to the pipeline, in order. For example, `"parser"` here references
`[components.parser]`. By default, spaCy will **update all components that can
be updated**. Trainable components that are created from scratch are initialized
with random weights. For sourced components, spaCy will keep the existing
weights and [resume training](/api/language#resume_training).

If you don't want a component to be updated, you can **freeze** it by adding it
to the `frozen_components` list in the `[training]` block. Frozen components are
**not updated** during training and are included in the final trained pipeline
as-is. They are also excluded when calling
[`nlp.initialize`](/api/language#initialize).

> #### Note on frozen components
>
> Even though frozen components are not **updated** during training, they will
> still **run** during evaluation. This is very important, because they may
> still impact your model's performance – for instance, a sentence boundary
> detector can impact what the parser or entity recognizer considers a valid
> parse. So the evaluation results should always reflect what your pipeline will
> produce at runtime. If you want a frozen component to run (without updating)
> during training as well, so that downstream components can use its
> **predictions**, you should add it to the list of
> [`annotating_components`](/usage/training#annotating-components).

```ini
[nlp]
lang = "en"
pipeline = ["parser", "ner", "textcat", "custom"]

[training]
frozen_components = ["parser", "custom"]
```

<Infobox variant="warning" title="Shared Tok2Vec listener layer" id="config-components-listeners">

When the components in your pipeline
[share an embedding layer](/usage/embeddings-transformers#embedding-layers), the
**performance** of your frozen component will be **degraded** if you continue
training other layers with the same underlying `Tok2Vec` instance. As a rule of
thumb, ensure that your frozen components are truly **independent** in the
pipeline.

To automatically replace a shared token-to-vector listener with an independent
copy of the token-to-vector layer, you can use the `replace_listeners` setting
of a sourced component, pointing to the listener layer(s) in the config. For
more details on how this works under the hood, see
[`Language.replace_listeners`](/api/language#replace_listeners).

```ini
[training]
frozen_components = ["tagger"]

[components.tagger]
source = "en_core_web_sm"
replace_listeners = ["model.tok2vec"]
```

</Infobox>

### Using predictions from preceding components {id="annotating-components",version="3.1"}

By default, components are updated in isolation during training, which means
that they don't see the predictions of any earlier components in the pipeline. A
component receives [`Example.predicted`](/api/example) as input and compares its
predictions to [`Example.reference`](/api/example) without saving its
annotations in the `predicted` doc.

Instead, if certain components should **set their annotations** during training,
use the setting `annotating_components` in the `[training]` block to specify a
list of components. For example, the feature `DEP` from the parser could be used
as a tagger feature by including `DEP` in the tok2vec `attrs` and including
`parser` in `annotating_components`:

```ini {title="config.cfg (excerpt)",highlight="7,12"}
[nlp]
pipeline = ["parser", "tagger"]

[components.tagger.model.tok2vec.embed]
@architectures = "spacy.MultiHashEmbed.v1"
width = ${components.tagger.model.tok2vec.encode.width}
attrs = ["NORM","DEP"]
rows = [5000,2500]
include_static_vectors = false

[training]
annotating_components = ["parser"]
```

Any component in the pipeline can be included as an annotating component,
including frozen components. Frozen components can set annotations during
training just as they would set annotations during evaluation or when the final
pipeline is run. The config excerpt below shows how a frozen `ner` component and
a `sentencizer` can provide the required `doc.sents` and `doc.ents` for the
entity linker during training:

```ini {title="config.cfg (excerpt)"}
[nlp]
pipeline = ["sentencizer", "ner", "entity_linker"]

[components.ner]
source = "en_core_web_sm"

[training]
frozen_components = ["ner"]
annotating_components = ["sentencizer", "ner"]
```

Similarly, a pretrained `tok2vec` layer can be frozen and specified in the list
of `annotating_components` to ensure that a downstream component can use the
embedding layer without updating it.

<Infobox variant="warning" title="Training speed with annotating components" id="annotating-components-speed">

Be aware that non-frozen annotating components with statistical models will
**run twice** on each batch, once to update the model and once to apply the
now-updated model to the predicted docs.

</Infobox>

### Using registered functions {id="config-functions"}

The training configuration defined in the config file doesn't have to only
consist of static values. Some settings can also be **functions**. For instance,
the `batch_size` can be a number that doesn't change, or a schedule, like a
sequence of compounding values, which has shown to be an effective trick (see
[Smith et al., 2017](https://arxiv.org/abs/1711.00489)).

```ini {title="With static value"}
[training]
batch_size = 128
```

To refer to a function instead, you can make `[training.batch_size]` its own
section and use the `@` syntax to specify the function and its arguments – in
this case [`compounding.v1`](https://thinc.ai/docs/api-schedules#compounding)
defined in the [function registry](/api/top-level#registry). All other values
defined in the block are passed to the function as keyword arguments when it's
initialized. You can also use this mechanism to register
[custom implementations and architectures](#custom-functions) and reference them
from your configs.

> #### How the config is resolved
>
> The config file is parsed into a regular dictionary and is resolved and
> validated **bottom-up**. Arguments provided for registered functions are
> checked against the function's signature and type annotations. The return
> value of a registered function can also be passed into another function – for
> instance, a learning rate schedule can be provided as the an argument of an
> optimizer.

```ini {title="With registered function"}
[training.batch_size]
@schedules = "compounding.v1"
start = 100
stop = 1000
compound = 1.001
```

### Model architectures {id="model-architectures"}

> #### 💡 Model type annotations
>
> In the documentation and code base, you may come across type annotations and
> descriptions of [Thinc](https://thinc.ai) model types, like ~~Model[List[Doc],
> List[Floats2d]]~~. This so-called generic type describes the layer and its
> input and output type – in this case, it takes a list of `Doc` objects as the
> input and list of 2-dimensional arrays of floats as the output. You can read
> more about defining Thinc models [here](https://thinc.ai/docs/usage-models).
> Also see the [type checking](https://thinc.ai/docs/usage-type-checking) for
> how to enable linting in your editor to see live feedback if your inputs and
> outputs don't match.

A **model architecture** is a function that wires up a Thinc
[`Model`](https://thinc.ai/docs/api-model) instance, which you can then use in a
component or as a layer of a larger network. You can use Thinc as a thin
[wrapper around frameworks](https://thinc.ai/docs/usage-frameworks) such as
PyTorch, TensorFlow or MXNet, or you can implement your logic in Thinc
[directly](https://thinc.ai/docs/usage-models). For more details and examples,
see the usage guide on [layers and architectures](/usage/layers-architectures).

spaCy's built-in components will never construct their `Model` instances
themselves, so you won't have to subclass the component to change its model
architecture. You can just **update the config** so that it refers to a
different registered function. Once the component has been created, its `Model`
instance has already been assigned, so you cannot change its model architecture.
The architecture is like a recipe for the network, and you can't change the
recipe once the dish has already been prepared. You have to make a new one.
spaCy includes a variety of built-in [architectures](/api/architectures) for
different tasks. For example:

| Architecture                                                      | Description                                                                                                                                                                                                                                               |
| ----------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| [HashEmbedCNN](/api/architectures#HashEmbedCNN)                   | Build spaCy’s "standard" embedding layer, which uses hash embedding with subword features and a CNN with layer-normalized maxout. ~~Model[List[Doc], List[Floats2d]]~~                                                                                    |
| [TransitionBasedParser](/api/architectures#TransitionBasedParser) | Build a [transition-based parser](https://explosion.ai/blog/parsing-english-in-python) model used in the default [`EntityRecognizer`](/api/entityrecognizer) and [`DependencyParser`](/api/dependencyparser). ~~Model[List[Docs], List[List[Floats2d]]]~~ |
| [TextCatEnsemble](/api/architectures#TextCatEnsemble)             | Stacked ensemble of a bag-of-words model and a neural network model with an internal CNN embedding layer. Used in the default [`TextCategorizer`](/api/textcategorizer). ~~Model[List[Doc], Floats2d]~~                                                   |

### Metrics, training output and weighted scores {id="metrics"}

When you train a pipeline using the [`spacy train`](/api/cli#train) command,
you'll see a table showing the metrics after each pass over the data. The
available metrics **depend on the pipeline components**. Pipeline components
also define which scores are shown and how they should be **weighted in the
final score** that decides about the best model.

The `training.score_weights` setting in your `config.cfg` lets you customize the
scores shown in the table and how they should be weighted. In this example, the
labeled dependency accuracy and NER F-score count towards the final score with
40% each and the tagging accuracy makes up the remaining 20%. The tokenization
accuracy and speed are both shown in the table, but not counted towards the
score.

> #### Why do I need score weights?
>
> At the end of your training process, you typically want to select the **best
> model** – but what "best" means depends on the available components and your
> specific use case. For instance, you may prefer a pipeline with higher NER and
> lower POS tagging accuracy over a pipeline with lower NER and higher POS
> accuracy. You can express this preference in the score weights, e.g. by
> assigning `ents_f` (NER F-score) a higher weight.

```ini
[training.score_weights]
dep_las = 0.4
dep_uas = null
ents_f = 0.4
tag_acc = 0.2
token_acc = 0.0
speed = 0.0
```

The `score_weights` don't _have to_ sum to `1.0` – but it's recommended. When
you generate a config for a given pipeline, the score weights are generated by
combining and normalizing the default score weights of the pipeline components.
The default score weights are defined by each pipeline component via the
`default_score_weights` setting on the
[`@Language.factory`](/api/language#factory) decorator. By default, all pipeline
components are weighted equally. If a score weight is set to `null`, it will be
excluded from the logs and the score won't be weighted.

<Accordion title="Understanding the training output and score types" spaced id="score-types">

| Name              | Description                                                                                                             |
| ----------------- | ----------------------------------------------------------------------------------------------------------------------- |
| **Loss**          | The training loss representing the amount of work left for the optimizer. Should decrease, but usually not to `0`.      |
| **Precision** (P) | Percentage of predicted annotations that were correct. Should increase.                                                 |
| **Recall** (R)    | Percentage of reference annotations recovered. Should increase.                                                         |
| **F-Score** (F)   | Harmonic mean of precision and recall. Should increase.                                                                 |
| **UAS** / **LAS** | Unlabeled and labeled attachment score for the dependency parser, i.e. the percentage of correct arcs. Should increase. |
| **Speed**         | Prediction speed in words per second (WPS). Should stay stable.                                                         |

Note that if the development data has raw text, some of the gold-standard
entities might not align to the predicted tokenization. These tokenization
errors are **excluded from the NER evaluation**. If your tokenization makes it
impossible for the model to predict 50% of your entities, your NER F-score might
still look good.

</Accordion>

## Custom functions {id="custom-functions"}

Registered functions in the training config files can refer to built-in
implementations, but you can also plug in fully **custom implementations**. All
you need to do is register your function using the `@spacy.registry` decorator
with the name of the respective [registry](/api/top-level#registry), e.g.
`@spacy.registry.architectures`, and a string name to assign to your function.
Registering custom functions allows you to **plug in models** defined in PyTorch
or TensorFlow, make **custom modifications** to the `nlp` object, create custom
optimizers or schedules, or **stream in data** and preprocess it on the fly
while training.

Each custom function can have any number of arguments that are passed in via the
[config](#config), just the built-in functions. If your function defines
**default argument values**, spaCy is able to auto-fill your config when you run
[`init fill-config`](/api/cli#init-fill-config). If you want to make sure that a
given parameter is always explicitly set in the config, avoid setting a default
value for it.

### Training with custom code {id="custom-code"}

> ```bash
> ### Training
> $ python -m spacy train config.cfg --code functions.py
> ```
>
> ```bash
> ### Packaging
> $ python -m spacy package ./model-best ./packages --code functions.py
> ```

The [`spacy train`](/api/cli#train) recipe lets you specify an optional argument
`--code` that points to a Python file. The file is imported before training and
allows you to add custom functions and architectures to the function registry
that can then be referenced from your `config.cfg`. This lets you train spaCy
pipelines with custom components, without having to re-implement the whole
training workflow. When you package your trained pipeline later using
[`spacy package`](/api/cli#package), you can provide one or more Python files to
be included in the package and imported in its `__init__.py`. This means that
any custom architectures, functions or
[components](/usage/processing-pipelines#custom-components) will be shipped with
your pipeline and registered when it's loaded. See the documentation on
[saving and loading pipelines](/usage/saving-loading#models-custom) for details.

<Infobox variant="warning">

Note that the unpackaged models produced by `spacy train` are data directories
that **do not include custom code**. You need to import the code in your script
before loading in unpackaged models. For more details, see
[`spacy.load`](/api/top-level#spacy.load).

</Infobox>

#### Example: Modifying the nlp object {id="custom-code-nlp-callbacks"}

For many use cases, you don't necessarily want to implement the whole `Language`
subclass and language data from scratch – it's often enough to make a few small
modifications, like adjusting the
[tokenization rules](/usage/linguistic-features#native-tokenizer-additions) or
[language defaults](/api/language#defaults) like stop words. The config lets you
provide five optional **callback functions** that give you access to the
language class and `nlp` object at different points of the lifecycle:

| Callback                      | Description                                                                                                                                                                                                                |
| ----------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `nlp.before_creation`         | Called before the `nlp` object is created and receives the language subclass like `English` (not the instance). Useful for writing to the [`Language.Defaults`](/api/language#defaults) aside from the tokenizer settings. |
| `nlp.after_creation`          | Called right after the `nlp` object is created, but before the pipeline components are added to the pipeline and receives the `nlp` object.                                                                                |
| `nlp.after_pipeline_creation` | Called right after the pipeline components are created and added and receives the `nlp` object. Useful for modifying pipeline components.                                                                                  |
| `initialize.before_init`      | Called before the pipeline components are initialized and receives the `nlp` object for in-place modification. Useful for modifying the tokenizer settings, similar to the v2 base model option.                           |
| `initialize.after_init`       | Called after the pipeline components are initialized and receives the `nlp` object for in-place modification.                                                                                                              |

The `@spacy.registry.callbacks` decorator lets you register your custom function
in the `callbacks` [registry](/api/top-level#registry) under a given name. You
can then reference the function in a config block using the `@callbacks` key. If
a block contains a key starting with an `@`, it's interpreted as a reference to
a function. Because you've registered the function, spaCy knows how to create it
when you reference `"customize_language_data"` in your config. Here's an example
of a callback that runs before the `nlp` object is created and adds a custom
stop word to the defaults:

> #### config.cfg
>
> ```ini
> [nlp.before_creation]
> @callbacks = "customize_language_data"
> ```

```python {title="functions.py",highlight="3,6"}
import spacy

@spacy.registry.callbacks("customize_language_data")
def create_callback():
    def customize_language_data(lang_cls):
        lang_cls.Defaults.stop_words.add("good")
        return lang_cls

    return customize_language_data
```

<Infobox variant="warning">

Remember that a registered function should always be a function that spaCy
**calls to create something**. In this case, it **creates a callback** – it's
not the callback itself.

</Infobox>

Any registered function – in this case `create_callback` – can also take
**arguments** that can be **set by the config**. This lets you implement and
keep track of different configurations, without having to hack at your code. You
can choose any arguments that make sense for your use case. In this example,
we're adding the arguments `extra_stop_words` (a list of strings) and `debug`
(boolean) for printing additional info when the function runs.

> #### config.cfg
>
> ```ini
> [nlp.before_creation]
> @callbacks = "customize_language_data"
> extra_stop_words = ["ooh", "aah"]
> debug = true
> ```

```python {title="functions.py",highlight="5,7-9"}
from typing import List
import spacy

@spacy.registry.callbacks("customize_language_data")
def create_callback(extra_stop_words: List[str] = [], debug: bool = False):
    def customize_language_data(lang_cls):
        lang_cls.Defaults.stop_words.update(extra_stop_words)
        if debug:
            print("Updated stop words")
        return lang_cls

    return customize_language_data
```

<Infobox title="Tip: Use Python type hints" emoji="💡">

spaCy's configs are powered by our machine learning library Thinc's
[configuration system](https://thinc.ai/docs/usage-config), which supports
[type hints](https://docs.python.org/3/library/typing.html) and even
[advanced type annotations](https://thinc.ai/docs/usage-config#advanced-types)
using [`pydantic`](https://github.com/samuelcolvin/pydantic). If your registered
function provides type hints, the values that are passed in will be checked
against the expected types. For example, `debug: bool` in the example above will
ensure that the value received as the argument `debug` is a boolean. If the
value can't be coerced into a boolean, spaCy will raise an error.
`debug: pydantic.StrictBool` will force the value to be a boolean and raise an
error if it's not – for instance, if your config defines `1` instead of `true`.

</Infobox>

With your `functions.py` defining additional code and the updated `config.cfg`,
you can now run [`spacy train`](/api/cli#train) and point the argument `--code`
to your Python file. Before loading the config, spaCy will import the
`functions.py` module and your custom functions will be registered.

```bash
$ python -m spacy train config.cfg --output ./output --code ./functions.py
```

#### Example: Modifying tokenizer settings {id="custom-tokenizer"}

Use the `initialize.before_init` callback to modify the tokenizer settings when
training a new pipeline. Write a registered callback that modifies the tokenizer
settings and specify this callback in your config:

> #### config.cfg
>
> ```ini
> [initialize]
>
> [initialize.before_init]
> @callbacks = "customize_tokenizer"
> ```

```python {title="functions.py"}
from spacy.util import registry, compile_suffix_regex

@registry.callbacks("customize_tokenizer")
def make_customize_tokenizer():
    def customize_tokenizer(nlp):
        # remove a suffix
        suffixes = list(nlp.Defaults.suffixes)
        suffixes.remove("\\[")
        suffix_regex = compile_suffix_regex(suffixes)
        nlp.tokenizer.suffix_search = suffix_regex.search

        # add a special case
        nlp.tokenizer.add_special_case("_SPECIAL_", [{"ORTH": "_SPECIAL_"}])
    return customize_tokenizer
```

When training, provide the function above with the `--code` option:

```bash
$ python -m spacy train config.cfg --code ./functions.py
```

Because this callback is only called in the one-time initialization step before
training, the callback code does not need to be packaged with the final pipeline
package. However, to make it easier for others to replicate your training setup,
you can choose to package the initialization callbacks with the pipeline package
or to publish them separately.

<Infobox variant="warning" title="nlp.before_creation vs. initialize.before_init">

- `nlp.before_creation` is the best place to modify language defaults other than
  the tokenizer settings.
- `initialize.before_init` is the best place to modify tokenizer settings when
  training a new pipeline.

Unlike the other language defaults, the tokenizer settings are saved with the
pipeline with `nlp.to_disk()`, so modifications made in `nlp.before_creation`
will be clobbered by the saved settings when the trained pipeline is loaded from
disk.

</Infobox>

#### Example: Custom logging function {id="custom-logging"}

During training, the results of each step are passed to a logger function. By
default, these results are written to the console with the
[`ConsoleLogger`](/api/top-level#ConsoleLogger). There is also built-in support
for writing the log files to [Weights & Biases](https://www.wandb.com/) with the
[`WandbLogger`](https://github.com/explosion/spacy-loggers#wandblogger). On each
step, the logger function receives a **dictionary** with the following keys:

| Key            | Value                                                                                                 |
| -------------- | ----------------------------------------------------------------------------------------------------- |
| `epoch`        | How many passes over the data have been completed. ~~int~~                                            |
| `step`         | How many steps have been completed. ~~int~~                                                           |
| `score`        | The main score from the last evaluation, measured on the dev set. ~~float~~                           |
| `other_scores` | The other scores from the last evaluation, measured on the dev set. ~~Dict[str, Any]~~                |
| `losses`       | The accumulated training losses, keyed by component name. ~~Dict[str, float]~~                        |
| `checkpoints`  | A list of previous results, where each result is a `(score, step)` tuple. ~~List[Tuple[float, int]]~~ |

You can easily implement and plug in your own logger that records the training
results in a custom way, or sends them to an experiment management tracker of
your choice. In this example, the function `my_custom_logger.v1` writes the
tabular results to a file:

> ```ini
> ### config.cfg (excerpt)
> [training.logger]
> @loggers = "my_custom_logger.v1"
> log_path = "my_file.tab"
> ```

```python {title="functions.py"}
import sys
from typing import IO, Tuple, Callable, Dict, Any, Optional
import spacy
from spacy import Language
from pathlib import Path

@spacy.registry.loggers("my_custom_logger.v1")
def custom_logger(log_path):
    def setup_logger(
        nlp: Language,
        stdout: IO=sys.stdout,
        stderr: IO=sys.stderr
    ) -> Tuple[Callable, Callable]:
        stdout.write(f"Logging to {log_path}\\n")
        log_file = Path(log_path).open("w", encoding="utf8")
        log_file.write("step\\t")
        log_file.write("score\\t")
        for pipe in nlp.pipe_names:
            log_file.write(f"loss_{pipe}\\t")
        log_file.write("\\n")

        def log_step(info: Optional[Dict[str, Any]]):
            if info:
                log_file.write(f"{info['step']}\\t")
                log_file.write(f"{info['score']}\\t")
                for pipe in nlp.pipe_names:
                    log_file.write(f"{info['losses'][pipe]}\\t")
                log_file.write("\\n")

        def finalize():
            log_file.close()

        return log_step, finalize

    return setup_logger
```

#### Example: Custom batch size schedule {id="custom-code-schedule"}

You can also implement your own batch size schedule to use during training. The
`@spacy.registry.schedules` decorator lets you register that function in the
`schedules` [registry](/api/top-level#registry) and assign it a string name:

> #### Why the version in the name?
>
> A big benefit of the config system is that it makes your experiments
> reproducible. We recommend versioning the functions you register, especially
> if you expect them to change (like a new model architecture). This way, you
> know that a config referencing `v1` means a different function than a config
> referencing `v2`.

```python {title="functions.py"}
import spacy

@spacy.registry.schedules("my_custom_schedule.v1")
def my_custom_schedule(start: int = 1, factor: float = 1.001):
   while True:
      yield start
      start = start * factor
```

In your config, you can now reference the schedule in the
`[training.batch_size]` block via `@schedules`. If a block contains a key
starting with an `@`, it's interpreted as a reference to a function. All other
settings in the block will be passed to the function as keyword arguments. Keep
in mind that the config shouldn't have any hidden defaults and all arguments on
the functions need to be represented in the config.

```ini {title="config.cfg (excerpt)"}
[training.batch_size]
@schedules = "my_custom_schedule.v1"
start = 2
factor = 1.005
```

### Defining custom architectures {id="custom-architectures"}

Built-in pipeline components such as the tagger or named entity recognizer are
constructed with default neural network [models](/api/architectures). You can
change the model architecture entirely by implementing your own custom models
and providing those in the config when creating the pipeline component. See the
documentation on [layers and model architectures](/usage/layers-architectures)
for more details.

> ```ini
> ### config.cfg
> [components.tagger]
> factory = "tagger"
>
> [components.tagger.model]
> @architectures = "custom_neural_network.v1"
> output_width = 512
> ```

```python {title="functions.py"}
from typing import List
from thinc.types import Floats2d
from thinc.api import Model
import spacy
from spacy.tokens import Doc

@spacy.registry.architectures("custom_neural_network.v1")
def custom_neural_network(output_width: int) -> Model[List[Doc], List[Floats2d]]:
    return create_model(output_width)
```

## Customizing the initialization {id="initialization"}

When you start training a new model from scratch,
[`spacy train`](/api/cli#train) will call
[`nlp.initialize`](/api/language#initialize) to initialize the pipeline and load
the required data. All settings for this are defined in the
[`[initialize]`](/api/data-formats#config-initialize) block of the config, so
you can keep track of how the initial `nlp` object was created. The
initialization process typically includes the following:

> #### config.cfg (excerpt)
>
> ```ini
> [initialize]
> vectors = ${paths.vectors}
> init_tok2vec = ${paths.init_tok2vec}
>
> [initialize.components]
> # Settings for components
> ```

1. Load in **data resources** defined in the `[initialize]` config, including
   **word vectors** and
   [pretrained](/usage/embeddings-transformers/#pretraining) **tok2vec
   weights**.
2. Call the `initialize` methods of the tokenizer (if implemented, e.g. for
   [Chinese](/usage/models#chinese)) and pipeline components with a callback to
   access the training data, the current `nlp` object and any **custom
   arguments** defined in the `[initialize]` config.
3. In **pipeline components**: if needed, use the data to
   [infer missing shapes](/usage/layers-architectures#thinc-shape-inference) and
   set up the label scheme if no labels are provided. Components may also load
   other data like lookup tables or dictionaries.

The initialization step allows the config to define **all settings** required
for the pipeline, while keeping a separation between settings and functions that
should only be used **before training** to set up the initial pipeline, and
logic and configuration that needs to be available **at runtime**. Without that
separation, it would be very difficult to use the same, reproducible config file
because the component settings required for training (load data from an external
file) wouldn't match the component settings required at runtime (load what's
included with the saved `nlp` object and don't depend on external file).

![Illustration of pipeline lifecycle](/images/lifecycle.svg)

<Infobox title="How components save and load data" emoji="📖">

For details and examples of how pipeline components can **save and load data
assets** like model weights or lookup tables, and how the component
initialization is implemented under the hood, see the usage guide on
[serializing and initializing component data](/usage/processing-pipelines#component-data-initialization).

</Infobox>

#### Initializing labels {id="initialization-labels"}

Built-in pipeline components like the
[`EntityRecognizer`](/api/entityrecognizer) or
[`DependencyParser`](/api/dependencyparser) need to know their available labels
and associated internal meta information to initialize their model weights.
Using the `get_examples` callback provided on initialization, they're able to
**read the labels off the training data** automatically, which is very
convenient – but it can also slow down the training process to compute this
information on every run.

The [`init labels`](/api/cli#init-labels) command lets you auto-generate JSON
files containing the label data for all supported components. You can then pass
in the labels in the `[initialize]` settings for the respective components to
allow them to initialize faster.

> #### config.cfg
>
> ```ini
> [initialize.components.ner]
>
> [initialize.components.ner.labels]
> @readers = "spacy.read_labels.v1"
> path = "corpus/labels/ner.json
> ```

```bash
$ python -m spacy init labels config.cfg ./corpus --paths.train ./corpus/train.spacy
```

Under the hood, the command delegates to the `label_data` property of the
pipeline components, for instance
[`EntityRecognizer.label_data`](/api/entityrecognizer#label_data).

<Infobox variant="warning" title="Important note">

The JSON format differs for each component and some components need additional
meta information about their labels. The format exported by
[`init labels`](/api/cli#init-labels) matches what the components need, so you
should always let spaCy **auto-generate the labels** for you.

</Infobox>

## Data utilities {id="data"}

spaCy includes various features and utilities to make it easy to train models
using your own data, manage training and evaluation corpora, convert existing
annotations and configure data augmentation strategies for more robust models.

### Converting existing corpora and annotations {id="data-convert"}

If you have training data in a standard format like `.conll` or `.conllu`, the
easiest way to convert it for use with spaCy is to run
[`spacy convert`](/api/cli#convert) and pass it a file and an output directory.
By default, the command will pick the converter based on the file extension.

```bash
$ python -m spacy convert ./train.gold.conll ./corpus
```

> #### 💡 Tip: Converting from Prodigy
>
> If you're using the [Prodigy](https://prodi.gy) annotation tool to create
> training data, you can run the
> [`data-to-spacy` command](https://prodi.gy/docs/recipes#data-to-spacy) to
> merge and export multiple datasets for use with
> [`spacy train`](/api/cli#train). Different types of annotations on the same
> text will be combined, giving you one corpus to train multiple components.

<Infobox title="Tip: Manage multi-step workflows with projects" emoji="💡">

Training workflows often consist of multiple steps, from preprocessing the data
all the way to packaging and deploying the trained model.
[spaCy projects](/usage/projects) let you define all steps in one file, manage
data assets, track changes and share your end-to-end processes with your team.

</Infobox>

The binary `.spacy` format is a serialized [`DocBin`](/api/docbin) containing
one or more [`Doc`](/api/doc) objects. It's extremely **efficient in storage**,
especially when packing multiple documents together. You can also create `Doc`
objects manually, so you can write your own custom logic to convert and store
existing annotations for use in spaCy.

```python {title="Training data from Doc objects",highlight="6-9"}
import spacy
from spacy.tokens import Doc, DocBin

nlp = spacy.blank("en")
docbin = DocBin()
words = ["Apple", "is", "looking", "at", "buying", "U.K.", "startup", "."]
spaces = [True, True, True, True, True, True, True, False]
ents = ["B-ORG", "O", "O", "O", "O", "B-GPE", "O", "O"]
doc = Doc(nlp.vocab, words=words, spaces=spaces, ents=ents)
docbin.add(doc)
docbin.to_disk("./train.spacy")
```

### Working with corpora {id="data-corpora"}

> #### Example
>
> ```ini
> [corpora]
>
> [corpora.train]
> @readers = "spacy.Corpus.v1"
> path = ${paths.train}
> gold_preproc = false
> max_length = 0
> limit = 0
> augmenter = null
>
> [training]
> train_corpus = "corpora.train"
> ```

The [`[corpora]`](/api/data-formats#config-corpora) block in your config lets
you define **data resources** to use for training, evaluation, pretraining or
any other custom workflows. `corpora.train` and `corpora.dev` are used as
conventions within spaCy's default configs, but you can also define any other
custom blocks. Each section in the corpora config should resolve to a
[`Corpus`](/api/corpus) – for example, using spaCy's built-in
[corpus reader](/api/top-level#corpus-readers) that takes a path to a binary
`.spacy` file. The `train_corpus` and `dev_corpus` fields in the
[`[training]`](/api/data-formats#config-training) block specify where to find
the corpus in your config. This makes it easy to **swap out** different corpora
by only changing a single config setting.

Instead of making `[corpora]` a block with multiple subsections for each portion
of the data, you can also use a single function that returns a dictionary of
corpora, keyed by corpus name, e.g. `"train"` and `"dev"`. This can be
especially useful if you need to split a single file into corpora for training
and evaluation, without loading the same file twice.

By default, the training data is loaded into memory and shuffled before each
epoch. If the corpus is **too large to fit into memory** during training, stream
the corpus using a custom reader as described in the next section.

### Custom data reading and batching {id="custom-code-readers-batchers"}

Some use-cases require **streaming in data** or manipulating datasets on the
fly, rather than generating all data beforehand and storing it to disk. Instead
of using the built-in [`Corpus`](/api/corpus) reader, which uses static file
paths, you can create and register a custom function that generates
[`Example`](/api/example) objects.

In the following example we assume a custom function `read_custom_data` which
loads or generates texts with relevant text classification annotations. Then,
small lexical variations of the input text are created before generating the
final [`Example`](/api/example) objects. The `@spacy.registry.readers` decorator
lets you register the function creating the custom reader in the `readers`
[registry](/api/top-level#registry) and assign it a string name, so it can be
used in your config. All arguments on the registered function become available
as **config settings** – in this case, `source`.

> #### config.cfg
>
> ```ini
> [corpora.train]
> @readers = "corpus_variants.v1"
> source = "s3://your_bucket/path/data.csv"
> ```

```python {title="functions.py",highlight="7-8"}
from typing import Callable, Iterator, List
import spacy
from spacy.training import Example
from spacy.language import Language
import random

@spacy.registry.readers("corpus_variants.v1")
def stream_data(source: str) -> Callable[[Language], Iterator[Example]]:
    def generate_stream(nlp):
        for text, cats in read_custom_data(source):
            # Create a random variant of the example text
            i = random.randint(0, len(text) - 1)
            variant = text[:i] + text[i].upper() + text[i + 1:]
            doc = nlp.make_doc(variant)
            example = Example.from_dict(doc, {"cats": cats})
            yield example

    return generate_stream
```

<Infobox variant="warning">

Remember that a registered function should always be a function that spaCy
**calls to create something**. In this case, it **creates the reader function**
– it's not the reader itself.

</Infobox>

If the corpus is **too large to load into memory** or the corpus reader is an
**infinite generator**, use the setting `max_epochs = -1` to indicate that the
train corpus should be streamed. With this setting the train corpus is merely
streamed and batched, not shuffled, so any shuffling needs to be implemented in
the corpus reader itself. In the example below, a corpus reader that generates
sentences containing even or odd numbers is used with an unlimited number of
examples for the train corpus and a limited number of examples for the dev
corpus. The dev corpus should always be finite and fit in memory during the
evaluation step. `max_steps` and/or `patience` are used to determine when the
training should stop.

> #### config.cfg
>
> ```ini
> [corpora.dev]
> @readers = "even_odd.v1"
> limit = 100
>
> [corpora.train]
> @readers = "even_odd.v1"
> limit = -1
>
> [training]
> max_epochs = -1
> patience = 500
> max_steps = 2000
> ```

```python {title="functions.py"}
from typing import Callable, Iterable, Iterator
from spacy import util
import random
from spacy.training import Example
from spacy import Language


@util.registry.readers("even_odd.v1")
def create_even_odd_corpus(limit: int = -1) -> Callable[[Language], Iterable[Example]]:
    return EvenOddCorpus(limit)


class EvenOddCorpus:
    def __init__(self, limit):
        self.limit = limit

    def __call__(self, nlp: Language) -> Iterator[Example]:
        i = 0
        while i < self.limit or self.limit < 0:
            r = random.randint(0, 1000)
            cat = r % 2 == 0
            text = "This is sentence " + str(r)
            yield Example.from_dict(
                nlp.make_doc(text), {"cats": {"EVEN": cat, "ODD": not cat}}
            )
            i += 1
```

> #### config.cfg
>
> ```ini
> [initialize.components.textcat.labels]
> @readers = "spacy.read_labels.v1"
> path = "labels/textcat.json"
> require = true
> ```

If the train corpus is streamed, the initialize step peeks at the first 100
examples in the corpus to find the labels for each component. If this isn't
sufficient, you'll need to [provide the labels](#initialization-labels) for each
component in the `[initialize]` block. [`init labels`](/api/cli#init-labels) can
be used to generate JSON files in the correct format, which you can extend with
the full label set.

We can also customize the **batching strategy** by registering a new batcher
function in the `batchers` [registry](/api/top-level#registry). A batcher turns
a stream of items into a stream of batches. spaCy has several useful built-in
[batching strategies](/api/top-level#batchers) with customizable sizes, but it's
also easy to implement your own. For instance, the following function takes the
stream of generated [`Example`](/api/example) objects, and removes those which
have the same underlying raw text, to avoid duplicates within each batch. Note
that in a more realistic implementation, you'd also want to check whether the
annotations are the same.

> #### config.cfg
>
> ```ini
> [training.batcher]
> @batchers = "filtering_batch.v1"
> size = 150
> ```

```python {title="functions.py"}
from typing import Callable, Iterable, Iterator, List
import spacy
from spacy.training import Example

@spacy.registry.batchers("filtering_batch.v1")
def filter_batch(size: int) -> Callable[[Iterable[Example]], Iterator[List[Example]]]:
    def create_filtered_batches(examples):
        batch = []
        for eg in examples:
            # Remove duplicate examples with the same text from batch
            if eg.text not in [x.text for x in batch]:
                batch.append(eg)
            if len(batch) == size:
                yield batch
                batch = []

    return create_filtered_batches
```

{/* TODO: Custom corpus class, Minibatching */}

### Data augmentation {id="data-augmentation"}

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.

The easiest way to use data augmentation during training is to provide an
`augmenter` to the training corpus, e.g. in the `[corpora.train]` section of
your config. The built-in [`orth_variants`](/api/top-level#orth_variants)
augmenter creates a data augmentation callback that uses orth-variant
replacement.

```ini {title="config.cfg (excerpt)",highlight="8,14"}
[corpora.train]
@readers = "spacy.Corpus.v1"
path = ${paths.train}
gold_preproc = false
max_length = 0
limit = 0

[corpora.train.augmenter]
@augmenters = "spacy.orth_variants.v1"
# Percentage of texts that will be augmented / lowercased
level = 0.1
lower = 0.5

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

The `orth_variants` argument lets you pass in a dictionary of replacement rules,
typically loaded from a JSON file. There are two types of orth variant rules:
`"single"` for single tokens that should be replaced (e.g. hyphens) and
`"paired"` for pairs of tokens (e.g. quotes).

```json {title="orth_variants.json"}
{
  "single": [{ "tags": ["NFP"], "variants": ["…", "..."] }],
  "paired": [
    {
      "tags": ["``", "''"],
      "variants": [
        ["'", "'"],
        ["‘", "’"]
      ]
    }
  ]
}
```

<Accordion title="Full examples for English and German" spaced>

```json
https://github.com/explosion/spacy-lookups-data/blob/master/spacy_lookups_data/data/en_orth_variants.json
```

```json
https://github.com/explosion/spacy-lookups-data/blob/master/spacy_lookups_data/data/de_orth_variants.json
```

</Accordion>

<Infobox title="Important note" variant="warning">

When adding data augmentation, keep in mind that it typically only makes sense
to apply it to the **training corpus**, not the development data.

</Infobox>

#### Writing custom data augmenters {id="data-augmentation-custom"}

Using the [`@spacy.augmenters`](/api/top-level#registry) registry, you can also
register your own data augmentation callbacks. The callback should be a function
that takes the current `nlp` object and a training [`Example`](/api/example) and
yields `Example` objects. Keep in mind that the augmenter should yield **all
examples** you want to use in your corpus, not only the augmented examples
(unless you want to augment all examples).

Here'a an example of a custom augmentation callback that produces text variants
in ["SpOnGeBoB cAsE"](https://knowyourmeme.com/memes/mocking-spongebob). The
registered function takes one argument `randomize` that can be set via the
config and decides whether the uppercase/lowercase transformation is applied
randomly or not. The augmenter yields two `Example` objects: the original
example and the augmented example.

> #### config.cfg
>
> ```ini
> [corpora.train.augmenter]
> @augmenters = "spongebob_augmenter.v1"
> randomize = false
> ```

```python
import spacy
import random

@spacy.registry.augmenters("spongebob_augmenter.v1")
def create_augmenter(randomize: bool = False):
    def augment(nlp, example):
        text = example.text
        if randomize:
            # Randomly uppercase/lowercase characters
            chars = [c.lower() if random.random() < 0.5 else c.upper() for c in text]
        else:
            # Uppercase followed by lowercase
            chars = [c.lower() if i % 2 else c.upper() for i, c in enumerate(text)]
        # Create augmented training example
        example_dict = example.to_dict()
        doc = nlp.make_doc("".join(chars))
        example_dict["token_annotation"]["ORTH"] = [t.text for t in doc]
        # Original example followed by augmented example
        yield example
        yield example.from_dict(doc, example_dict)

    return augment
```

An easy way to create modified `Example` objects is to use the
[`Example.from_dict`](/api/example#from_dict) method with a new reference
[`Doc`](/api/doc) created from the modified text. In this case, only the
capitalization changes, so only the `ORTH` values of the tokens will be
different between the original and augmented examples.

Note that if your data augmentation strategy involves changing the tokenization
(for instance, removing or adding tokens) and your training examples include
token-based annotations like the dependency parse or entity labels, you'll need
to take care to adjust the `Example` object so its annotations match and remain
valid.

## Parallel & distributed training with Ray {id="parallel-training"}

> #### Installation
>
> ```bash
> $ pip install -U %%SPACY_PKG_NAME[ray]%%SPACY_PKG_FLAGS
> # Check that the CLI is registered
> $ python -m spacy ray --help
> ```

[Ray](https://ray.io/) is a fast and simple framework for building and running
**distributed applications**. You can use Ray to train spaCy on one or more
remote machines, potentially speeding up your training process. Parallel
training won't always be faster though – it depends on your batch size, models,
and hardware.

<Infobox variant="warning">

To use Ray with spaCy, you need the
[`spacy-ray`](https://github.com/explosion/spacy-ray) package installed.
Installing the package will automatically add the `ray` command to the spaCy
CLI.

</Infobox>

The [`spacy ray train`](/api/cli#ray-train) command follows the same API as
[`spacy train`](/api/cli#train), with a few extra options to configure the Ray
setup. You can optionally set the `--address` option to point to your Ray
cluster. If it's not set, Ray will run locally.

```bash
python -m spacy ray train config.cfg --n-workers 2
```

<Project id="integrations/ray">

Get started with parallel training using our project template. It trains a
simple model on a Universal Dependencies Treebank and lets you parallelize the
training with Ray.

</Project>

### How parallel training works {id="parallel-training-details"}

Each worker receives a shard of the **data** and builds a copy of the **model
and optimizer** from the [`config.cfg`](#config). It also has a communication
channel to **pass gradients and parameters** to the other workers. Additionally,
each worker is given ownership of a subset of the parameter arrays. Every
parameter array is owned by exactly one worker, and the workers are given a
mapping so they know which worker owns which parameter.

![Illustration of setup](/images/spacy-ray.svg)

As training proceeds, every worker will be computing gradients for **all** of
the model parameters. When they compute gradients for parameters they don't own,
they'll **send them to the worker** that does own that parameter, along with a
version identifier so that the owner can decide whether to discard the gradient.
Workers use the gradients they receive and the ones they compute locally to
update the parameters they own, and then broadcast the updated array and a new
version ID to the other workers.

This training procedure is **asynchronous** and **non-blocking**. Workers always
push their gradient increments and parameter updates, they do not have to pull
them and block on the result, so the transfers can happen in the background,
overlapped with the actual training work. The workers also do not have to stop
and wait for each other ("synchronize") at the start of each batch. This is very
useful for spaCy, because spaCy is often trained on long documents, which means
**batches can vary in size** significantly. Uneven workloads make synchronous
gradient descent inefficient, because if one batch is slow, all of the other
workers are stuck waiting for it to complete before they can continue.

## Internal training API {id="api"}

<Infobox variant="danger">

spaCy gives you full control over the training loop. However, for most use
cases, it's recommended to train your pipelines via the
[`spacy train`](/api/cli#train) command with a [`config.cfg`](#config) to keep
track of your settings and hyperparameters, instead of writing your own training
scripts from scratch. [Custom registered functions](#custom-code) should
typically give you everything you need to train fully custom pipelines with
[`spacy train`](/api/cli#train).

</Infobox>

### Training from a Python script {id="api-train",version="3.2"}

If you want to run the training from a Python script instead of using the
[`spacy train`](/api/cli#train) CLI command, you can call into the
[`train`](/api/cli#train-function) helper function directly. It takes the path
to the config file, an optional output directory and an optional dictionary of
[config overrides](#config-overrides).

```python
from spacy.cli.train import train

train("./config.cfg", overrides={"paths.train": "./train.spacy", "paths.dev": "./dev.spacy"})
```

### Internal training loop API {id="api-loop"}

<Infobox variant="warning">

This section documents how the training loop and updates to the `nlp` object
work internally. You typically shouldn't have to implement this in Python unless
you're writing your own trainable components. To train a pipeline, use
[`spacy train`](/api/cli#train) or the [`train`](/api/cli#train-function) helper
function instead.

</Infobox>

The [`Example`](/api/example) object contains annotated training data, also
called the **gold standard**. It's initialized with a [`Doc`](/api/doc) object
that will hold the predictions, and another `Doc` object that holds the
gold-standard annotations. It also includes the **alignment** between those two
documents if they differ in tokenization. The `Example` class ensures that spaCy
can rely on one **standardized format** that's passed through the pipeline. For
instance, let's say we want to define gold-standard part-of-speech tags:

```python
words = ["I", "like", "stuff"]
predicted = Doc(vocab, words=words)
# create the reference Doc with gold-standard TAG annotations
tags = ["NOUN", "VERB", "NOUN"]
tag_ids = [vocab.strings.add(tag) for tag in tags]
reference = Doc(vocab, words=words).from_array("TAG", numpy.array(tag_ids, dtype="uint64"))
example = Example(predicted, reference)
```

As this is quite verbose, there's an alternative way to create the reference
`Doc` with the gold-standard annotations. The function `Example.from_dict` takes
a dictionary with keyword arguments specifying the annotations, like `tags` or
`entities`. Using the resulting `Example` object and its gold-standard
annotations, the model can be updated to learn a sentence of three words with
their assigned part-of-speech tags.

```python
words = ["I", "like", "stuff"]
tags = ["NOUN", "VERB", "NOUN"]
predicted = Doc(nlp.vocab, words=words)
example = Example.from_dict(predicted, {"tags": tags})
```

Here's another example that shows how to define gold-standard named entities.
The letters added before the labels refer to the tags of the
[BILUO scheme](/usage/linguistic-features#updating-biluo) – `O` is a token
outside an entity, `U` a single entity unit, `B` the beginning of an entity, `I`
a token inside an entity and `L` the last token of an entity.

```python
doc = Doc(nlp.vocab, words=["Facebook", "released", "React", "in", "2014"])
example = Example.from_dict(doc, {"entities": ["U-ORG", "O", "U-TECHNOLOGY", "O", "U-DATE"]})
```

<Infobox title="Migrating from v2.x" variant="warning">

As of v3.0, the [`Example`](/api/example) object replaces the `GoldParse` class.
It can be constructed in a very similar way – from a `Doc` and a dictionary of
annotations. For more details, see the
[migration guide](/usage/v3#migrating-training).

```diff
- gold = GoldParse(doc, entities=entities)
+ example = Example.from_dict(doc, {"entities": entities})
```

</Infobox>

Of course, it's not enough to only show a model a single example once.
Especially if you only have few examples, you'll want to train for a **number of
iterations**. At each iteration, the training data is **shuffled** to ensure the
model doesn't make any generalizations based on the order of examples. Another
technique to improve the learning results is to set a **dropout rate**, a rate
at which to randomly "drop" individual features and representations. This makes
it harder for the model to memorize the training data. For example, a `0.25`
dropout means that each feature or internal representation has a 1/4 likelihood
of being dropped.

> - [`nlp`](/api/language): The `nlp` object with the pipeline components and
>   their models.
> - [`nlp.initialize`](/api/language#initialize): Initialize the pipeline and
>   return an optimizer to update the component model weights.
> - [`Optimizer`](https://thinc.ai/docs/api-optimizers): Function that holds
>   state between updates.
> - [`nlp.update`](/api/language#update): Update component models with examples.
> - [`Example`](/api/example): object holding predictions and gold-standard
>   annotations.
> - [`nlp.to_disk`](/api/language#to_disk): Save the updated pipeline to a
>   directory.

```python {title="Example training loop"}
optimizer = nlp.initialize()
for itn in range(100):
    random.shuffle(train_data)
    for raw_text, entity_offsets in train_data:
        doc = nlp.make_doc(raw_text)
        example = Example.from_dict(doc, {"entities": entity_offsets})
        nlp.update([example], sgd=optimizer)
nlp.to_disk("/output")
```

The [`nlp.update`](/api/language#update) method takes the following arguments:

| Name       | Description                                                                                                                                                            |
| ---------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `examples` | [`Example`](/api/example) objects. The `update` method takes a sequence of them, so you can batch up your training examples.                                           |
| `drop`     | Dropout rate. Makes it harder for the model to just memorize the data.                                                                                                 |
| `sgd`      | An [`Optimizer`](https://thinc.ai/docs/api-optimizers) object, which updates the model's weights. If not set, spaCy will create a new one and save it for further use. |

<Infobox title="Migrating from v2.x" variant="warning">

As of v3.0, the [`Example`](/api/example) object replaces the `GoldParse` class
and the "simple training style" of calling `nlp.update` with a text and a
dictionary of annotations. Updating your code to use the `Example` object should
be very straightforward: you can call
[`Example.from_dict`](/api/example#from_dict) with a [`Doc`](/api/doc) and the
dictionary of annotations:

```diff
text = "Facebook released React in 2014"
annotations = {"entities": ["U-ORG", "O", "U-TECHNOLOGY", "O", "U-DATE"]}
+ example = Example.from_dict(nlp.make_doc(text), annotations)
- nlp.update([text], [annotations])
+ nlp.update([example])
```

</Infobox>