2020-07-22 14:42:59 +03:00
|
|
|
|
from typing import Optional, Any, Dict, Callable, Iterable, Union, List, Pattern
|
2020-10-10 20:14:48 +03:00
|
|
|
|
from typing import Tuple
|
2020-07-22 14:42:59 +03:00
|
|
|
|
from dataclasses import dataclass
|
2017-05-25 04:10:54 +03:00
|
|
|
|
import random
|
2017-07-25 19:57:59 +03:00
|
|
|
|
import itertools
|
2017-10-17 19:18:10 +03:00
|
|
|
|
import functools
|
2017-10-27 22:07:59 +03:00
|
|
|
|
from contextlib import contextmanager
|
2020-08-28 16:20:14 +03:00
|
|
|
|
from copy import deepcopy
|
2020-02-27 20:42:27 +03:00
|
|
|
|
from pathlib import Path
|
2020-02-28 14:20:23 +03:00
|
|
|
|
import warnings
|
2020-09-29 13:14:08 +03:00
|
|
|
|
from thinc.api import Model, get_current_ops, Config, Optimizer
|
💫 Replace ujson, msgpack and dill/pickle/cloudpickle with srsly (#3003)
Remove hacks and wrappers, keep code in sync across our libraries and move spaCy a few steps closer to only depending on packages with binary wheels 🎉
See here: https://github.com/explosion/srsly
Serialization is hard, especially across Python versions and multiple platforms. After dealing with many subtle bugs over the years (encodings, locales, large files) our libraries like spaCy and Prodigy have steadily grown a number of utility functions to wrap the multiple serialization formats we need to support (especially json, msgpack and pickle). These wrapping functions ended up duplicated across our codebases, so we wanted to put them in one place.
At the same time, we noticed that having a lot of small dependencies was making maintainence harder, and making installation slower. To solve this, we've made srsly standalone, by including the component packages directly within it. This way we can provide all the serialization utilities we need in a single binary wheel.
srsly currently includes forks of the following packages:
ujson
msgpack
msgpack-numpy
cloudpickle
* WIP: replace json/ujson with srsly
* Replace ujson in examples
Use regular json instead of srsly to make code easier to read and follow
* Update requirements
* Fix imports
* Fix typos
* Replace msgpack with srsly
* Fix warning
2018-12-03 03:28:22 +03:00
|
|
|
|
import srsly
|
2019-10-08 13:20:55 +03:00
|
|
|
|
import multiprocessing as mp
|
|
|
|
|
from itertools import chain, cycle
|
2020-07-29 12:02:31 +03:00
|
|
|
|
from timeit import default_timer as timer
|
2017-05-18 12:25:19 +03:00
|
|
|
|
|
2020-02-12 13:50:42 +03:00
|
|
|
|
from .tokens.underscore import Underscore
|
2020-07-24 15:50:26 +03:00
|
|
|
|
from .vocab import Vocab, create_vocab
|
2020-08-01 14:40:06 +03:00
|
|
|
|
from .pipe_analysis import validate_attrs, analyze_pipes, print_pipe_analysis
|
2020-09-09 11:31:03 +03:00
|
|
|
|
from .training import Example, validate_examples
|
2020-09-29 17:05:48 +03:00
|
|
|
|
from .training.initialize import init_vocab, init_tok2vec
|
2017-10-07 01:26:05 +03:00
|
|
|
|
from .scorer import Scorer
|
2020-10-08 22:33:49 +03:00
|
|
|
|
from .util import registry, SimpleFrozenList, _pipe
|
2020-08-14 15:06:22 +03:00
|
|
|
|
from .util import SimpleFrozenDict, combine_score_weights, CONFIG_SECTION_ORDER
|
2020-07-24 15:50:26 +03:00
|
|
|
|
from .lang.tokenizer_exceptions import URL_MATCH, BASE_EXCEPTIONS
|
2017-10-27 15:40:14 +03:00
|
|
|
|
from .lang.punctuation import TOKENIZER_PREFIXES, TOKENIZER_SUFFIXES
|
|
|
|
|
from .lang.punctuation import TOKENIZER_INFIXES
|
2020-07-22 23:18:46 +03:00
|
|
|
|
from .tokens import Doc
|
2020-07-24 15:50:26 +03:00
|
|
|
|
from .tokenizer import Tokenizer
|
2020-04-28 14:37:37 +03:00
|
|
|
|
from .errors import Errors, Warnings
|
2020-09-29 17:05:48 +03:00
|
|
|
|
from .schemas import ConfigSchema, ConfigSchemaNlp, ConfigSchemaInit
|
|
|
|
|
from .schemas import ConfigSchemaPretrain, validate_init_settings
|
2020-07-02 18:10:27 +03:00
|
|
|
|
from .git_info import GIT_VERSION
|
2017-04-15 13:05:47 +03:00
|
|
|
|
from . import util
|
2017-10-07 01:26:05 +03:00
|
|
|
|
from . import about
|
2020-09-18 16:45:55 +03:00
|
|
|
|
from .lookups import load_lookups
|
2016-10-09 13:24:24 +03:00
|
|
|
|
|
2015-08-27 10:16:11 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
# This is the base config will all settings (training etc.)
|
|
|
|
|
DEFAULT_CONFIG_PATH = Path(__file__).parent / "default_config.cfg"
|
2020-08-14 15:06:22 +03:00
|
|
|
|
DEFAULT_CONFIG = util.load_config(DEFAULT_CONFIG_PATH)
|
2020-08-24 16:56:03 +03:00
|
|
|
|
# This is the base config for the [pretraining] block and currently not included
|
|
|
|
|
# in the main config and only added via the 'init fill-config' command
|
|
|
|
|
DEFAULT_CONFIG_PRETRAIN_PATH = Path(__file__).parent / "default_config_pretraining.cfg"
|
2019-10-27 15:35:49 +03:00
|
|
|
|
|
|
|
|
|
|
2020-07-12 15:03:23 +03:00
|
|
|
|
class BaseDefaults:
|
2020-07-29 00:12:42 +03:00
|
|
|
|
"""Language data defaults, available via Language.Defaults. Can be
|
|
|
|
|
overwritten by language subclasses by defining their own subclasses of
|
|
|
|
|
Language.Defaults.
|
|
|
|
|
"""
|
2020-07-29 16:14:07 +03:00
|
|
|
|
|
2020-08-14 15:06:22 +03:00
|
|
|
|
config: Config = Config(section_order=CONFIG_SECTION_ORDER)
|
2020-07-24 15:50:26 +03:00
|
|
|
|
tokenizer_exceptions: Dict[str, List[dict]] = BASE_EXCEPTIONS
|
|
|
|
|
prefixes: Optional[List[Union[str, Pattern]]] = TOKENIZER_PREFIXES
|
|
|
|
|
suffixes: Optional[List[Union[str, Pattern]]] = TOKENIZER_SUFFIXES
|
|
|
|
|
infixes: Optional[List[Union[str, Pattern]]] = TOKENIZER_INFIXES
|
|
|
|
|
token_match: Optional[Pattern] = None
|
|
|
|
|
url_match: Optional[Pattern] = URL_MATCH
|
|
|
|
|
syntax_iterators: Dict[str, Callable] = {}
|
|
|
|
|
lex_attr_getters: Dict[int, Callable[[str], Any]] = {}
|
|
|
|
|
stop_words = set()
|
|
|
|
|
writing_system = {"direction": "ltr", "has_case": True, "has_letters": True}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@registry.tokenizers("spacy.Tokenizer.v1")
|
|
|
|
|
def create_tokenizer() -> Callable[["Language"], Tokenizer]:
|
2020-07-29 00:12:42 +03:00
|
|
|
|
"""Registered function to create a tokenizer. Returns a factory that takes
|
|
|
|
|
the nlp object and returns a Tokenizer instance using the language detaults.
|
|
|
|
|
"""
|
2020-07-29 16:14:07 +03:00
|
|
|
|
|
2020-07-24 15:50:26 +03:00
|
|
|
|
def tokenizer_factory(nlp: "Language") -> Tokenizer:
|
|
|
|
|
prefixes = nlp.Defaults.prefixes
|
|
|
|
|
suffixes = nlp.Defaults.suffixes
|
|
|
|
|
infixes = nlp.Defaults.infixes
|
|
|
|
|
prefix_search = util.compile_prefix_regex(prefixes).search if prefixes else None
|
|
|
|
|
suffix_search = util.compile_suffix_regex(suffixes).search if suffixes else None
|
|
|
|
|
infix_finditer = util.compile_infix_regex(infixes).finditer if infixes else None
|
|
|
|
|
return Tokenizer(
|
|
|
|
|
nlp.vocab,
|
|
|
|
|
rules=nlp.Defaults.tokenizer_exceptions,
|
|
|
|
|
prefix_search=prefix_search,
|
|
|
|
|
suffix_search=suffix_search,
|
|
|
|
|
infix_finditer=infix_finditer,
|
|
|
|
|
token_match=nlp.Defaults.token_match,
|
|
|
|
|
url_match=nlp.Defaults.url_match,
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
return tokenizer_factory
|
|
|
|
|
|
|
|
|
|
|
2020-09-18 20:43:19 +03:00
|
|
|
|
@registry.misc("spacy.LookupsDataLoader.v1")
|
2020-09-18 16:45:55 +03:00
|
|
|
|
def load_lookups_data(lang, tables):
|
2020-09-18 20:43:19 +03:00
|
|
|
|
util.logger.debug(f"Loading lookups from spacy-lookups-data: {tables}")
|
2020-09-18 16:45:55 +03:00
|
|
|
|
lookups = load_lookups(lang=lang, tables=tables)
|
|
|
|
|
return lookups
|
|
|
|
|
|
|
|
|
|
|
2020-07-12 15:03:23 +03:00
|
|
|
|
class Language:
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""A text-processing pipeline. Usually you'll load this once per process,
|
|
|
|
|
and pass the instance around your application.
|
2017-05-19 19:47:24 +03:00
|
|
|
|
|
|
|
|
|
Defaults (class): Settings, data and factory methods for creating the `nlp`
|
|
|
|
|
object and processing pipeline.
|
2020-05-24 18:20:58 +03:00
|
|
|
|
lang (str): Two-letter language ID, i.e. ISO code.
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language
|
2019-03-08 13:42:26 +03:00
|
|
|
|
"""
|
2019-03-11 01:36:47 +03:00
|
|
|
|
|
2016-09-24 21:26:17 +03:00
|
|
|
|
Defaults = BaseDefaults
|
2020-07-22 14:42:59 +03:00
|
|
|
|
lang: str = None
|
|
|
|
|
default_config = DEFAULT_CONFIG
|
2015-08-25 16:37:17 +03:00
|
|
|
|
|
2020-07-24 15:50:26 +03:00
|
|
|
|
factories = SimpleFrozenDict(error=Errors.E957)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
_factory_meta: Dict[str, "FactoryMeta"] = {} # meta by factory
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
def __init__(
|
2020-02-28 13:57:41 +03:00
|
|
|
|
self,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
vocab: Union[Vocab, bool] = True,
|
2020-07-27 01:27:53 +03:00
|
|
|
|
*,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
max_length: int = 10 ** 6,
|
|
|
|
|
meta: Dict[str, Any] = {},
|
|
|
|
|
create_tokenizer: Optional[Callable[["Language"], Callable[[str], Doc]]] = None,
|
2020-12-09 11:13:26 +03:00
|
|
|
|
batch_size: int = 1000,
|
2020-02-28 13:57:41 +03:00
|
|
|
|
**kwargs,
|
2020-07-29 00:12:42 +03:00
|
|
|
|
) -> None:
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""Initialise a Language object.
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
vocab (Vocab): A `Vocab` object. If `True`, a vocab is created.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
meta (dict): Custom meta data for the Language class. Is written to by
|
|
|
|
|
models to add model meta data.
|
2020-07-25 13:14:28 +03:00
|
|
|
|
max_length (int): Maximum number of characters in a single text. The
|
|
|
|
|
current models may run out memory on extremely long texts, due to
|
|
|
|
|
large internal allocations. You should segment these texts into
|
|
|
|
|
meaningful units, e.g. paragraphs, subsections etc, before passing
|
|
|
|
|
them to spaCy. Default maximum length is 1,000,000 charas (1mb). As
|
|
|
|
|
a rule of thumb, if all pipeline components are enabled, spaCy's
|
|
|
|
|
default models currently requires roughly 1GB of temporary memory per
|
2018-03-29 22:45:26 +03:00
|
|
|
|
100,000 characters in one text.
|
2020-07-25 13:14:28 +03:00
|
|
|
|
create_tokenizer (Callable): Function that takes the nlp object and
|
|
|
|
|
returns a tokenizer.
|
2020-12-09 11:13:26 +03:00
|
|
|
|
batch_size (int): Default batch size for pipe and evaluate.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#init
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""
|
2020-07-22 14:42:59 +03:00
|
|
|
|
# We're only calling this to import all factories provided via entry
|
|
|
|
|
# points. The factory decorator applied to these functions takes care
|
|
|
|
|
# of the rest.
|
|
|
|
|
util.registry._entry_point_factories.get_all()
|
|
|
|
|
|
2020-08-13 18:38:30 +03:00
|
|
|
|
self._config = DEFAULT_CONFIG.merge(self.default_config)
|
2017-07-23 01:50:18 +03:00
|
|
|
|
self._meta = dict(meta)
|
2017-10-25 12:57:43 +03:00
|
|
|
|
self._path = None
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self._optimizer = None
|
|
|
|
|
# Component meta and configs are only needed on the instance
|
|
|
|
|
self._pipe_meta: Dict[str, "FactoryMeta"] = {} # meta by component
|
|
|
|
|
self._pipe_configs: Dict[str, Config] = {} # config by component
|
|
|
|
|
|
2020-09-15 14:25:34 +03:00
|
|
|
|
if not isinstance(vocab, Vocab) and vocab is not True:
|
|
|
|
|
raise ValueError(Errors.E918.format(vocab=vocab, vocab_type=type(Vocab)))
|
2017-05-16 12:21:59 +03:00
|
|
|
|
if vocab is True:
|
2020-07-22 14:42:59 +03:00
|
|
|
|
vectors_name = meta.get("vectors", {}).get("name")
|
2020-09-21 11:59:07 +03:00
|
|
|
|
vocab = create_vocab(self.lang, self.Defaults, vectors_name=vectors_name)
|
2019-08-01 18:13:01 +03:00
|
|
|
|
else:
|
|
|
|
|
if (self.lang and vocab.lang) and (self.lang != vocab.lang):
|
|
|
|
|
raise ValueError(Errors.E150.format(nlp=self.lang, vocab=vocab.lang))
|
2020-07-25 16:01:15 +03:00
|
|
|
|
self.vocab: Vocab = vocab
|
2020-07-22 14:42:59 +03:00
|
|
|
|
if self.lang is None:
|
|
|
|
|
self.lang = self.vocab.lang
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self._components = []
|
2020-08-29 13:08:33 +03:00
|
|
|
|
self._disabled = set()
|
2018-03-29 22:45:26 +03:00
|
|
|
|
self.max_length = max_length
|
2020-07-22 14:42:59 +03:00
|
|
|
|
# Create the default tokenizer from the default config
|
|
|
|
|
if not create_tokenizer:
|
|
|
|
|
tokenizer_cfg = {"tokenizer": self._config["nlp"]["tokenizer"]}
|
2020-09-27 23:21:31 +03:00
|
|
|
|
create_tokenizer = registry.resolve(tokenizer_cfg)["tokenizer"]
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self.tokenizer = create_tokenizer(self)
|
2020-12-09 11:13:26 +03:00
|
|
|
|
self.batch_size = batch_size
|
2020-07-22 14:42:59 +03:00
|
|
|
|
|
|
|
|
|
def __init_subclass__(cls, **kwargs):
|
|
|
|
|
super().__init_subclass__(**kwargs)
|
2020-08-13 18:38:30 +03:00
|
|
|
|
cls.default_config = DEFAULT_CONFIG.merge(cls.Defaults.config)
|
2020-07-24 15:50:26 +03:00
|
|
|
|
cls.default_config["nlp"]["lang"] = cls.lang
|
2015-10-12 11:33:11 +03:00
|
|
|
|
|
2017-10-25 12:57:43 +03:00
|
|
|
|
@property
|
|
|
|
|
def path(self):
|
|
|
|
|
return self._path
|
|
|
|
|
|
2017-07-23 01:50:18 +03:00
|
|
|
|
@property
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def meta(self) -> Dict[str, Any]:
|
2020-07-29 00:12:42 +03:00
|
|
|
|
"""Custom meta data of the language class. If a model is loaded, this
|
|
|
|
|
includes details from the model's meta.json.
|
|
|
|
|
|
|
|
|
|
RETURNS (Dict[str, Any]): The meta.
|
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#meta
|
2020-07-29 00:12:42 +03:00
|
|
|
|
"""
|
2020-05-30 16:01:58 +03:00
|
|
|
|
spacy_version = util.get_model_version_range(about.__version__)
|
2019-08-01 18:13:01 +03:00
|
|
|
|
if self.vocab.lang:
|
|
|
|
|
self._meta.setdefault("lang", self.vocab.lang)
|
|
|
|
|
else:
|
|
|
|
|
self._meta.setdefault("lang", self.lang)
|
2020-09-03 14:13:03 +03:00
|
|
|
|
self._meta.setdefault("name", "pipeline")
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
self._meta.setdefault("version", "0.0.0")
|
2020-05-30 16:01:58 +03:00
|
|
|
|
self._meta.setdefault("spacy_version", spacy_version)
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
self._meta.setdefault("description", "")
|
|
|
|
|
self._meta.setdefault("author", "")
|
|
|
|
|
self._meta.setdefault("email", "")
|
|
|
|
|
self._meta.setdefault("url", "")
|
|
|
|
|
self._meta.setdefault("license", "")
|
2020-07-02 18:10:27 +03:00
|
|
|
|
self._meta.setdefault("spacy_git_version", GIT_VERSION)
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
self._meta["vectors"] = {
|
|
|
|
|
"width": self.vocab.vectors_length,
|
|
|
|
|
"vectors": len(self.vocab.vectors),
|
|
|
|
|
"keys": self.vocab.vectors.n_keys,
|
|
|
|
|
"name": self.vocab.vectors.name,
|
|
|
|
|
}
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self._meta["labels"] = dict(self.pipe_labels)
|
2020-07-28 17:14:23 +03:00
|
|
|
|
# TODO: Adding this back to prevent breaking people's code etc., but
|
|
|
|
|
# we should consider removing it
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self._meta["pipeline"] = list(self.pipe_names)
|
2020-09-04 15:42:12 +03:00
|
|
|
|
self._meta["components"] = list(self.component_names)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self._meta["disabled"] = list(self.disabled)
|
2017-07-23 01:50:18 +03:00
|
|
|
|
return self._meta
|
|
|
|
|
|
|
|
|
|
@meta.setter
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def meta(self, value: Dict[str, Any]) -> None:
|
2017-07-23 01:50:18 +03:00
|
|
|
|
self._meta = value
|
|
|
|
|
|
2017-06-04 23:52:09 +03:00
|
|
|
|
@property
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def config(self) -> Config:
|
2020-07-29 00:12:42 +03:00
|
|
|
|
"""Trainable config for the current language instance. Includes the
|
|
|
|
|
current pipeline components, as well as default training config.
|
|
|
|
|
|
|
|
|
|
RETURNS (thinc.api.Config): The config.
|
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#config
|
2020-07-29 00:12:42 +03:00
|
|
|
|
"""
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self._config.setdefault("nlp", {})
|
2020-07-26 14:18:43 +03:00
|
|
|
|
self._config.setdefault("training", {})
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self._config["nlp"]["lang"] = self.lang
|
|
|
|
|
# We're storing the filled config for each pipeline component and so
|
|
|
|
|
# we can populate the config again later
|
|
|
|
|
pipeline = {}
|
2020-07-26 14:18:43 +03:00
|
|
|
|
score_weights = []
|
2020-08-28 22:04:02 +03:00
|
|
|
|
for pipe_name in self.component_names:
|
2020-07-22 14:42:59 +03:00
|
|
|
|
pipe_meta = self.get_pipe_meta(pipe_name)
|
|
|
|
|
pipe_config = self.get_pipe_config(pipe_name)
|
2020-07-22 18:29:31 +03:00
|
|
|
|
pipeline[pipe_name] = {"factory": pipe_meta.factory, **pipe_config}
|
2020-07-27 13:27:40 +03:00
|
|
|
|
if pipe_meta.default_score_weights:
|
|
|
|
|
score_weights.append(pipe_meta.default_score_weights)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self._config["nlp"]["pipeline"] = list(self.component_names)
|
|
|
|
|
self._config["nlp"]["disabled"] = list(self.disabled)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self._config["components"] = pipeline
|
2020-09-24 11:27:33 +03:00
|
|
|
|
# We're merging the existing score weights back into the combined
|
|
|
|
|
# weights to make sure we're preserving custom settings in the config
|
|
|
|
|
# but also reflect updates (e.g. new components added)
|
2020-09-24 11:42:47 +03:00
|
|
|
|
prev_weights = self._config["training"].get("score_weights", {})
|
|
|
|
|
combined_score_weights = combine_score_weights(score_weights, prev_weights)
|
2020-09-24 11:27:33 +03:00
|
|
|
|
self._config["training"]["score_weights"] = combined_score_weights
|
2020-07-22 14:42:59 +03:00
|
|
|
|
if not srsly.is_json_serializable(self._config):
|
|
|
|
|
raise ValueError(Errors.E961.format(config=self._config))
|
2020-02-27 20:42:27 +03:00
|
|
|
|
return self._config
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
@config.setter
|
|
|
|
|
def config(self, value: Config) -> None:
|
|
|
|
|
self._config = value
|
|
|
|
|
|
2020-08-29 13:08:33 +03:00
|
|
|
|
@property
|
|
|
|
|
def disabled(self) -> List[str]:
|
|
|
|
|
"""Get the names of all disabled components.
|
|
|
|
|
|
|
|
|
|
RETURNS (List[str]): The disabled components.
|
|
|
|
|
"""
|
2020-08-29 13:58:22 +03:00
|
|
|
|
# Make sure the disabled components are returned in the order they
|
|
|
|
|
# appear in the pipeline (which isn't guaranteed by the set)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
names = [name for name, _ in self._components if name in self._disabled]
|
|
|
|
|
return SimpleFrozenList(names, error=Errors.E926.format(attr="disabled"))
|
2020-08-29 13:08:33 +03:00
|
|
|
|
|
2017-10-07 01:25:54 +03:00
|
|
|
|
@property
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def factory_names(self) -> List[str]:
|
|
|
|
|
"""Get names of all available factories.
|
|
|
|
|
|
|
|
|
|
RETURNS (List[str]): The factory names.
|
|
|
|
|
"""
|
2020-08-29 16:20:11 +03:00
|
|
|
|
names = list(self.factories.keys())
|
|
|
|
|
return SimpleFrozenList(names)
|
|
|
|
|
|
|
|
|
|
@property
|
|
|
|
|
def components(self) -> List[Tuple[str, Callable[[Doc], Doc]]]:
|
|
|
|
|
"""Get all (name, component) tuples in the pipeline, including the
|
|
|
|
|
currently disabled components.
|
|
|
|
|
"""
|
|
|
|
|
return SimpleFrozenList(
|
|
|
|
|
self._components, error=Errors.E926.format(attr="components")
|
|
|
|
|
)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
|
2020-08-28 16:20:14 +03:00
|
|
|
|
@property
|
2020-08-28 22:04:02 +03:00
|
|
|
|
def component_names(self) -> List[str]:
|
2020-08-28 16:20:14 +03:00
|
|
|
|
"""Get the names of the available pipeline components. Includes all
|
|
|
|
|
active and inactive pipeline components.
|
|
|
|
|
|
|
|
|
|
RETURNS (List[str]): List of component name strings, in order.
|
|
|
|
|
"""
|
2020-08-29 16:20:11 +03:00
|
|
|
|
names = [pipe_name for pipe_name, _ in self._components]
|
|
|
|
|
return SimpleFrozenList(names, error=Errors.E926.format(attr="component_names"))
|
2020-08-28 16:20:14 +03:00
|
|
|
|
|
|
|
|
|
@property
|
|
|
|
|
def pipeline(self) -> List[Tuple[str, Callable[[Doc], Doc]]]:
|
|
|
|
|
"""The processing pipeline consisting of (name, component) tuples. The
|
|
|
|
|
components are called on the Doc in order as it passes through the
|
|
|
|
|
pipeline.
|
|
|
|
|
|
|
|
|
|
RETURNS (List[Tuple[str, Callable[[Doc], Doc]]]): The pipeline.
|
|
|
|
|
"""
|
2020-08-29 16:20:11 +03:00
|
|
|
|
pipes = [(n, p) for n, p in self._components if n not in self._disabled]
|
|
|
|
|
return SimpleFrozenList(pipes, error=Errors.E926.format(attr="pipeline"))
|
2020-07-22 14:42:59 +03:00
|
|
|
|
|
|
|
|
|
@property
|
|
|
|
|
def pipe_names(self) -> List[str]:
|
2020-08-28 16:20:14 +03:00
|
|
|
|
"""Get names of available active pipeline components.
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
RETURNS (List[str]): List of component name strings, in order.
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""
|
2020-08-29 16:20:11 +03:00
|
|
|
|
names = [pipe_name for pipe_name, _ in self.pipeline]
|
|
|
|
|
return SimpleFrozenList(names, error=Errors.E926.format(attr="pipe_names"))
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2019-10-27 15:35:49 +03:00
|
|
|
|
@property
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def pipe_factories(self) -> Dict[str, str]:
|
2019-10-27 15:35:49 +03:00
|
|
|
|
"""Get the component factories for the available pipeline components.
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
RETURNS (Dict[str, str]): Factory names, keyed by component names.
|
2019-10-27 15:35:49 +03:00
|
|
|
|
"""
|
|
|
|
|
factories = {}
|
2020-08-29 16:20:11 +03:00
|
|
|
|
for pipe_name, pipe in self._components:
|
2020-07-22 14:42:59 +03:00
|
|
|
|
factories[pipe_name] = self.get_pipe_meta(pipe_name).factory
|
2020-08-29 16:20:11 +03:00
|
|
|
|
return SimpleFrozenDict(factories)
|
2019-10-27 15:35:49 +03:00
|
|
|
|
|
2019-09-12 11:56:28 +03:00
|
|
|
|
@property
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def pipe_labels(self) -> Dict[str, List[str]]:
|
2019-09-12 14:03:38 +03:00
|
|
|
|
"""Get the labels set by the pipeline components, if available (if
|
|
|
|
|
the component exposes a labels property).
|
2019-09-12 11:56:28 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
RETURNS (Dict[str, List[str]]): Labels keyed by component name.
|
2019-09-12 11:56:28 +03:00
|
|
|
|
"""
|
2019-12-22 03:53:56 +03:00
|
|
|
|
labels = {}
|
2020-08-29 16:20:11 +03:00
|
|
|
|
for name, pipe in self._components:
|
2019-09-12 11:56:28 +03:00
|
|
|
|
if hasattr(pipe, "labels"):
|
|
|
|
|
labels[name] = list(pipe.labels)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
return SimpleFrozenDict(labels)
|
2019-09-12 11:56:28 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
@classmethod
|
|
|
|
|
def has_factory(cls, name: str) -> bool:
|
|
|
|
|
"""RETURNS (bool): Whether a factory of that name is registered."""
|
|
|
|
|
internal_name = cls.get_factory_name(name)
|
|
|
|
|
return name in registry.factories or internal_name in registry.factories
|
|
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
|
def get_factory_name(cls, name: str) -> str:
|
|
|
|
|
"""Get the internal factory name based on the language subclass.
|
|
|
|
|
|
|
|
|
|
name (str): The factory name.
|
|
|
|
|
RETURNS (str): The internal factory name.
|
|
|
|
|
"""
|
|
|
|
|
if cls.lang is None:
|
|
|
|
|
return name
|
|
|
|
|
return f"{cls.lang}.{name}"
|
|
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
|
def get_factory_meta(cls, name: str) -> "FactoryMeta":
|
|
|
|
|
"""Get the meta information for a given factory name.
|
|
|
|
|
|
|
|
|
|
name (str): The component factory name.
|
|
|
|
|
RETURNS (FactoryMeta): The meta for the given factory name.
|
|
|
|
|
"""
|
|
|
|
|
internal_name = cls.get_factory_name(name)
|
|
|
|
|
if internal_name in cls._factory_meta:
|
|
|
|
|
return cls._factory_meta[internal_name]
|
|
|
|
|
if name in cls._factory_meta:
|
|
|
|
|
return cls._factory_meta[name]
|
|
|
|
|
raise ValueError(Errors.E967.format(meta="factory", name=name))
|
|
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
|
def set_factory_meta(cls, name: str, value: "FactoryMeta") -> None:
|
|
|
|
|
"""Set the meta information for a given factory name.
|
|
|
|
|
|
|
|
|
|
name (str): The component factory name.
|
|
|
|
|
value (FactoryMeta): The meta to set.
|
|
|
|
|
"""
|
|
|
|
|
cls._factory_meta[cls.get_factory_name(name)] = value
|
|
|
|
|
|
|
|
|
|
def get_pipe_meta(self, name: str) -> "FactoryMeta":
|
|
|
|
|
"""Get the meta information for a given component name.
|
|
|
|
|
|
|
|
|
|
name (str): The component name.
|
|
|
|
|
RETURNS (FactoryMeta): The meta for the given component name.
|
|
|
|
|
"""
|
|
|
|
|
if name not in self._pipe_meta:
|
|
|
|
|
raise ValueError(Errors.E967.format(meta="component", name=name))
|
|
|
|
|
return self._pipe_meta[name]
|
|
|
|
|
|
|
|
|
|
def get_pipe_config(self, name: str) -> Config:
|
|
|
|
|
"""Get the config used to create a pipeline component.
|
|
|
|
|
|
|
|
|
|
name (str): The component name.
|
|
|
|
|
RETURNS (Config): The config used to create the pipeline component.
|
|
|
|
|
"""
|
|
|
|
|
if name not in self._pipe_configs:
|
|
|
|
|
raise ValueError(Errors.E960.format(name=name))
|
|
|
|
|
pipe_config = self._pipe_configs[name]
|
|
|
|
|
return pipe_config
|
|
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
|
def factory(
|
|
|
|
|
cls,
|
|
|
|
|
name: str,
|
|
|
|
|
*,
|
|
|
|
|
default_config: Dict[str, Any] = SimpleFrozenDict(),
|
2020-08-29 16:20:11 +03:00
|
|
|
|
assigns: Iterable[str] = SimpleFrozenList(),
|
|
|
|
|
requires: Iterable[str] = SimpleFrozenList(),
|
2020-07-22 14:42:59 +03:00
|
|
|
|
retokenizes: bool = False,
|
2020-07-27 13:27:40 +03:00
|
|
|
|
default_score_weights: Dict[str, float] = SimpleFrozenDict(),
|
2020-07-22 14:42:59 +03:00
|
|
|
|
func: Optional[Callable] = None,
|
|
|
|
|
) -> Callable:
|
|
|
|
|
"""Register a new pipeline component factory. Can be used as a decorator
|
|
|
|
|
on a function or classmethod, or called as a function with the factory
|
|
|
|
|
provided as the func keyword argument. To create a component and add
|
|
|
|
|
it to the pipeline, you can use nlp.add_pipe(name).
|
|
|
|
|
|
|
|
|
|
name (str): The name of the component factory.
|
|
|
|
|
default_config (Dict[str, Any]): Default configuration, describing the
|
|
|
|
|
default values of the factory arguments.
|
|
|
|
|
assigns (Iterable[str]): Doc/Token attributes assigned by this component,
|
|
|
|
|
e.g. "token.ent_id". Used for pipeline analyis.
|
|
|
|
|
requires (Iterable[str]): Doc/Token attributes required by this component,
|
|
|
|
|
e.g. "token.ent_id". Used for pipeline analyis.
|
|
|
|
|
retokenizes (bool): Whether the component changes the tokenization.
|
|
|
|
|
Used for pipeline analysis.
|
2020-07-28 12:22:24 +03:00
|
|
|
|
default_score_weights (Dict[str, float]): The scores to report during
|
|
|
|
|
training, and their default weight towards the final score used to
|
|
|
|
|
select the best model. Weights should sum to 1.0 per component and
|
2020-09-24 11:27:33 +03:00
|
|
|
|
will be combined and normalized for the whole pipeline. If None,
|
|
|
|
|
the score won't be shown in the logs or be weighted.
|
2020-07-22 14:42:59 +03:00
|
|
|
|
func (Optional[Callable]): Factory function if not used as a decorator.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#factory
|
2020-07-22 14:42:59 +03:00
|
|
|
|
"""
|
|
|
|
|
if not isinstance(name, str):
|
|
|
|
|
raise ValueError(Errors.E963.format(decorator="factory"))
|
|
|
|
|
if not isinstance(default_config, dict):
|
|
|
|
|
err = Errors.E962.format(
|
|
|
|
|
style="default config", name=name, cfg_type=type(default_config)
|
|
|
|
|
)
|
|
|
|
|
raise ValueError(err)
|
|
|
|
|
|
|
|
|
|
def add_factory(factory_func: Callable) -> Callable:
|
2020-08-28 17:27:22 +03:00
|
|
|
|
internal_name = cls.get_factory_name(name)
|
|
|
|
|
if internal_name in registry.factories:
|
|
|
|
|
# We only check for the internal name here – it's okay if it's a
|
|
|
|
|
# subclass and the base class has a factory of the same name. We
|
|
|
|
|
# also only raise if the function is different to prevent raising
|
|
|
|
|
# if module is reloaded.
|
|
|
|
|
existing_func = registry.factories.get(internal_name)
|
|
|
|
|
if not util.is_same_func(factory_func, existing_func):
|
|
|
|
|
err = Errors.E004.format(
|
|
|
|
|
name=name, func=existing_func, new_func=factory_func
|
|
|
|
|
)
|
|
|
|
|
raise ValueError(err)
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
arg_names = util.get_arg_names(factory_func)
|
|
|
|
|
if "nlp" not in arg_names or "name" not in arg_names:
|
|
|
|
|
raise ValueError(Errors.E964.format(name=name))
|
|
|
|
|
# Officially register the factory so we can later call
|
2020-09-27 23:21:31 +03:00
|
|
|
|
# registry.resolve and refer to it in the config as
|
2020-07-22 14:42:59 +03:00
|
|
|
|
# @factories = "spacy.Language.xyz". We use the class name here so
|
|
|
|
|
# different classes can have different factories.
|
|
|
|
|
registry.factories.register(internal_name, func=factory_func)
|
|
|
|
|
factory_meta = FactoryMeta(
|
|
|
|
|
factory=name,
|
|
|
|
|
default_config=default_config,
|
|
|
|
|
assigns=validate_attrs(assigns),
|
|
|
|
|
requires=validate_attrs(requires),
|
2020-09-24 11:27:33 +03:00
|
|
|
|
scores=list(default_score_weights.keys()),
|
2020-07-27 13:27:40 +03:00
|
|
|
|
default_score_weights=default_score_weights,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
retokenizes=retokenizes,
|
|
|
|
|
)
|
|
|
|
|
cls.set_factory_meta(name, factory_meta)
|
|
|
|
|
# We're overwriting the class attr with a frozen dict to handle
|
|
|
|
|
# backwards-compat (writing to Language.factories directly). This
|
|
|
|
|
# wouldn't work with an instance property and just produce a
|
|
|
|
|
# confusing error – here we can show a custom error
|
|
|
|
|
cls.factories = SimpleFrozenDict(
|
|
|
|
|
registry.factories.get_all(), error=Errors.E957
|
|
|
|
|
)
|
|
|
|
|
return factory_func
|
|
|
|
|
|
|
|
|
|
if func is not None: # Support non-decorator use cases
|
|
|
|
|
return add_factory(func)
|
|
|
|
|
return add_factory
|
|
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
|
def component(
|
|
|
|
|
cls,
|
|
|
|
|
name: Optional[str] = None,
|
|
|
|
|
*,
|
2020-08-29 16:20:11 +03:00
|
|
|
|
assigns: Iterable[str] = SimpleFrozenList(),
|
|
|
|
|
requires: Iterable[str] = SimpleFrozenList(),
|
2020-07-22 14:42:59 +03:00
|
|
|
|
retokenizes: bool = False,
|
|
|
|
|
func: Optional[Callable[[Doc], Doc]] = None,
|
|
|
|
|
) -> Callable:
|
|
|
|
|
"""Register a new pipeline component. Can be used for stateless function
|
|
|
|
|
components that don't require a separate factory. Can be used as a
|
|
|
|
|
decorator on a function or classmethod, or called as a function with the
|
|
|
|
|
factory provided as the func keyword argument. To create a component and
|
|
|
|
|
add it to the pipeline, you can use nlp.add_pipe(name).
|
|
|
|
|
|
|
|
|
|
name (str): The name of the component factory.
|
|
|
|
|
assigns (Iterable[str]): Doc/Token attributes assigned by this component,
|
|
|
|
|
e.g. "token.ent_id". Used for pipeline analyis.
|
|
|
|
|
requires (Iterable[str]): Doc/Token attributes required by this component,
|
|
|
|
|
e.g. "token.ent_id". Used for pipeline analyis.
|
|
|
|
|
retokenizes (bool): Whether the component changes the tokenization.
|
|
|
|
|
Used for pipeline analysis.
|
|
|
|
|
func (Optional[Callable]): Factory function if not used as a decorator.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#component
|
2020-07-22 14:42:59 +03:00
|
|
|
|
"""
|
|
|
|
|
if name is not None and not isinstance(name, str):
|
|
|
|
|
raise ValueError(Errors.E963.format(decorator="component"))
|
|
|
|
|
component_name = name if name is not None else util.get_object_name(func)
|
|
|
|
|
|
|
|
|
|
def add_component(component_func: Callable[[Doc], Doc]) -> Callable:
|
|
|
|
|
if isinstance(func, type): # function is a class
|
|
|
|
|
raise ValueError(Errors.E965.format(name=component_name))
|
|
|
|
|
|
|
|
|
|
def factory_func(nlp: cls, name: str) -> Callable[[Doc], Doc]:
|
|
|
|
|
return component_func
|
|
|
|
|
|
2020-08-28 17:27:22 +03:00
|
|
|
|
internal_name = cls.get_factory_name(name)
|
|
|
|
|
if internal_name in registry.factories:
|
|
|
|
|
# We only check for the internal name here – it's okay if it's a
|
|
|
|
|
# subclass and the base class has a factory of the same name. We
|
|
|
|
|
# also only raise if the function is different to prevent raising
|
|
|
|
|
# if module is reloaded. It's hacky, but we need to check the
|
|
|
|
|
# existing functure for a closure and whether that's identical
|
|
|
|
|
# to the component function (because factory_func created above
|
|
|
|
|
# will always be different, even for the same function)
|
|
|
|
|
existing_func = registry.factories.get(internal_name)
|
|
|
|
|
closure = existing_func.__closure__
|
|
|
|
|
wrapped = [c.cell_contents for c in closure][0] if closure else None
|
|
|
|
|
if util.is_same_func(wrapped, component_func):
|
|
|
|
|
factory_func = existing_func # noqa: F811
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
cls.factory(
|
|
|
|
|
component_name,
|
|
|
|
|
assigns=assigns,
|
|
|
|
|
requires=requires,
|
|
|
|
|
retokenizes=retokenizes,
|
|
|
|
|
func=factory_func,
|
|
|
|
|
)
|
|
|
|
|
return component_func
|
|
|
|
|
|
|
|
|
|
if func is not None: # Support non-decorator use cases
|
|
|
|
|
return add_component(func)
|
|
|
|
|
return add_component
|
|
|
|
|
|
2020-07-31 19:34:35 +03:00
|
|
|
|
def analyze_pipes(
|
|
|
|
|
self,
|
|
|
|
|
*,
|
|
|
|
|
keys: List[str] = ["assigns", "requires", "scores", "retokenizes"],
|
2020-08-01 14:40:06 +03:00
|
|
|
|
pretty: bool = False,
|
2020-07-31 19:34:35 +03:00
|
|
|
|
) -> Optional[Dict[str, Any]]:
|
|
|
|
|
"""Analyze the current pipeline components, print a summary of what
|
|
|
|
|
they assign or require and check that all requirements are met.
|
|
|
|
|
|
|
|
|
|
keys (List[str]): The meta values to display in the table. Corresponds
|
|
|
|
|
to values in FactoryMeta, defined by @Language.factory decorator.
|
2020-08-01 14:40:06 +03:00
|
|
|
|
pretty (bool): Pretty-print the results.
|
|
|
|
|
RETURNS (dict): The data.
|
2020-07-31 19:34:35 +03:00
|
|
|
|
"""
|
2020-08-01 14:40:06 +03:00
|
|
|
|
analysis = analyze_pipes(self, keys=keys)
|
|
|
|
|
if pretty:
|
|
|
|
|
print_pipe_analysis(analysis, keys=keys)
|
|
|
|
|
return analysis
|
2020-07-31 19:34:35 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def get_pipe(self, name: str) -> Callable[[Doc], Doc]:
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""Get a pipeline component for a given component name.
|
|
|
|
|
|
2020-05-24 18:20:58 +03:00
|
|
|
|
name (str): Name of pipeline component to get.
|
2017-10-07 01:25:54 +03:00
|
|
|
|
RETURNS (callable): The pipeline component.
|
2019-03-15 18:23:17 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#get_pipe
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""
|
2020-08-29 16:20:11 +03:00
|
|
|
|
for pipe_name, component in self._components:
|
2017-10-07 01:25:54 +03:00
|
|
|
|
if pipe_name == name:
|
|
|
|
|
return component
|
2020-08-28 22:04:02 +03:00
|
|
|
|
raise KeyError(Errors.E001.format(name=name, opts=self.component_names))
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def create_pipe(
|
|
|
|
|
self,
|
|
|
|
|
factory_name: str,
|
|
|
|
|
name: Optional[str] = None,
|
2020-07-29 00:12:42 +03:00
|
|
|
|
*,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
config: Optional[Dict[str, Any]] = SimpleFrozenDict(),
|
2020-08-13 18:38:30 +03:00
|
|
|
|
raw_config: Optional[Config] = None,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
validate: bool = True,
|
|
|
|
|
) -> Callable[[Doc], Doc]:
|
|
|
|
|
"""Create a pipeline component. Mostly used internally. To create and
|
|
|
|
|
add a component to the pipeline, you can use nlp.add_pipe.
|
|
|
|
|
|
|
|
|
|
factory_name (str): Name of component factory.
|
|
|
|
|
name (Optional[str]): Optional name to assign to component instance.
|
|
|
|
|
Defaults to factory name if not set.
|
|
|
|
|
config (Optional[Dict[str, Any]]): Config parameters to use for this
|
|
|
|
|
component. Will be merged with default config, if available.
|
2020-08-13 18:38:30 +03:00
|
|
|
|
raw_config (Optional[Config]): Internals: the non-interpolated config.
|
2020-07-22 14:42:59 +03:00
|
|
|
|
validate (bool): Whether to validate the component config against the
|
|
|
|
|
arguments and types expected by the factory.
|
|
|
|
|
RETURNS (Callable[[Doc], Doc]): The pipeline component.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#create_pipe
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""
|
2020-07-22 14:42:59 +03:00
|
|
|
|
name = name if name is not None else factory_name
|
|
|
|
|
if not isinstance(config, dict):
|
|
|
|
|
err = Errors.E962.format(style="config", name=name, cfg_type=type(config))
|
|
|
|
|
raise ValueError(err)
|
|
|
|
|
if not srsly.is_json_serializable(config):
|
|
|
|
|
raise ValueError(Errors.E961.format(config=config))
|
|
|
|
|
if not self.has_factory(factory_name):
|
|
|
|
|
err = Errors.E002.format(
|
|
|
|
|
name=factory_name,
|
|
|
|
|
opts=", ".join(self.factory_names),
|
|
|
|
|
method="create_pipe",
|
|
|
|
|
lang=util.get_object_name(self),
|
|
|
|
|
lang_code=self.lang,
|
2020-05-21 19:39:06 +03:00
|
|
|
|
)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
raise ValueError(err)
|
|
|
|
|
pipe_meta = self.get_factory_meta(factory_name)
|
|
|
|
|
config = config or {}
|
|
|
|
|
# This is unideal, but the alternative would mean you always need to
|
|
|
|
|
# specify the full config settings, which is not really viable.
|
|
|
|
|
if pipe_meta.default_config:
|
2020-08-13 18:38:30 +03:00
|
|
|
|
config = Config(pipe_meta.default_config).merge(config)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
# We need to create a top-level key because Thinc doesn't allow resolving
|
|
|
|
|
# top-level references to registered functions. Also gives nicer errors.
|
|
|
|
|
# The name allows components to know their pipe name and use it in the
|
|
|
|
|
# losses etc. (even if multiple instances of the same factory are used)
|
|
|
|
|
internal_name = self.get_factory_name(factory_name)
|
|
|
|
|
# If the language-specific factory doesn't exist, try again with the
|
|
|
|
|
# not-specific name
|
|
|
|
|
if internal_name not in registry.factories:
|
|
|
|
|
internal_name = factory_name
|
|
|
|
|
config = {"nlp": self, "name": name, **config, "@factories": internal_name}
|
|
|
|
|
cfg = {factory_name: config}
|
|
|
|
|
# We're calling the internal _fill here to avoid constructing the
|
|
|
|
|
# registered functions twice
|
2020-09-27 23:21:31 +03:00
|
|
|
|
resolved = registry.resolve(cfg, validate=validate)
|
|
|
|
|
filled = registry.fill({"cfg": cfg[factory_name]}, validate=validate)["cfg"]
|
|
|
|
|
filled = Config(filled)
|
2020-07-22 18:29:31 +03:00
|
|
|
|
filled["factory"] = factory_name
|
2020-07-26 16:11:24 +03:00
|
|
|
|
filled.pop("@factories", None)
|
2020-09-15 15:24:17 +03:00
|
|
|
|
# Remove the extra values we added because we don't want to keep passing
|
|
|
|
|
# them around, copying them etc.
|
|
|
|
|
filled.pop("nlp", None)
|
|
|
|
|
filled.pop("name", None)
|
2020-08-13 18:38:30 +03:00
|
|
|
|
# Merge the final filled config with the raw config (including non-
|
|
|
|
|
# interpolated variables)
|
|
|
|
|
if raw_config:
|
|
|
|
|
filled = filled.merge(raw_config)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self._pipe_configs[name] = filled
|
|
|
|
|
return resolved[factory_name]
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2020-08-05 00:39:19 +03:00
|
|
|
|
def create_pipe_from_source(
|
2020-09-08 23:44:25 +03:00
|
|
|
|
self, source_name: str, source: "Language", *, name: str
|
2020-08-05 00:39:19 +03:00
|
|
|
|
) -> Tuple[Callable[[Doc], Doc], str]:
|
|
|
|
|
"""Create a pipeline component by copying it from an existing model.
|
|
|
|
|
|
|
|
|
|
source_name (str): Name of the component in the source pipeline.
|
|
|
|
|
source (Language): The source nlp object to copy from.
|
|
|
|
|
name (str): Optional alternative name to use in current pipeline.
|
|
|
|
|
RETURNS (Tuple[Callable, str]): The component and its factory name.
|
|
|
|
|
"""
|
|
|
|
|
# TODO: handle errors and mismatches (vectors etc.)
|
|
|
|
|
if not isinstance(source, self.__class__):
|
|
|
|
|
raise ValueError(Errors.E945.format(name=source_name, source=type(source)))
|
|
|
|
|
if not source.has_pipe(source_name):
|
|
|
|
|
raise KeyError(
|
|
|
|
|
Errors.E944.format(
|
|
|
|
|
name=source_name,
|
|
|
|
|
model=f"{source.meta['lang']}_{source.meta['name']}",
|
|
|
|
|
opts=", ".join(source.pipe_names),
|
|
|
|
|
)
|
|
|
|
|
)
|
|
|
|
|
pipe = source.get_pipe(source_name)
|
2020-08-13 18:38:30 +03:00
|
|
|
|
# Make sure the source config is interpolated so we don't end up with
|
|
|
|
|
# orphaned variables in our final config
|
|
|
|
|
source_config = source.config.interpolate()
|
|
|
|
|
pipe_config = util.copy_config(source_config["components"][source_name])
|
2020-08-05 00:39:19 +03:00
|
|
|
|
self._pipe_configs[name] = pipe_config
|
2021-01-16 04:26:15 +03:00
|
|
|
|
for s in source.vocab.strings:
|
|
|
|
|
self.vocab.strings.add(s)
|
2020-08-05 00:39:19 +03:00
|
|
|
|
return pipe, pipe_config["factory"]
|
|
|
|
|
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
def add_pipe(
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self,
|
|
|
|
|
factory_name: str,
|
|
|
|
|
name: Optional[str] = None,
|
|
|
|
|
*,
|
|
|
|
|
before: Optional[Union[str, int]] = None,
|
|
|
|
|
after: Optional[Union[str, int]] = None,
|
|
|
|
|
first: Optional[bool] = None,
|
|
|
|
|
last: Optional[bool] = None,
|
2020-08-05 00:39:19 +03:00
|
|
|
|
source: Optional["Language"] = None,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
config: Optional[Dict[str, Any]] = SimpleFrozenDict(),
|
2020-08-13 18:38:30 +03:00
|
|
|
|
raw_config: Optional[Config] = None,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
validate: bool = True,
|
|
|
|
|
) -> Callable[[Doc], Doc]:
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""Add a component to the processing pipeline. Valid components are
|
2017-10-27 15:40:14 +03:00
|
|
|
|
callables that take a `Doc` object, modify it and return it. Only one
|
|
|
|
|
of before/after/first/last can be set. Default behaviour is "last".
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
factory_name (str): Name of the component factory.
|
2020-05-24 18:20:58 +03:00
|
|
|
|
name (str): Name of pipeline component. Overwrites existing
|
2017-10-07 01:25:54 +03:00
|
|
|
|
component.name attribute if available. If no name is set and
|
|
|
|
|
the component exposes no name attribute, component.__name__ is
|
2017-10-27 15:40:14 +03:00
|
|
|
|
used. An error is raised if a name already exists in the pipeline.
|
2020-07-22 14:42:59 +03:00
|
|
|
|
before (Union[str, int]): Name or index of the component to insert new
|
|
|
|
|
component directly before.
|
|
|
|
|
after (Union[str, int]): Name or index of the component to insert new
|
|
|
|
|
component directly after.
|
|
|
|
|
first (bool): If True, insert component first in the pipeline.
|
|
|
|
|
last (bool): If True, insert component last in the pipeline.
|
2020-08-05 00:39:19 +03:00
|
|
|
|
source (Language): Optional loaded nlp object to copy the pipeline
|
|
|
|
|
component from.
|
2020-07-22 14:42:59 +03:00
|
|
|
|
config (Optional[Dict[str, Any]]): Config parameters to use for this
|
|
|
|
|
component. Will be merged with default config, if available.
|
2020-08-13 18:38:30 +03:00
|
|
|
|
raw_config (Optional[Config]): Internals: the non-interpolated config.
|
2020-07-22 14:42:59 +03:00
|
|
|
|
validate (bool): Whether to validate the component config against the
|
|
|
|
|
arguments and types expected by the factory.
|
|
|
|
|
RETURNS (Callable[[Doc], Doc]): The pipeline component.
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#add_pipe
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""
|
2020-07-22 14:42:59 +03:00
|
|
|
|
if not isinstance(factory_name, str):
|
|
|
|
|
bad_val = repr(factory_name)
|
|
|
|
|
err = Errors.E966.format(component=bad_val, name=name)
|
|
|
|
|
raise ValueError(err)
|
|
|
|
|
name = name if name is not None else factory_name
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if name in self.component_names:
|
|
|
|
|
raise ValueError(Errors.E007.format(name=name, opts=self.component_names))
|
2020-08-05 00:39:19 +03:00
|
|
|
|
if source is not None:
|
|
|
|
|
# We're loading the component from a model. After loading the
|
|
|
|
|
# component, we know its real factory name
|
|
|
|
|
pipe_component, factory_name = self.create_pipe_from_source(
|
|
|
|
|
factory_name, source, name=name
|
|
|
|
|
)
|
|
|
|
|
else:
|
|
|
|
|
if not self.has_factory(factory_name):
|
|
|
|
|
err = Errors.E002.format(
|
|
|
|
|
name=factory_name,
|
|
|
|
|
opts=", ".join(self.factory_names),
|
|
|
|
|
method="add_pipe",
|
|
|
|
|
lang=util.get_object_name(self),
|
|
|
|
|
lang_code=self.lang,
|
|
|
|
|
)
|
|
|
|
|
pipe_component = self.create_pipe(
|
2020-08-13 18:38:30 +03:00
|
|
|
|
factory_name,
|
|
|
|
|
name=name,
|
|
|
|
|
config=config,
|
|
|
|
|
raw_config=raw_config,
|
|
|
|
|
validate=validate,
|
2020-08-05 00:39:19 +03:00
|
|
|
|
)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
pipe_index = self._get_pipe_index(before, after, first, last)
|
|
|
|
|
self._pipe_meta[name] = self.get_factory_meta(factory_name)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self._components.insert(pipe_index, (name, pipe_component))
|
2020-07-22 14:42:59 +03:00
|
|
|
|
return pipe_component
|
2017-06-04 23:52:09 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def _get_pipe_index(
|
|
|
|
|
self,
|
|
|
|
|
before: Optional[Union[str, int]] = None,
|
|
|
|
|
after: Optional[Union[str, int]] = None,
|
|
|
|
|
first: Optional[bool] = None,
|
|
|
|
|
last: Optional[bool] = None,
|
|
|
|
|
) -> int:
|
|
|
|
|
"""Determine where to insert a pipeline component based on the before/
|
|
|
|
|
after/first/last values.
|
|
|
|
|
|
|
|
|
|
before (str): Name or index of the component to insert directly before.
|
|
|
|
|
after (str): Name or index of component to insert directly after.
|
|
|
|
|
first (bool): If True, insert component first in the pipeline.
|
|
|
|
|
last (bool): If True, insert component last in the pipeline.
|
|
|
|
|
RETURNS (int): The index of the new pipeline component.
|
|
|
|
|
"""
|
|
|
|
|
all_args = {"before": before, "after": after, "first": first, "last": last}
|
|
|
|
|
if sum(arg is not None for arg in [before, after, first, last]) >= 2:
|
2020-08-28 22:04:02 +03:00
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E006.format(args=all_args, opts=self.component_names)
|
|
|
|
|
)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
if last or not any(value is not None for value in [first, before, after]):
|
2020-08-29 16:20:11 +03:00
|
|
|
|
return len(self._components)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
elif first:
|
|
|
|
|
return 0
|
|
|
|
|
elif isinstance(before, str):
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if before not in self.component_names:
|
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E001.format(name=before, opts=self.component_names)
|
|
|
|
|
)
|
|
|
|
|
return self.component_names.index(before)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
elif isinstance(after, str):
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if after not in self.component_names:
|
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E001.format(name=after, opts=self.component_names)
|
|
|
|
|
)
|
|
|
|
|
return self.component_names.index(after) + 1
|
2020-07-22 14:42:59 +03:00
|
|
|
|
# We're only accepting indices referring to components that exist
|
|
|
|
|
# (can't just do isinstance here because bools are instance of int, too)
|
|
|
|
|
elif type(before) == int:
|
2020-08-29 16:20:11 +03:00
|
|
|
|
if before >= len(self._components) or before < 0:
|
2020-08-28 16:20:14 +03:00
|
|
|
|
err = Errors.E959.format(
|
2020-08-28 22:04:02 +03:00
|
|
|
|
dir="before", idx=before, opts=self.component_names
|
2020-08-28 16:20:14 +03:00
|
|
|
|
)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
raise ValueError(err)
|
|
|
|
|
return before
|
|
|
|
|
elif type(after) == int:
|
2020-08-29 16:20:11 +03:00
|
|
|
|
if after >= len(self._components) or after < 0:
|
2020-08-28 22:04:02 +03:00
|
|
|
|
err = Errors.E959.format(
|
|
|
|
|
dir="after", idx=after, opts=self.component_names
|
|
|
|
|
)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
raise ValueError(err)
|
|
|
|
|
return after + 1
|
2020-08-28 22:04:02 +03:00
|
|
|
|
raise ValueError(Errors.E006.format(args=all_args, opts=self.component_names))
|
2020-07-22 14:42:59 +03:00
|
|
|
|
|
|
|
|
|
def has_pipe(self, name: str) -> bool:
|
2017-10-17 12:20:07 +03:00
|
|
|
|
"""Check if a component name is present in the pipeline. Equivalent to
|
|
|
|
|
`name in nlp.pipe_names`.
|
|
|
|
|
|
2020-05-24 18:20:58 +03:00
|
|
|
|
name (str): Name of the component.
|
2017-10-27 15:40:14 +03:00
|
|
|
|
RETURNS (bool): Whether a component of the name exists in the pipeline.
|
2019-03-15 18:23:17 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#has_pipe
|
2017-10-17 12:20:07 +03:00
|
|
|
|
"""
|
|
|
|
|
return name in self.pipe_names
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def replace_pipe(
|
|
|
|
|
self,
|
|
|
|
|
name: str,
|
|
|
|
|
factory_name: str,
|
2020-07-29 00:12:42 +03:00
|
|
|
|
*,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
config: Dict[str, Any] = SimpleFrozenDict(),
|
|
|
|
|
validate: bool = True,
|
2020-10-08 11:34:01 +03:00
|
|
|
|
) -> Callable[[Doc], Doc]:
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""Replace a component in the pipeline.
|
|
|
|
|
|
2020-05-24 18:20:58 +03:00
|
|
|
|
name (str): Name of the component to replace.
|
2020-07-22 14:42:59 +03:00
|
|
|
|
factory_name (str): Factory name of replacement component.
|
|
|
|
|
config (Optional[Dict[str, Any]]): Config parameters to use for this
|
|
|
|
|
component. Will be merged with default config, if available.
|
|
|
|
|
validate (bool): Whether to validate the component config against the
|
|
|
|
|
arguments and types expected by the factory.
|
2020-10-08 11:34:01 +03:00
|
|
|
|
RETURNS (Callable[[Doc], Doc]): The new pipeline component.
|
2019-03-15 18:23:17 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#replace_pipe
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""
|
|
|
|
|
if name not in self.pipe_names:
|
2018-04-03 16:50:31 +03:00
|
|
|
|
raise ValueError(Errors.E001.format(name=name, opts=self.pipe_names))
|
2020-07-22 14:42:59 +03:00
|
|
|
|
if hasattr(factory_name, "__call__"):
|
|
|
|
|
err = Errors.E968.format(component=repr(factory_name), name=name)
|
|
|
|
|
raise ValueError(err)
|
|
|
|
|
# We need to delegate to Language.add_pipe here instead of just writing
|
|
|
|
|
# to Language.pipeline to make sure the configs are handled correctly
|
|
|
|
|
pipe_index = self.pipe_names.index(name)
|
|
|
|
|
self.remove_pipe(name)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
if not len(self._components) or pipe_index == len(self._components):
|
2020-08-05 10:30:58 +03:00
|
|
|
|
# we have no components to insert before/after, or we're replacing the last component
|
2020-10-08 11:34:01 +03:00
|
|
|
|
return self.add_pipe(
|
|
|
|
|
factory_name, name=name, config=config, validate=validate
|
|
|
|
|
)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
else:
|
2020-10-08 11:34:01 +03:00
|
|
|
|
return self.add_pipe(
|
2020-08-23 22:15:12 +03:00
|
|
|
|
factory_name,
|
|
|
|
|
name=name,
|
|
|
|
|
before=pipe_index,
|
|
|
|
|
config=config,
|
|
|
|
|
validate=validate,
|
|
|
|
|
)
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def rename_pipe(self, old_name: str, new_name: str) -> None:
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""Rename a pipeline component.
|
|
|
|
|
|
2020-05-24 18:20:58 +03:00
|
|
|
|
old_name (str): Name of the component to rename.
|
|
|
|
|
new_name (str): New name of the component.
|
2019-03-15 18:23:17 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#rename_pipe
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if old_name not in self.component_names:
|
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E001.format(name=old_name, opts=self.component_names)
|
|
|
|
|
)
|
|
|
|
|
if new_name in self.component_names:
|
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E007.format(name=new_name, opts=self.component_names)
|
|
|
|
|
)
|
|
|
|
|
i = self.component_names.index(old_name)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self._components[i] = (new_name, self._components[i][1])
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self._pipe_meta[new_name] = self._pipe_meta.pop(old_name)
|
|
|
|
|
self._pipe_configs[new_name] = self._pipe_configs.pop(old_name)
|
2020-10-04 15:43:45 +03:00
|
|
|
|
# Make sure [initialize] config is adjusted
|
|
|
|
|
if old_name in self._config["initialize"]["components"]:
|
|
|
|
|
init_cfg = self._config["initialize"]["components"].pop(old_name)
|
|
|
|
|
self._config["initialize"]["components"][new_name] = init_cfg
|
2017-10-07 01:25:54 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def remove_pipe(self, name: str) -> Tuple[str, Callable[[Doc], Doc]]:
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""Remove a component from the pipeline.
|
|
|
|
|
|
2020-05-24 18:20:58 +03:00
|
|
|
|
name (str): Name of the component to remove.
|
2017-10-07 02:04:50 +03:00
|
|
|
|
RETURNS (tuple): A `(name, component)` tuple of the removed component.
|
2019-03-15 18:23:17 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#remove_pipe
|
2017-10-07 01:25:54 +03:00
|
|
|
|
"""
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if name not in self.component_names:
|
|
|
|
|
raise ValueError(Errors.E001.format(name=name, opts=self.component_names))
|
2020-08-29 16:20:11 +03:00
|
|
|
|
removed = self._components.pop(self.component_names.index(name))
|
2020-07-22 14:42:59 +03:00
|
|
|
|
# We're only removing the component itself from the metas/configs here
|
|
|
|
|
# because factory may be used for something else
|
|
|
|
|
self._pipe_meta.pop(name)
|
|
|
|
|
self._pipe_configs.pop(name)
|
2020-10-04 15:43:45 +03:00
|
|
|
|
# Make sure name is removed from the [initialize] config
|
|
|
|
|
if name in self._config["initialize"]["components"]:
|
|
|
|
|
self._config["initialize"]["components"].pop(name)
|
2020-08-28 16:20:14 +03:00
|
|
|
|
# Make sure the name is also removed from the set of disabled components
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if name in self.disabled:
|
2020-08-29 13:08:33 +03:00
|
|
|
|
self._disabled.remove(name)
|
2019-10-30 21:04:17 +03:00
|
|
|
|
return removed
|
2017-06-04 23:52:09 +03:00
|
|
|
|
|
2020-08-28 16:20:14 +03:00
|
|
|
|
def disable_pipe(self, name: str) -> None:
|
|
|
|
|
"""Disable a pipeline component. The component will still exist on
|
2020-08-28 21:34:46 +03:00
|
|
|
|
the nlp object, but it won't be run as part of the pipeline. Does
|
|
|
|
|
nothing if the component is already disabled.
|
2020-08-28 16:20:14 +03:00
|
|
|
|
|
|
|
|
|
name (str): The name of the component to disable.
|
|
|
|
|
"""
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if name not in self.component_names:
|
|
|
|
|
raise ValueError(Errors.E001.format(name=name, opts=self.component_names))
|
2020-08-29 13:08:33 +03:00
|
|
|
|
self._disabled.add(name)
|
2020-08-28 16:20:14 +03:00
|
|
|
|
|
|
|
|
|
def enable_pipe(self, name: str) -> None:
|
|
|
|
|
"""Enable a previously disabled pipeline component so it's run as part
|
2020-08-28 21:34:46 +03:00
|
|
|
|
of the pipeline. Does nothing if the component is already enabled.
|
2020-08-28 16:20:14 +03:00
|
|
|
|
|
|
|
|
|
name (str): The name of the component to enable.
|
|
|
|
|
"""
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if name not in self.component_names:
|
|
|
|
|
raise ValueError(Errors.E001.format(name=name, opts=self.component_names))
|
|
|
|
|
if name in self.disabled:
|
2020-08-29 13:08:33 +03:00
|
|
|
|
self._disabled.remove(name)
|
2020-08-28 16:20:14 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def __call__(
|
|
|
|
|
self,
|
|
|
|
|
text: str,
|
2020-07-29 00:12:42 +03:00
|
|
|
|
*,
|
2020-08-29 16:20:11 +03:00
|
|
|
|
disable: Iterable[str] = SimpleFrozenList(),
|
2020-07-22 14:42:59 +03:00
|
|
|
|
component_cfg: Optional[Dict[str, Dict[str, Any]]] = None,
|
|
|
|
|
) -> Doc:
|
2017-10-07 01:26:05 +03:00
|
|
|
|
"""Apply the pipeline to some text. The text can span multiple sentences,
|
2020-05-21 00:06:39 +03:00
|
|
|
|
and can contain arbitrary whitespace. Alignment into the original string
|
2015-08-25 16:37:17 +03:00
|
|
|
|
is preserved.
|
2016-12-18 18:54:52 +03:00
|
|
|
|
|
2020-05-24 18:20:58 +03:00
|
|
|
|
text (str): The text to be processed.
|
2017-05-26 13:33:54 +03:00
|
|
|
|
disable (list): Names of the pipeline components to disable.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
component_cfg (Dict[str, dict]): An optional dictionary with extra
|
|
|
|
|
keyword arguments for specific components.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
RETURNS (Doc): A container for accessing the annotations.
|
2016-11-01 14:25:36 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#call
|
2015-08-25 16:37:17 +03:00
|
|
|
|
"""
|
2016-10-14 18:38:29 +03:00
|
|
|
|
doc = self.make_doc(text)
|
2019-03-11 01:36:47 +03:00
|
|
|
|
if component_cfg is None:
|
|
|
|
|
component_cfg = {}
|
2017-10-07 01:25:54 +03:00
|
|
|
|
for name, proc in self.pipeline:
|
2017-05-26 13:33:54 +03:00
|
|
|
|
if name in disable:
|
2017-05-16 12:21:59 +03:00
|
|
|
|
continue
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
if not hasattr(proc, "__call__"):
|
2018-04-03 16:50:31 +03:00
|
|
|
|
raise ValueError(Errors.E003.format(component=type(proc), name=name))
|
2020-02-27 20:42:27 +03:00
|
|
|
|
try:
|
|
|
|
|
doc = proc(doc, **component_cfg.get(name, {}))
|
2020-10-01 10:21:00 +03:00
|
|
|
|
except KeyError as e:
|
2020-10-03 12:43:56 +03:00
|
|
|
|
# This typically happens if a component is not initialized
|
|
|
|
|
raise ValueError(Errors.E109.format(name=name)) from e
|
2018-04-03 16:50:31 +03:00
|
|
|
|
if doc is None:
|
|
|
|
|
raise ValueError(Errors.E005.format(name=name))
|
2016-05-17 17:55:42 +03:00
|
|
|
|
return doc
|
2015-08-25 16:37:17 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def disable_pipes(self, *names) -> "DisabledPipes":
|
2017-10-27 15:40:14 +03:00
|
|
|
|
"""Disable one or more pipeline components. If used as a context
|
|
|
|
|
manager, the pipeline will be restored to the initial state at the end
|
|
|
|
|
of the block. Otherwise, a DisabledPipes object is returned, that has
|
|
|
|
|
a `.restore()` method you can use to undo your changes.
|
2017-10-25 14:46:41 +03:00
|
|
|
|
|
2020-05-18 23:27:10 +03:00
|
|
|
|
This method has been deprecated since 3.0
|
2017-10-27 15:40:14 +03:00
|
|
|
|
"""
|
2020-05-18 23:27:10 +03:00
|
|
|
|
warnings.warn(Warnings.W096, DeprecationWarning)
|
2019-10-25 17:19:08 +03:00
|
|
|
|
if len(names) == 1 and isinstance(names[0], (list, tuple)):
|
|
|
|
|
names = names[0] # support list of names instead of spread
|
2020-08-29 13:08:46 +03:00
|
|
|
|
return self.select_pipes(disable=names)
|
2020-05-18 23:27:10 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def select_pipes(
|
|
|
|
|
self,
|
2020-07-29 00:12:42 +03:00
|
|
|
|
*,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
disable: Optional[Union[str, Iterable[str]]] = None,
|
|
|
|
|
enable: Optional[Union[str, Iterable[str]]] = None,
|
|
|
|
|
) -> "DisabledPipes":
|
2020-05-18 23:27:10 +03:00
|
|
|
|
"""Disable one or more pipeline components. If used as a context
|
|
|
|
|
manager, the pipeline will be restored to the initial state at the end
|
|
|
|
|
of the block. Otherwise, a DisabledPipes object is returned, that has
|
|
|
|
|
a `.restore()` method you can use to undo your changes.
|
|
|
|
|
|
|
|
|
|
disable (str or iterable): The name(s) of the pipes to disable
|
|
|
|
|
enable (str or iterable): The name(s) of the pipes to enable - all others will be disabled
|
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#select_pipes
|
2020-05-18 23:27:10 +03:00
|
|
|
|
"""
|
|
|
|
|
if enable is None and disable is None:
|
|
|
|
|
raise ValueError(Errors.E991)
|
|
|
|
|
if disable is not None and isinstance(disable, str):
|
|
|
|
|
disable = [disable]
|
|
|
|
|
if enable is not None:
|
|
|
|
|
if isinstance(enable, str):
|
|
|
|
|
enable = [enable]
|
|
|
|
|
to_disable = [pipe for pipe in self.pipe_names if pipe not in enable]
|
|
|
|
|
# raise an error if the enable and disable keywords are not consistent
|
|
|
|
|
if disable is not None and disable != to_disable:
|
2020-05-19 17:20:03 +03:00
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E992.format(
|
|
|
|
|
enable=enable, disable=disable, names=self.pipe_names
|
|
|
|
|
)
|
|
|
|
|
)
|
2020-05-18 23:27:10 +03:00
|
|
|
|
disable = to_disable
|
2020-10-09 13:06:20 +03:00
|
|
|
|
# DisabledPipes will restore the pipes in 'disable' when it's done, so we need to exclude
|
|
|
|
|
# those pipes that were already disabled.
|
|
|
|
|
disable = [d for d in disable if d not in self._disabled]
|
2020-05-18 23:27:10 +03:00
|
|
|
|
return DisabledPipes(self, disable)
|
2017-10-25 14:46:41 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def make_doc(self, text: str) -> Doc:
|
|
|
|
|
"""Turn a text into a Doc object.
|
|
|
|
|
|
|
|
|
|
text (str): The text to process.
|
|
|
|
|
RETURNS (Doc): The processed doc.
|
|
|
|
|
"""
|
2020-12-08 09:24:02 +03:00
|
|
|
|
if len(text) > self.max_length:
|
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E088.format(length=len(text), max_length=self.max_length)
|
|
|
|
|
)
|
2017-05-29 16:40:45 +03:00
|
|
|
|
return self.tokenizer(text)
|
|
|
|
|
|
2020-05-21 19:39:06 +03:00
|
|
|
|
def update(
|
|
|
|
|
self,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
examples: Iterable[Example],
|
2020-07-29 00:12:42 +03:00
|
|
|
|
_: Optional[Any] = None,
|
2020-05-21 19:39:06 +03:00
|
|
|
|
*,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
drop: float = 0.0,
|
|
|
|
|
sgd: Optional[Optimizer] = None,
|
|
|
|
|
losses: Optional[Dict[str, float]] = None,
|
|
|
|
|
component_cfg: Optional[Dict[str, Dict[str, Any]]] = None,
|
2020-08-29 16:20:11 +03:00
|
|
|
|
exclude: Iterable[str] = SimpleFrozenList(),
|
2020-05-21 19:39:06 +03:00
|
|
|
|
):
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""Update the models in the pipeline.
|
|
|
|
|
|
2020-07-09 20:43:39 +03:00
|
|
|
|
examples (Iterable[Example]): A batch of examples
|
2020-07-29 00:12:42 +03:00
|
|
|
|
_: Should not be set - serves to catch backwards-incompatible scripts.
|
2019-10-14 13:28:53 +03:00
|
|
|
|
drop (float): The dropout rate.
|
2020-07-09 20:43:39 +03:00
|
|
|
|
sgd (Optimizer): An optimizer.
|
|
|
|
|
losses (Dict[str, float]): Dictionary to update with the loss, keyed by component.
|
|
|
|
|
component_cfg (Dict[str, Dict]): Config parameters for specific pipeline
|
2019-05-24 15:06:26 +03:00
|
|
|
|
components, keyed by component name.
|
2020-08-05 00:39:19 +03:00
|
|
|
|
exclude (Iterable[str]): Names of components that shouldn't be updated.
|
2020-07-09 20:43:39 +03:00
|
|
|
|
RETURNS (Dict[str, float]): The updated losses dictionary
|
2017-05-19 00:57:38 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#update
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""
|
2020-07-29 00:12:42 +03:00
|
|
|
|
if _ is not None:
|
2020-05-20 12:41:12 +03:00
|
|
|
|
raise ValueError(Errors.E989)
|
2020-07-09 20:43:39 +03:00
|
|
|
|
if losses is None:
|
|
|
|
|
losses = {}
|
2019-11-11 19:35:27 +03:00
|
|
|
|
if len(examples) == 0:
|
2020-07-09 20:43:39 +03:00
|
|
|
|
return losses
|
2020-08-12 00:29:31 +03:00
|
|
|
|
validate_examples(examples, "Language.update")
|
2021-01-19 18:47:44 +03:00
|
|
|
|
examples = _copy_examples(examples)
|
2017-08-20 15:42:07 +03:00
|
|
|
|
if sgd is None:
|
|
|
|
|
if self._optimizer is None:
|
2020-09-29 12:42:19 +03:00
|
|
|
|
self._optimizer = self.create_optimizer()
|
2017-08-20 15:42:07 +03:00
|
|
|
|
sgd = self._optimizer
|
2019-03-11 01:36:47 +03:00
|
|
|
|
if component_cfg is None:
|
|
|
|
|
component_cfg = {}
|
2020-05-22 16:55:45 +03:00
|
|
|
|
for i, (name, proc) in enumerate(self.pipeline):
|
2020-01-29 19:06:46 +03:00
|
|
|
|
component_cfg.setdefault(name, {})
|
|
|
|
|
component_cfg[name].setdefault("drop", drop)
|
|
|
|
|
for name, proc in self.pipeline:
|
2020-08-05 00:39:19 +03:00
|
|
|
|
if name in exclude or not hasattr(proc, "update"):
|
2017-05-22 02:43:31 +03:00
|
|
|
|
continue
|
2020-01-29 19:06:46 +03:00
|
|
|
|
proc.update(examples, sgd=None, losses=losses, **component_cfg[name])
|
2020-07-08 22:36:51 +03:00
|
|
|
|
if sgd not in (None, False):
|
2020-01-29 19:06:46 +03:00
|
|
|
|
for name, proc in self.pipeline:
|
2020-08-12 00:29:31 +03:00
|
|
|
|
if (
|
|
|
|
|
name not in exclude
|
2020-10-05 18:43:42 +03:00
|
|
|
|
and hasattr(proc, "is_trainable")
|
2020-10-08 22:33:49 +03:00
|
|
|
|
and proc.is_trainable
|
2020-08-12 00:29:31 +03:00
|
|
|
|
and proc.model not in (True, False, None)
|
|
|
|
|
):
|
2020-10-05 17:23:33 +03:00
|
|
|
|
proc.finish_update(sgd)
|
2020-07-09 20:43:39 +03:00
|
|
|
|
return losses
|
2017-05-16 17:17:30 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def rehearse(
|
|
|
|
|
self,
|
|
|
|
|
examples: Iterable[Example],
|
2020-07-29 00:12:42 +03:00
|
|
|
|
*,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
sgd: Optional[Optimizer] = None,
|
|
|
|
|
losses: Optional[Dict[str, float]] = None,
|
|
|
|
|
component_cfg: Optional[Dict[str, Dict[str, Any]]] = None,
|
2020-08-29 16:20:11 +03:00
|
|
|
|
exclude: Iterable[str] = SimpleFrozenList(),
|
2020-07-22 14:42:59 +03:00
|
|
|
|
) -> Dict[str, float]:
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
"""Make a "rehearsal" update to the models in the pipeline, to prevent
|
|
|
|
|
forgetting. Rehearsal updates run an initial copy of the model over some
|
|
|
|
|
data, and update the model so its current predictions are more like the
|
2019-10-02 11:37:39 +03:00
|
|
|
|
initial ones. This is useful for keeping a pretrained model on-track,
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
even if you're updating it with a smaller set of examples.
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
examples (Iterable[Example]): A batch of `Example` objects.
|
|
|
|
|
sgd (Optional[Optimizer]): An optimizer.
|
|
|
|
|
component_cfg (Dict[str, Dict]): Config parameters for specific pipeline
|
|
|
|
|
components, keyed by component name.
|
2020-08-05 00:39:19 +03:00
|
|
|
|
exclude (Iterable[str]): Names of components that shouldn't be updated.
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
RETURNS (dict): Results from the update.
|
|
|
|
|
|
|
|
|
|
EXAMPLE:
|
|
|
|
|
>>> raw_text_batches = minibatch(raw_texts)
|
2020-07-06 14:02:36 +03:00
|
|
|
|
>>> for labelled_batch in minibatch(examples):
|
2019-11-11 19:35:27 +03:00
|
|
|
|
>>> nlp.update(labelled_batch)
|
2020-07-06 14:02:36 +03:00
|
|
|
|
>>> raw_batch = [Example.from_dict(nlp.make_doc(text), {}) for text in next(raw_text_batches)]
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
>>> nlp.rehearse(raw_batch)
|
2020-07-29 00:12:42 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#rehearse
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
"""
|
2019-11-11 19:35:27 +03:00
|
|
|
|
if len(examples) == 0:
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
return
|
2020-08-12 00:29:31 +03:00
|
|
|
|
validate_examples(examples, "Language.rehearse")
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
if sgd is None:
|
|
|
|
|
if self._optimizer is None:
|
2020-09-29 12:42:19 +03:00
|
|
|
|
self._optimizer = self.create_optimizer()
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
sgd = self._optimizer
|
|
|
|
|
pipes = list(self.pipeline)
|
|
|
|
|
random.shuffle(pipes)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
if component_cfg is None:
|
|
|
|
|
component_cfg = {}
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
grads = {}
|
|
|
|
|
|
|
|
|
|
def get_grads(W, dW, key=None):
|
|
|
|
|
grads[key] = (W, dW)
|
|
|
|
|
|
2020-01-29 19:06:46 +03:00
|
|
|
|
get_grads.learn_rate = sgd.learn_rate
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
get_grads.b1 = sgd.b1
|
|
|
|
|
get_grads.b2 = sgd.b2
|
|
|
|
|
for name, proc in pipes:
|
2020-08-05 00:39:19 +03:00
|
|
|
|
if name in exclude or not hasattr(proc, "rehearse"):
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
continue
|
|
|
|
|
grads = {}
|
2020-02-03 15:02:12 +03:00
|
|
|
|
proc.rehearse(
|
2020-07-22 14:42:59 +03:00
|
|
|
|
examples, sgd=get_grads, losses=losses, **component_cfg.get(name, {})
|
2020-02-03 15:02:12 +03:00
|
|
|
|
)
|
2020-01-29 19:06:46 +03:00
|
|
|
|
for key, (W, dW) in grads.items():
|
|
|
|
|
sgd(W, dW, key=key)
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
return losses
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def begin_training(
|
|
|
|
|
self,
|
2020-07-29 00:12:42 +03:00
|
|
|
|
get_examples: Optional[Callable[[], Iterable[Example]]] = None,
|
|
|
|
|
*,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
sgd: Optional[Optimizer] = None,
|
2020-09-28 22:35:09 +03:00
|
|
|
|
) -> Optimizer:
|
|
|
|
|
warnings.warn(Warnings.W089, DeprecationWarning)
|
2020-09-29 13:14:08 +03:00
|
|
|
|
return self.initialize(get_examples, sgd=sgd)
|
2020-09-28 22:35:09 +03:00
|
|
|
|
|
|
|
|
|
def initialize(
|
|
|
|
|
self,
|
|
|
|
|
get_examples: Optional[Callable[[], Iterable[Example]]] = None,
|
|
|
|
|
*,
|
|
|
|
|
sgd: Optional[Optimizer] = None,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
) -> Optimizer:
|
2020-07-29 00:12:42 +03:00
|
|
|
|
"""Initialize the pipe for training, using data examples if available.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
|
2020-07-29 00:12:42 +03:00
|
|
|
|
get_examples (Callable[[], Iterable[Example]]): Optional function that
|
|
|
|
|
returns gold-standard Example objects.
|
2020-09-29 13:14:08 +03:00
|
|
|
|
sgd (Optional[Optimizer]): An optimizer to use for updates. If not
|
2020-09-29 12:42:19 +03:00
|
|
|
|
provided, will be created using the .create_optimizer() method.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
RETURNS (thinc.api.Optimizer): The optimizer.
|
2019-03-15 18:23:17 +03:00
|
|
|
|
|
2020-09-28 22:35:09 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#initialize
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""
|
2019-11-11 19:35:27 +03:00
|
|
|
|
if get_examples is None:
|
2020-09-08 23:44:25 +03:00
|
|
|
|
util.logger.debug(
|
2020-09-28 22:35:09 +03:00
|
|
|
|
"No 'get_examples' callback provided to 'Language.initialize', creating dummy examples"
|
2020-09-08 23:44:25 +03:00
|
|
|
|
)
|
|
|
|
|
doc = Doc(self.vocab, words=["x", "y", "z"])
|
|
|
|
|
get_examples = lambda: [Example.from_dict(doc, {})]
|
|
|
|
|
if not hasattr(get_examples, "__call__"):
|
2020-10-10 20:14:48 +03:00
|
|
|
|
err = Errors.E930.format(
|
|
|
|
|
method="Language.initialize", obj=type(get_examples)
|
|
|
|
|
)
|
2020-10-08 22:33:49 +03:00
|
|
|
|
raise TypeError(err)
|
2020-09-29 17:05:48 +03:00
|
|
|
|
# Make sure the config is interpolated so we can resolve subsections
|
|
|
|
|
config = self.config.interpolate()
|
|
|
|
|
# These are the settings provided in the [initialize] block in the config
|
|
|
|
|
I = registry.resolve(config["initialize"], schema=ConfigSchemaInit)
|
2021-01-12 13:29:31 +03:00
|
|
|
|
before_init = I["before_init"]
|
|
|
|
|
if before_init is not None:
|
|
|
|
|
before_init(self)
|
2020-09-29 17:05:48 +03:00
|
|
|
|
init_vocab(
|
2020-09-29 22:39:28 +03:00
|
|
|
|
self, data=I["vocab_data"], lookups=I["lookups"], vectors=I["vectors"]
|
2020-09-29 17:05:48 +03:00
|
|
|
|
)
|
|
|
|
|
pretrain_cfg = config.get("pretraining")
|
|
|
|
|
if pretrain_cfg:
|
|
|
|
|
P = registry.resolve(pretrain_cfg, schema=ConfigSchemaPretrain)
|
2020-09-29 17:47:55 +03:00
|
|
|
|
init_tok2vec(self, P, I)
|
2020-09-29 12:42:19 +03:00
|
|
|
|
if self.vocab.vectors.data.shape[1] >= 1:
|
|
|
|
|
ops = get_current_ops()
|
|
|
|
|
self.vocab.vectors.data = ops.asarray(self.vocab.vectors.data)
|
2020-09-29 12:52:45 +03:00
|
|
|
|
if hasattr(self.tokenizer, "initialize"):
|
|
|
|
|
tok_settings = validate_init_settings(
|
|
|
|
|
self.tokenizer.initialize,
|
2020-09-29 17:05:48 +03:00
|
|
|
|
I["tokenizer"],
|
2020-09-29 12:52:45 +03:00
|
|
|
|
section="tokenizer",
|
|
|
|
|
name="tokenizer",
|
|
|
|
|
)
|
|
|
|
|
self.tokenizer.initialize(get_examples, nlp=self, **tok_settings)
|
2017-10-07 01:25:54 +03:00
|
|
|
|
for name, proc in self.pipeline:
|
2020-09-28 22:35:09 +03:00
|
|
|
|
if hasattr(proc, "initialize"):
|
2020-09-29 17:05:48 +03:00
|
|
|
|
p_settings = I["components"].get(name, {})
|
2020-09-29 12:52:45 +03:00
|
|
|
|
p_settings = validate_init_settings(
|
|
|
|
|
proc.initialize, p_settings, section="components", name=name
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
)
|
2020-09-29 13:21:52 +03:00
|
|
|
|
proc.initialize(get_examples, nlp=self, **p_settings)
|
2020-01-29 19:06:46 +03:00
|
|
|
|
self._link_components()
|
2020-09-29 17:05:48 +03:00
|
|
|
|
self._optimizer = sgd
|
2020-09-29 12:42:19 +03:00
|
|
|
|
if sgd is not None:
|
|
|
|
|
self._optimizer = sgd
|
|
|
|
|
elif self._optimizer is None:
|
|
|
|
|
self._optimizer = self.create_optimizer()
|
2021-01-12 13:29:31 +03:00
|
|
|
|
after_init = I["after_init"]
|
|
|
|
|
if after_init is not None:
|
|
|
|
|
after_init(self)
|
2017-08-20 15:42:07 +03:00
|
|
|
|
return self._optimizer
|
2017-05-21 17:07:06 +03:00
|
|
|
|
|
2020-09-29 12:42:19 +03:00
|
|
|
|
def resume_training(self, *, sgd: Optional[Optimizer] = None) -> Optimizer:
|
2019-10-02 11:37:39 +03:00
|
|
|
|
"""Continue training a pretrained model.
|
2018-12-18 15:48:10 +03:00
|
|
|
|
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
Create and return an optimizer, and initialize "rehearsal" for any pipeline
|
|
|
|
|
component that has a .rehearse() method. Rehearsal is used to prevent
|
2020-07-29 00:12:42 +03:00
|
|
|
|
models from "forgetting" their initialized "knowledge". To perform
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
rehearsal, collect samples of text you want the models to retain performance
|
2020-07-06 14:02:36 +03:00
|
|
|
|
on, and call nlp.rehearse() with a batch of Example objects.
|
2020-07-22 14:42:59 +03:00
|
|
|
|
|
|
|
|
|
RETURNS (Optimizer): The optimizer.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#resume_training
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
"""
|
2020-09-29 12:42:19 +03:00
|
|
|
|
ops = get_current_ops()
|
|
|
|
|
if self.vocab.vectors.data.shape[1] >= 1:
|
|
|
|
|
self.vocab.vectors.data = ops.asarray(self.vocab.vectors.data)
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
for name, proc in self.pipeline:
|
|
|
|
|
if hasattr(proc, "_rehearsal_model"):
|
|
|
|
|
proc._rehearsal_model = deepcopy(proc.model)
|
2020-09-29 12:42:19 +03:00
|
|
|
|
if sgd is not None:
|
|
|
|
|
self._optimizer = sgd
|
|
|
|
|
elif self._optimizer is None:
|
|
|
|
|
self._optimizer = self.create_optimizer()
|
💫 Better support for semi-supervised learning (#3035)
The new spacy pretrain command implemented BERT/ULMFit/etc-like transfer learning, using our Language Modelling with Approximate Outputs version of BERT's cloze task. Pretraining is convenient, but in some ways it's a bit of a strange solution. All we're doing is initialising the weights. At the same time, we're putting a lot of work into our optimisation so that it's less sensitive to initial conditions, and more likely to find good optima. I discuss this a bit in the pseudo-rehearsal blog post: https://explosion.ai/blog/pseudo-rehearsal-catastrophic-forgetting
Support semi-supervised learning in spacy train
One obvious way to improve these pretraining methods is to do multi-task learning, instead of just transfer learning. This has been shown to work very well: https://arxiv.org/pdf/1809.08370.pdf . This patch makes it easy to do this sort of thing.
Add a new argument to spacy train, --raw-text. This takes a jsonl file with unlabelled data that can be used in arbitrary ways to do semi-supervised learning.
Add a new method to the Language class and to pipeline components, .rehearse(). This is like .update(), but doesn't expect GoldParse objects. It takes a batch of Doc objects, and performs an update on some semi-supervised objective.
Move the BERT-LMAO objective out from spacy/cli/pretrain.py into spacy/_ml.py, so we can create a new pipeline component, ClozeMultitask. This can be specified as a parser or NER multitask in the spacy train command. Example usage:
python -m spacy train en ./tmp ~/data/en-core-web/train/nw.json ~/data/en-core-web/dev/nw.json --pipeline parser --raw-textt ~/data/unlabelled/reddit-100k.jsonl --vectors en_vectors_web_lg --parser-multitasks cloze
Implement rehearsal methods for pipeline components
The new --raw-text argument and nlp.rehearse() method also gives us a good place to implement the the idea in the pseudo-rehearsal blog post in the parser. This works as follows:
Add a new nlp.resume_training() method. This allocates copies of pre-trained models in the pipeline, setting things up for the rehearsal updates. It also returns an optimizer object. This also greatly reduces confusion around the nlp.begin_training() method, which randomises the weights, making it not suitable for adding new labels or otherwise fine-tuning a pre-trained model.
Implement rehearsal updates on the Parser class, making it available for the dependency parser and NER. During rehearsal, the initial model is used to supervise the model being trained. The current model is asked to match the predictions of the initial model on some data. This minimises catastrophic forgetting, by keeping the model's predictions close to the original. See the blog post for details.
Implement rehearsal updates for tagger
Implement rehearsal updates for text categoriz
2018-12-10 18:25:33 +03:00
|
|
|
|
return self._optimizer
|
|
|
|
|
|
2019-03-11 01:36:47 +03:00
|
|
|
|
def evaluate(
|
2020-07-22 14:42:59 +03:00
|
|
|
|
self,
|
|
|
|
|
examples: Iterable[Example],
|
2020-07-29 00:12:42 +03:00
|
|
|
|
*,
|
2020-12-09 11:13:26 +03:00
|
|
|
|
batch_size: Optional[int] = None,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
scorer: Optional[Scorer] = None,
|
|
|
|
|
component_cfg: Optional[Dict[str, Dict[str, Any]]] = None,
|
2020-07-31 12:02:17 +03:00
|
|
|
|
scorer_cfg: Optional[Dict[str, Any]] = None,
|
2020-07-29 00:12:42 +03:00
|
|
|
|
) -> Dict[str, Union[float, dict]]:
|
2019-05-24 15:06:36 +03:00
|
|
|
|
"""Evaluate a model's pipeline components.
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
examples (Iterable[Example]): `Example` objects.
|
2020-12-09 12:21:39 +03:00
|
|
|
|
batch_size (Optional[int]): Batch size to use.
|
2020-07-22 14:42:59 +03:00
|
|
|
|
scorer (Optional[Scorer]): Scorer to use. If not passed in, a new one
|
2019-05-24 15:06:36 +03:00
|
|
|
|
will be created.
|
|
|
|
|
component_cfg (dict): An optional dictionary with extra keyword
|
|
|
|
|
arguments for specific components.
|
2020-07-31 12:02:17 +03:00
|
|
|
|
scorer_cfg (dict): An optional dictionary with extra keyword arguments
|
|
|
|
|
for the scorer.
|
2019-05-24 15:06:36 +03:00
|
|
|
|
RETURNS (Scorer): The scorer containing the evaluation results.
|
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#evaluate
|
2019-05-24 15:06:36 +03:00
|
|
|
|
"""
|
2020-12-31 02:45:50 +03:00
|
|
|
|
examples = list(examples)
|
2020-08-12 00:29:31 +03:00
|
|
|
|
validate_examples(examples, "Language.evaluate")
|
2021-01-19 18:47:44 +03:00
|
|
|
|
examples = _copy_examples(examples)
|
2020-12-09 11:13:26 +03:00
|
|
|
|
if batch_size is None:
|
|
|
|
|
batch_size = self.batch_size
|
2019-03-15 17:20:09 +03:00
|
|
|
|
if component_cfg is None:
|
|
|
|
|
component_cfg = {}
|
2020-07-31 12:02:17 +03:00
|
|
|
|
if scorer_cfg is None:
|
|
|
|
|
scorer_cfg = {}
|
Refactor the Scorer to improve flexibility (#5731)
* Refactor the Scorer to improve flexibility
Refactor the `Scorer` to improve flexibility for arbitrary pipeline
components.
* Individual pipeline components provide their own `evaluate` methods
that score a list of `Example`s and return a dictionary of scores
* `Scorer` is initialized either:
* with a provided pipeline containing components to be scored
* with a default pipeline containing the built-in statistical
components (senter, tagger, morphologizer, parser, ner)
* `Scorer.score` evaluates a list of `Example`s and returns a dictionary
of scores referring to the scores provided by the components in the
pipeline
Significant differences:
* `tags_acc` is renamed to `tag_acc` to be consistent with `token_acc`
and the new `morph_acc`, `pos_acc`, and `lemma_acc`
* Scoring is no longer cumulative: `Scorer.score` scores a list of
examples rather than a single example and does not retain any state
about previously scored examples
* PRF values in the returned scores are no longer multiplied by 100
* Add kwargs to Morphologizer.evaluate
* Create generalized scoring methods in Scorer
* Generalized static scoring methods are added to `Scorer`
* Methods require an attribute (either on Token or Doc) that is
used to key the returned scores
Naming differences:
* `uas`, `las`, and `las_per_type` in the scores dict are renamed to
`dep_uas`, `dep_las`, and `dep_las_per_type`
Scoring differences:
* `Doc.sents` is now scored as spans rather than on sentence-initial
token positions so that `Doc.sents` and `Doc.ents` can be scored with
the same method (this lowers scores since a single incorrect sentence
start results in two incorrect spans)
* Simplify / extend hasattr check for eval method
* Add hasattr check to tokenizer scoring
* Simplify to hasattr check for component scoring
* Reset Example alignment if docs are set
Reset the Example alignment if either doc is set in case the
tokenization has changed.
* Add PRF tokenization scoring for tokens as spans
Add PRF scores for tokens as character spans. The scores are:
* token_acc: # correct tokens / # gold tokens
* token_p/r/f: PRF for (token.idx, token.idx + len(token))
* Add docstring to Scorer.score_tokenization
* Rename component.evaluate() to component.score()
* Update Scorer API docs
* Update scoring for positive_label in textcat
* Fix TextCategorizer.score kwargs
* Update Language.evaluate docs
* Update score names in default config
2020-07-25 13:53:02 +03:00
|
|
|
|
if scorer is None:
|
2020-07-31 12:02:17 +03:00
|
|
|
|
kwargs = dict(scorer_cfg)
|
Refactor the Scorer to improve flexibility (#5731)
* Refactor the Scorer to improve flexibility
Refactor the `Scorer` to improve flexibility for arbitrary pipeline
components.
* Individual pipeline components provide their own `evaluate` methods
that score a list of `Example`s and return a dictionary of scores
* `Scorer` is initialized either:
* with a provided pipeline containing components to be scored
* with a default pipeline containing the built-in statistical
components (senter, tagger, morphologizer, parser, ner)
* `Scorer.score` evaluates a list of `Example`s and returns a dictionary
of scores referring to the scores provided by the components in the
pipeline
Significant differences:
* `tags_acc` is renamed to `tag_acc` to be consistent with `token_acc`
and the new `morph_acc`, `pos_acc`, and `lemma_acc`
* Scoring is no longer cumulative: `Scorer.score` scores a list of
examples rather than a single example and does not retain any state
about previously scored examples
* PRF values in the returned scores are no longer multiplied by 100
* Add kwargs to Morphologizer.evaluate
* Create generalized scoring methods in Scorer
* Generalized static scoring methods are added to `Scorer`
* Methods require an attribute (either on Token or Doc) that is
used to key the returned scores
Naming differences:
* `uas`, `las`, and `las_per_type` in the scores dict are renamed to
`dep_uas`, `dep_las`, and `dep_las_per_type`
Scoring differences:
* `Doc.sents` is now scored as spans rather than on sentence-initial
token positions so that `Doc.sents` and `Doc.ents` can be scored with
the same method (this lowers scores since a single incorrect sentence
start results in two incorrect spans)
* Simplify / extend hasattr check for eval method
* Add hasattr check to tokenizer scoring
* Simplify to hasattr check for component scoring
* Reset Example alignment if docs are set
Reset the Example alignment if either doc is set in case the
tokenization has changed.
* Add PRF tokenization scoring for tokens as spans
Add PRF scores for tokens as character spans. The scores are:
* token_acc: # correct tokens / # gold tokens
* token_p/r/f: PRF for (token.idx, token.idx + len(token))
* Add docstring to Scorer.score_tokenization
* Rename component.evaluate() to component.score()
* Update Scorer API docs
* Update scoring for positive_label in textcat
* Fix TextCategorizer.score kwargs
* Update Language.evaluate docs
* Update score names in default config
2020-07-25 13:53:02 +03:00
|
|
|
|
kwargs.setdefault("nlp", self)
|
|
|
|
|
scorer = Scorer(**kwargs)
|
2020-12-31 02:45:50 +03:00
|
|
|
|
# reset annotation in predicted docs and time tokenization
|
2020-07-29 12:02:31 +03:00
|
|
|
|
start_time = timer()
|
2020-12-31 02:45:50 +03:00
|
|
|
|
# apply all pipeline components
|
2017-10-07 01:25:54 +03:00
|
|
|
|
for name, pipe in self.pipeline:
|
2019-03-11 01:36:47 +03:00
|
|
|
|
kwargs = component_cfg.get(name, {})
|
|
|
|
|
kwargs.setdefault("batch_size", batch_size)
|
2020-12-31 02:45:50 +03:00
|
|
|
|
for doc, eg in zip(
|
|
|
|
|
_pipe((eg.predicted for eg in examples), pipe, kwargs), examples
|
|
|
|
|
):
|
|
|
|
|
eg.predicted = doc
|
2020-07-29 12:02:31 +03:00
|
|
|
|
end_time = timer()
|
|
|
|
|
results = scorer.score(examples)
|
2020-12-31 02:45:50 +03:00
|
|
|
|
n_words = sum(len(eg.predicted) for eg in examples)
|
2020-07-29 12:02:31 +03:00
|
|
|
|
results["speed"] = n_words / (end_time - start_time)
|
|
|
|
|
return results
|
2017-05-16 12:21:59 +03:00
|
|
|
|
|
2020-09-29 12:42:19 +03:00
|
|
|
|
def create_optimizer(self):
|
|
|
|
|
"""Create an optimizer, usually using the [training.optimizer] config."""
|
2020-09-29 13:00:08 +03:00
|
|
|
|
subconfig = {"optimizer": self.config["training"]["optimizer"]}
|
|
|
|
|
return registry.resolve(subconfig)["optimizer"]
|
2017-05-16 12:21:59 +03:00
|
|
|
|
|
2017-05-18 12:25:19 +03:00
|
|
|
|
@contextmanager
|
2020-09-03 13:51:04 +03:00
|
|
|
|
def use_params(self, params: Optional[dict]):
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""Replace weights of models in the pipeline with those provided in the
|
|
|
|
|
params dictionary. Can be used as a contextmanager, in which case,
|
|
|
|
|
models go back to their original weights after the block.
|
|
|
|
|
|
|
|
|
|
params (dict): A dictionary of parameters keyed by model ID.
|
|
|
|
|
|
|
|
|
|
EXAMPLE:
|
|
|
|
|
>>> with nlp.use_params(optimizer.averages):
|
2020-07-29 00:12:42 +03:00
|
|
|
|
>>> nlp.to_disk("/tmp/checkpoint")
|
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#use_params
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""
|
2020-09-03 13:51:04 +03:00
|
|
|
|
if not params:
|
|
|
|
|
yield
|
|
|
|
|
else:
|
|
|
|
|
contexts = [
|
|
|
|
|
pipe.use_params(params)
|
|
|
|
|
for name, pipe in self.pipeline
|
|
|
|
|
if hasattr(pipe, "use_params") and hasattr(pipe, "model")
|
|
|
|
|
]
|
|
|
|
|
# TODO: Having trouble with contextlib
|
|
|
|
|
# Workaround: these aren't actually context managers atm.
|
|
|
|
|
for context in contexts:
|
|
|
|
|
try:
|
|
|
|
|
next(context)
|
|
|
|
|
except StopIteration:
|
|
|
|
|
pass
|
|
|
|
|
yield
|
|
|
|
|
for context in contexts:
|
|
|
|
|
try:
|
|
|
|
|
next(context)
|
|
|
|
|
except StopIteration:
|
|
|
|
|
pass
|
2017-05-18 12:25:19 +03:00
|
|
|
|
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
def pipe(
|
|
|
|
|
self,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
texts: Iterable[str],
|
2020-07-29 00:12:42 +03:00
|
|
|
|
*,
|
2020-07-22 14:42:59 +03:00
|
|
|
|
as_tuples: bool = False,
|
2020-12-09 11:13:26 +03:00
|
|
|
|
batch_size: Optional[int] = None,
|
2020-08-29 16:20:11 +03:00
|
|
|
|
disable: Iterable[str] = SimpleFrozenList(),
|
2020-07-22 14:42:59 +03:00
|
|
|
|
component_cfg: Optional[Dict[str, Dict[str, Any]]] = None,
|
|
|
|
|
n_process: int = 1,
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
):
|
2017-10-27 15:40:14 +03:00
|
|
|
|
"""Process texts as a stream, and yield `Doc` objects in order.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
|
2020-07-09 20:43:39 +03:00
|
|
|
|
texts (Iterable[str]): A sequence of texts to process.
|
2019-03-15 18:23:17 +03:00
|
|
|
|
as_tuples (bool): If set to True, inputs should be a sequence of
|
2017-08-19 13:21:33 +03:00
|
|
|
|
(text, context) tuples. Output will then be a sequence of
|
|
|
|
|
(doc, context) tuples. Defaults to False.
|
2020-12-09 11:13:26 +03:00
|
|
|
|
batch_size (Optional[int]): The number of texts to buffer.
|
2020-07-09 20:43:39 +03:00
|
|
|
|
disable (List[str]): Names of the pipeline components to disable.
|
|
|
|
|
component_cfg (Dict[str, Dict]): An optional dictionary with extra keyword
|
2019-03-15 18:23:17 +03:00
|
|
|
|
arguments for specific components.
|
2020-07-09 20:43:39 +03:00
|
|
|
|
n_process (int): Number of processors to process texts. If -1, set `multiprocessing.cpu_count()`.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
YIELDS (Doc): Documents in the order of the original text.
|
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#pipe
|
2017-04-15 12:59:21 +03:00
|
|
|
|
"""
|
2019-10-08 13:20:55 +03:00
|
|
|
|
if n_process == -1:
|
|
|
|
|
n_process = mp.cpu_count()
|
2017-08-19 13:21:33 +03:00
|
|
|
|
if as_tuples:
|
2017-07-25 19:57:59 +03:00
|
|
|
|
text_context1, text_context2 = itertools.tee(texts)
|
|
|
|
|
texts = (tc[0] for tc in text_context1)
|
|
|
|
|
contexts = (tc[1] for tc in text_context2)
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
docs = self.pipe(
|
2019-03-11 01:36:47 +03:00
|
|
|
|
texts,
|
|
|
|
|
batch_size=batch_size,
|
|
|
|
|
disable=disable,
|
2019-11-04 22:29:03 +03:00
|
|
|
|
n_process=n_process,
|
2019-03-11 01:36:47 +03:00
|
|
|
|
component_cfg=component_cfg,
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
)
|
2019-12-22 03:53:56 +03:00
|
|
|
|
for doc, context in zip(docs, contexts):
|
2017-07-25 19:57:59 +03:00
|
|
|
|
yield (doc, context)
|
|
|
|
|
return
|
2019-03-11 01:36:47 +03:00
|
|
|
|
if component_cfg is None:
|
|
|
|
|
component_cfg = {}
|
2020-12-09 11:13:26 +03:00
|
|
|
|
if batch_size is None:
|
|
|
|
|
batch_size = self.batch_size
|
2019-10-08 13:20:55 +03:00
|
|
|
|
|
|
|
|
|
pipes = (
|
|
|
|
|
[]
|
2020-01-06 16:57:34 +03:00
|
|
|
|
) # contains functools.partial objects to easily create multiprocess worker.
|
2017-10-07 01:25:54 +03:00
|
|
|
|
for name, proc in self.pipeline:
|
2017-05-26 13:33:54 +03:00
|
|
|
|
if name in disable:
|
2017-05-16 12:21:59 +03:00
|
|
|
|
continue
|
2019-03-11 01:36:47 +03:00
|
|
|
|
kwargs = component_cfg.get(name, {})
|
|
|
|
|
# Allow component_cfg to overwrite the top-level kwargs.
|
|
|
|
|
kwargs.setdefault("batch_size", batch_size)
|
2020-10-08 22:33:49 +03:00
|
|
|
|
f = functools.partial(_pipe, proc=proc, kwargs=kwargs)
|
2019-10-08 13:20:55 +03:00
|
|
|
|
pipes.append(f)
|
|
|
|
|
|
|
|
|
|
if n_process != 1:
|
|
|
|
|
docs = self._multiprocessing_pipe(texts, pipes, n_process, batch_size)
|
|
|
|
|
else:
|
|
|
|
|
# if n_process == 1, no processes are forked.
|
|
|
|
|
docs = (self.make_doc(text) for text in texts)
|
|
|
|
|
for pipe in pipes:
|
|
|
|
|
docs = pipe(docs)
|
2017-05-19 21:25:42 +03:00
|
|
|
|
for doc in docs:
|
2016-02-03 04:04:55 +03:00
|
|
|
|
yield doc
|
2016-02-01 11:01:13 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def _multiprocessing_pipe(
|
|
|
|
|
self,
|
|
|
|
|
texts: Iterable[str],
|
|
|
|
|
pipes: Iterable[Callable[[Doc], Doc]],
|
|
|
|
|
n_process: int,
|
|
|
|
|
batch_size: int,
|
|
|
|
|
) -> None:
|
2019-10-08 13:20:55 +03:00
|
|
|
|
# raw_texts is used later to stop iteration.
|
|
|
|
|
texts, raw_texts = itertools.tee(texts)
|
|
|
|
|
# for sending texts to worker
|
|
|
|
|
texts_q = [mp.Queue() for _ in range(n_process)]
|
2020-01-06 16:57:34 +03:00
|
|
|
|
# for receiving byte-encoded docs from worker
|
2019-10-08 13:20:55 +03:00
|
|
|
|
bytedocs_recv_ch, bytedocs_send_ch = zip(
|
|
|
|
|
*[mp.Pipe(False) for _ in range(n_process)]
|
|
|
|
|
)
|
|
|
|
|
|
2020-05-21 21:05:03 +03:00
|
|
|
|
batch_texts = util.minibatch(texts, batch_size)
|
2019-10-08 13:20:55 +03:00
|
|
|
|
# Sender sends texts to the workers.
|
|
|
|
|
# This is necessary to properly handle infinite length of texts.
|
|
|
|
|
# (In this case, all data cannot be sent to the workers at once)
|
|
|
|
|
sender = _Sender(batch_texts, texts_q, chunk_size=n_process)
|
2020-01-06 16:57:34 +03:00
|
|
|
|
# send twice to make process busy
|
2019-10-08 13:20:55 +03:00
|
|
|
|
sender.send()
|
|
|
|
|
sender.send()
|
|
|
|
|
|
|
|
|
|
procs = [
|
2020-02-12 13:50:42 +03:00
|
|
|
|
mp.Process(
|
|
|
|
|
target=_apply_pipes,
|
2020-03-26 15:38:14 +03:00
|
|
|
|
args=(self.make_doc, pipes, rch, sch, Underscore.get_state()),
|
2020-02-12 13:50:42 +03:00
|
|
|
|
)
|
2019-10-08 13:20:55 +03:00
|
|
|
|
for rch, sch in zip(texts_q, bytedocs_send_ch)
|
|
|
|
|
]
|
|
|
|
|
for proc in procs:
|
|
|
|
|
proc.start()
|
|
|
|
|
|
|
|
|
|
# Cycle channels not to break the order of docs.
|
2020-01-06 16:57:34 +03:00
|
|
|
|
# The received object is a batch of byte-encoded docs, so flatten them with chain.from_iterable.
|
2019-10-08 13:20:55 +03:00
|
|
|
|
byte_docs = chain.from_iterable(recv.recv() for recv in cycle(bytedocs_recv_ch))
|
|
|
|
|
docs = (Doc(self.vocab).from_bytes(byte_doc) for byte_doc in byte_docs)
|
|
|
|
|
try:
|
|
|
|
|
for i, (_, doc) in enumerate(zip(raw_texts, docs), 1):
|
|
|
|
|
yield doc
|
|
|
|
|
if i % batch_size == 0:
|
|
|
|
|
# tell `sender` that one batch was consumed.
|
|
|
|
|
sender.step()
|
|
|
|
|
finally:
|
|
|
|
|
for proc in procs:
|
|
|
|
|
proc.terminate()
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def _link_components(self) -> None:
|
2020-01-29 19:06:46 +03:00
|
|
|
|
"""Register 'listeners' within pipeline components, to allow them to
|
|
|
|
|
effectively share weights.
|
|
|
|
|
"""
|
2020-09-16 18:51:29 +03:00
|
|
|
|
# I had though, "Why do we do this inside the Language object? Shouldn't
|
|
|
|
|
# it be the tok2vec/transformer/etc's job?
|
|
|
|
|
# The problem is we need to do it during deserialization...And the
|
|
|
|
|
# components don't receive the pipeline then. So this does have to be
|
|
|
|
|
# here :(
|
2020-01-29 19:06:46 +03:00
|
|
|
|
for i, (name1, proc1) in enumerate(self.pipeline):
|
|
|
|
|
if hasattr(proc1, "find_listeners"):
|
2020-09-21 11:59:07 +03:00
|
|
|
|
for name2, proc2 in self.pipeline[i + 1 :]:
|
2021-01-20 03:12:35 +03:00
|
|
|
|
proc1.find_listeners(proc2)
|
2020-01-29 19:06:46 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
@classmethod
|
|
|
|
|
def from_config(
|
|
|
|
|
cls,
|
|
|
|
|
config: Union[Dict[str, Any], Config] = {},
|
2020-07-27 01:27:53 +03:00
|
|
|
|
*,
|
2020-08-05 00:39:19 +03:00
|
|
|
|
vocab: Union[Vocab, bool] = True,
|
2020-08-29 16:20:11 +03:00
|
|
|
|
disable: Iterable[str] = SimpleFrozenList(),
|
|
|
|
|
exclude: Iterable[str] = SimpleFrozenList(),
|
2020-09-15 12:12:12 +03:00
|
|
|
|
meta: Dict[str, Any] = SimpleFrozenDict(),
|
2020-07-22 14:42:59 +03:00
|
|
|
|
auto_fill: bool = True,
|
|
|
|
|
validate: bool = True,
|
|
|
|
|
) -> "Language":
|
|
|
|
|
"""Create the nlp object from a loaded config. Will set up the tokenizer
|
|
|
|
|
and language data, add pipeline components etc. If no config is provided,
|
|
|
|
|
the default config of the given language is used.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
|
|
|
|
|
config (Dict[str, Any] / Config): The loaded config.
|
2020-08-05 00:39:19 +03:00
|
|
|
|
vocab (Vocab): A Vocab object. If True, a vocab is created.
|
2020-08-28 16:20:14 +03:00
|
|
|
|
disable (Iterable[str]): Names of pipeline components to disable.
|
|
|
|
|
Disabled pipes will be loaded but they won't be run unless you
|
|
|
|
|
explicitly enable them by calling nlp.enable_pipe.
|
|
|
|
|
exclude (Iterable[str]): Names of pipeline components to exclude.
|
|
|
|
|
Excluded components won't be loaded.
|
2020-09-15 12:12:12 +03:00
|
|
|
|
meta (Dict[str, Any]): Meta overrides for nlp.meta.
|
2020-07-29 00:12:42 +03:00
|
|
|
|
auto_fill (bool): Automatically fill in missing values in config based
|
|
|
|
|
on defaults and function argument annotations.
|
|
|
|
|
validate (bool): Validate the component config and arguments against
|
|
|
|
|
the types expected by the factory.
|
|
|
|
|
RETURNS (Language): The initialized Language class.
|
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#from_config
|
2020-07-22 14:42:59 +03:00
|
|
|
|
"""
|
|
|
|
|
if auto_fill:
|
2020-08-14 15:06:22 +03:00
|
|
|
|
config = Config(
|
|
|
|
|
cls.default_config, section_order=CONFIG_SECTION_ORDER
|
|
|
|
|
).merge(config)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
if "nlp" not in config:
|
|
|
|
|
raise ValueError(Errors.E985.format(config=config))
|
2020-09-29 22:08:13 +03:00
|
|
|
|
config_lang = config["nlp"].get("lang")
|
2020-09-15 15:24:06 +03:00
|
|
|
|
if config_lang is not None and config_lang != cls.lang:
|
2020-07-22 14:42:59 +03:00
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E958.format(
|
2020-07-24 15:50:26 +03:00
|
|
|
|
bad_lang_code=config["nlp"]["lang"],
|
2020-07-22 14:42:59 +03:00
|
|
|
|
lang_code=cls.lang,
|
|
|
|
|
lang=util.get_object_name(cls),
|
|
|
|
|
)
|
|
|
|
|
)
|
2020-07-24 15:50:26 +03:00
|
|
|
|
config["nlp"]["lang"] = cls.lang
|
2020-07-22 14:42:59 +03:00
|
|
|
|
# This isn't very elegant, but we remove the [components] block here to prevent
|
|
|
|
|
# it from getting resolved (causes problems because we expect to pass in
|
|
|
|
|
# the nlp and name args for each component). If we're auto-filling, we're
|
|
|
|
|
# using the nlp.config with all defaults.
|
|
|
|
|
config = util.copy_config(config)
|
|
|
|
|
orig_pipeline = config.pop("components", {})
|
|
|
|
|
config["components"] = {}
|
2020-09-27 23:50:36 +03:00
|
|
|
|
if auto_fill:
|
|
|
|
|
filled = registry.fill(config, validate=validate, schema=ConfigSchema)
|
|
|
|
|
else:
|
|
|
|
|
filled = config
|
2020-07-22 14:42:59 +03:00
|
|
|
|
filled["components"] = orig_pipeline
|
|
|
|
|
config["components"] = orig_pipeline
|
2020-09-27 23:50:36 +03:00
|
|
|
|
resolved_nlp = registry.resolve(
|
|
|
|
|
filled["nlp"], validate=validate, schema=ConfigSchemaNlp
|
|
|
|
|
)
|
2020-09-27 23:21:31 +03:00
|
|
|
|
create_tokenizer = resolved_nlp["tokenizer"]
|
|
|
|
|
before_creation = resolved_nlp["before_creation"]
|
|
|
|
|
after_creation = resolved_nlp["after_creation"]
|
|
|
|
|
after_pipeline_creation = resolved_nlp["after_pipeline_creation"]
|
2020-08-05 20:47:54 +03:00
|
|
|
|
lang_cls = cls
|
|
|
|
|
if before_creation is not None:
|
|
|
|
|
lang_cls = before_creation(cls)
|
|
|
|
|
if (
|
|
|
|
|
not isinstance(lang_cls, type)
|
|
|
|
|
or not issubclass(lang_cls, cls)
|
|
|
|
|
or lang_cls is not cls
|
|
|
|
|
):
|
|
|
|
|
raise ValueError(Errors.E943.format(value=type(lang_cls)))
|
2020-08-13 18:38:30 +03:00
|
|
|
|
# Note that we don't load vectors here, instead they get loaded explicitly
|
|
|
|
|
# inside stuff like the spacy train function. If we loaded them here,
|
|
|
|
|
# then we would load them twice at runtime: once when we make from config,
|
|
|
|
|
# and then again when we load from disk.
|
2020-09-15 12:12:12 +03:00
|
|
|
|
nlp = lang_cls(vocab=vocab, create_tokenizer=create_tokenizer, meta=meta)
|
2020-08-05 20:47:54 +03:00
|
|
|
|
if after_creation is not None:
|
|
|
|
|
nlp = after_creation(nlp)
|
|
|
|
|
if not isinstance(nlp, cls):
|
|
|
|
|
raise ValueError(Errors.E942.format(name="creation", value=type(nlp)))
|
2020-08-13 18:38:30 +03:00
|
|
|
|
# To create the components we need to use the final interpolated config
|
|
|
|
|
# so all values are available (if component configs use variables).
|
|
|
|
|
# Later we replace the component config with the raw config again.
|
|
|
|
|
interpolated = filled.interpolate() if not filled.is_interpolated else filled
|
|
|
|
|
pipeline = interpolated.get("components", {})
|
2020-08-05 00:39:19 +03:00
|
|
|
|
# If components are loaded from a source (existing models), we cache
|
|
|
|
|
# them here so they're only loaded once
|
|
|
|
|
source_nlps = {}
|
2020-07-24 15:50:26 +03:00
|
|
|
|
for pipe_name in config["nlp"]["pipeline"]:
|
2020-07-22 14:42:59 +03:00
|
|
|
|
if pipe_name not in pipeline:
|
|
|
|
|
opts = ", ".join(pipeline.keys())
|
|
|
|
|
raise ValueError(Errors.E956.format(name=pipe_name, opts=opts))
|
2020-07-22 18:29:31 +03:00
|
|
|
|
pipe_cfg = util.copy_config(pipeline[pipe_name])
|
2020-08-13 18:38:30 +03:00
|
|
|
|
raw_config = Config(filled["components"][pipe_name])
|
2020-08-28 16:20:14 +03:00
|
|
|
|
if pipe_name not in exclude:
|
2020-08-05 00:39:19 +03:00
|
|
|
|
if "factory" not in pipe_cfg and "source" not in pipe_cfg:
|
2020-07-22 14:42:59 +03:00
|
|
|
|
err = Errors.E984.format(name=pipe_name, config=pipe_cfg)
|
|
|
|
|
raise ValueError(err)
|
2020-08-05 00:39:19 +03:00
|
|
|
|
if "factory" in pipe_cfg:
|
|
|
|
|
factory = pipe_cfg.pop("factory")
|
|
|
|
|
# The pipe name (key in the config) here is the unique name
|
|
|
|
|
# of the component, not necessarily the factory
|
|
|
|
|
nlp.add_pipe(
|
2020-08-13 18:38:30 +03:00
|
|
|
|
factory,
|
|
|
|
|
name=pipe_name,
|
|
|
|
|
config=pipe_cfg,
|
|
|
|
|
validate=validate,
|
|
|
|
|
raw_config=raw_config,
|
2020-08-05 00:39:19 +03:00
|
|
|
|
)
|
|
|
|
|
else:
|
|
|
|
|
model = pipe_cfg["source"]
|
|
|
|
|
if model not in source_nlps:
|
|
|
|
|
# We only need the components here and we need to init
|
|
|
|
|
# model with the same vocab as the current nlp object
|
2021-01-16 04:26:15 +03:00
|
|
|
|
source_nlps[model] = util.load_model(model, vocab=nlp.vocab)
|
2020-08-05 00:39:19 +03:00
|
|
|
|
source_name = pipe_cfg.get("component", pipe_name)
|
|
|
|
|
nlp.add_pipe(source_name, source=source_nlps[model], name=pipe_name)
|
2020-08-28 16:20:14 +03:00
|
|
|
|
disabled_pipes = [*config["nlp"]["disabled"], *disable]
|
2020-08-29 13:08:33 +03:00
|
|
|
|
nlp._disabled = set(p for p in disabled_pipes if p not in exclude)
|
2020-12-09 11:13:26 +03:00
|
|
|
|
nlp.batch_size = config["nlp"]["batch_size"]
|
2020-07-22 14:42:59 +03:00
|
|
|
|
nlp.config = filled if auto_fill else config
|
2020-08-05 20:47:54 +03:00
|
|
|
|
if after_pipeline_creation is not None:
|
|
|
|
|
nlp = after_pipeline_creation(nlp)
|
|
|
|
|
if not isinstance(nlp, cls):
|
|
|
|
|
raise ValueError(
|
|
|
|
|
Errors.E942.format(name="pipeline_creation", value=type(nlp))
|
|
|
|
|
)
|
2020-07-22 14:42:59 +03:00
|
|
|
|
return nlp
|
|
|
|
|
|
2020-07-29 16:14:07 +03:00
|
|
|
|
def to_disk(
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self, path: Union[str, Path], *, exclude: Iterable[str] = SimpleFrozenList()
|
2020-07-29 16:14:07 +03:00
|
|
|
|
) -> None:
|
2017-05-26 13:33:54 +03:00
|
|
|
|
"""Save the current state to a directory. If a model is loaded, this
|
|
|
|
|
will include the model.
|
2017-04-17 02:40:26 +03:00
|
|
|
|
|
2020-05-24 19:51:10 +03:00
|
|
|
|
path (str / Path): Path to a directory, which will be created if
|
2019-03-10 21:16:45 +03:00
|
|
|
|
it doesn't exist.
|
|
|
|
|
exclude (list): Names of components or serialization fields to exclude.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#to_disk
|
2017-05-17 13:04:50 +03:00
|
|
|
|
"""
|
|
|
|
|
path = util.ensure_path(path)
|
2019-12-22 03:53:56 +03:00
|
|
|
|
serializers = {}
|
2019-08-01 18:13:01 +03:00
|
|
|
|
serializers["tokenizer"] = lambda p: self.tokenizer.to_disk(
|
|
|
|
|
p, exclude=["vocab"]
|
|
|
|
|
)
|
2020-04-06 19:54:32 +03:00
|
|
|
|
serializers["meta.json"] = lambda p: srsly.write_json(p, self.meta)
|
2020-02-27 20:42:27 +03:00
|
|
|
|
serializers["config.cfg"] = lambda p: self.config.to_disk(p)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
for name, proc in self._components:
|
2019-03-10 21:16:45 +03:00
|
|
|
|
if name in exclude:
|
2017-05-31 14:42:39 +03:00
|
|
|
|
continue
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
if not hasattr(proc, "to_disk"):
|
2017-05-31 14:42:39 +03:00
|
|
|
|
continue
|
2019-03-10 21:16:45 +03:00
|
|
|
|
serializers[name] = lambda p, proc=proc: proc.to_disk(p, exclude=["vocab"])
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
serializers["vocab"] = lambda p: self.vocab.to_disk(p)
|
2019-03-10 21:16:45 +03:00
|
|
|
|
util.to_disk(path, serializers, exclude)
|
2017-05-31 14:42:39 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def from_disk(
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self, path: Union[str, Path], *, exclude: Iterable[str] = SimpleFrozenList()
|
2020-07-22 14:42:59 +03:00
|
|
|
|
) -> "Language":
|
2017-05-19 00:57:38 +03:00
|
|
|
|
"""Loads state from a directory. Modifies the object in place and
|
2017-05-26 13:33:54 +03:00
|
|
|
|
returns it. If the saved `Language` object contains a model, the
|
|
|
|
|
model will be loaded.
|
2017-05-17 13:04:50 +03:00
|
|
|
|
|
2020-05-24 19:51:10 +03:00
|
|
|
|
path (str / Path): A path to a directory.
|
2019-03-10 21:16:45 +03:00
|
|
|
|
exclude (list): Names of components or serialization fields to exclude.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
RETURNS (Language): The modified `Language` object.
|
2017-05-17 13:04:50 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#from_disk
|
2017-05-17 13:04:50 +03:00
|
|
|
|
"""
|
2020-06-20 16:52:00 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def deserialize_meta(path: Path) -> None:
|
2020-05-27 15:48:54 +03:00
|
|
|
|
if path.exists():
|
|
|
|
|
data = srsly.read_json(path)
|
|
|
|
|
self.meta.update(data)
|
|
|
|
|
# self.meta always overrides meta["vectors"] with the metadata
|
|
|
|
|
# from self.vocab.vectors, so set the name directly
|
|
|
|
|
self.vocab.vectors.name = data.get("vectors", {}).get("name")
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def deserialize_vocab(path: Path) -> None:
|
2020-05-27 15:48:54 +03:00
|
|
|
|
if path.exists():
|
|
|
|
|
self.vocab.from_disk(path)
|
|
|
|
|
|
2017-05-17 13:04:50 +03:00
|
|
|
|
path = util.ensure_path(path)
|
2019-12-22 03:53:56 +03:00
|
|
|
|
deserializers = {}
|
2020-02-27 20:42:27 +03:00
|
|
|
|
if Path(path / "config.cfg").exists():
|
2020-08-27 17:44:36 +03:00
|
|
|
|
deserializers["config.cfg"] = lambda p: self.config.from_disk(
|
|
|
|
|
p, interpolate=False
|
|
|
|
|
)
|
2020-05-27 15:48:54 +03:00
|
|
|
|
deserializers["meta.json"] = deserialize_meta
|
|
|
|
|
deserializers["vocab"] = deserialize_vocab
|
2019-08-01 18:13:01 +03:00
|
|
|
|
deserializers["tokenizer"] = lambda p: self.tokenizer.from_disk(
|
|
|
|
|
p, exclude=["vocab"]
|
|
|
|
|
)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
for name, proc in self._components:
|
2019-03-10 21:16:45 +03:00
|
|
|
|
if name in exclude:
|
2017-05-31 14:42:39 +03:00
|
|
|
|
continue
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
if not hasattr(proc, "from_disk"):
|
2017-05-31 14:42:39 +03:00
|
|
|
|
continue
|
2019-08-01 18:13:01 +03:00
|
|
|
|
deserializers[name] = lambda p, proc=proc: proc.from_disk(
|
|
|
|
|
p, exclude=["vocab"]
|
|
|
|
|
)
|
2019-03-10 21:16:45 +03:00
|
|
|
|
if not (path / "vocab").exists() and "vocab" not in exclude:
|
|
|
|
|
# Convert to list here in case exclude is (default) tuple
|
|
|
|
|
exclude = list(exclude) + ["vocab"]
|
2017-06-01 15:38:35 +03:00
|
|
|
|
util.from_disk(path, deserializers, exclude)
|
2017-10-25 12:57:43 +03:00
|
|
|
|
self._path = path
|
2020-01-29 19:06:46 +03:00
|
|
|
|
self._link_components()
|
2017-05-31 14:42:39 +03:00
|
|
|
|
return self
|
2017-05-17 13:04:50 +03:00
|
|
|
|
|
2020-08-29 16:20:11 +03:00
|
|
|
|
def to_bytes(self, *, exclude: Iterable[str] = SimpleFrozenList()) -> bytes:
|
2017-05-17 13:04:50 +03:00
|
|
|
|
"""Serialize the current state to a binary string.
|
2016-12-18 18:54:52 +03:00
|
|
|
|
|
2019-03-10 21:16:45 +03:00
|
|
|
|
exclude (list): Names of components or serialization fields to exclude.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
RETURNS (bytes): The serialized form of the `Language` object.
|
2019-03-10 21:16:45 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#to_bytes
|
2017-05-17 13:04:50 +03:00
|
|
|
|
"""
|
2019-12-22 03:53:56 +03:00
|
|
|
|
serializers = {}
|
2019-03-10 21:16:45 +03:00
|
|
|
|
serializers["vocab"] = lambda: self.vocab.to_bytes()
|
|
|
|
|
serializers["tokenizer"] = lambda: self.tokenizer.to_bytes(exclude=["vocab"])
|
|
|
|
|
serializers["meta.json"] = lambda: srsly.json_dumps(self.meta)
|
2020-02-27 20:42:27 +03:00
|
|
|
|
serializers["config.cfg"] = lambda: self.config.to_bytes()
|
2020-08-29 16:20:11 +03:00
|
|
|
|
for name, proc in self._components:
|
2019-03-10 21:16:45 +03:00
|
|
|
|
if name in exclude:
|
2017-05-29 12:45:45 +03:00
|
|
|
|
continue
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
if not hasattr(proc, "to_bytes"):
|
2017-05-29 12:45:45 +03:00
|
|
|
|
continue
|
2019-03-10 21:16:45 +03:00
|
|
|
|
serializers[name] = lambda proc=proc: proc.to_bytes(exclude=["vocab"])
|
2017-10-17 19:18:10 +03:00
|
|
|
|
return util.to_bytes(serializers, exclude)
|
2017-04-15 13:05:47 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def from_bytes(
|
2020-08-29 16:20:11 +03:00
|
|
|
|
self, bytes_data: bytes, *, exclude: Iterable[str] = SimpleFrozenList()
|
2020-07-22 14:42:59 +03:00
|
|
|
|
) -> "Language":
|
2017-05-17 13:04:50 +03:00
|
|
|
|
"""Load state from a binary string.
|
|
|
|
|
|
2017-05-19 00:57:38 +03:00
|
|
|
|
bytes_data (bytes): The data to load from.
|
2019-03-10 21:16:45 +03:00
|
|
|
|
exclude (list): Names of components or serialization fields to exclude.
|
2017-05-19 00:57:38 +03:00
|
|
|
|
RETURNS (Language): The `Language` object.
|
2019-03-10 21:16:45 +03:00
|
|
|
|
|
2020-09-04 13:58:50 +03:00
|
|
|
|
DOCS: https://nightly.spacy.io/api/language#from_bytes
|
2017-05-17 13:04:50 +03:00
|
|
|
|
"""
|
2020-06-20 16:52:00 +03:00
|
|
|
|
|
2020-05-27 15:48:54 +03:00
|
|
|
|
def deserialize_meta(b):
|
|
|
|
|
data = srsly.json_loads(b)
|
|
|
|
|
self.meta.update(data)
|
|
|
|
|
# self.meta always overrides meta["vectors"] with the metadata
|
|
|
|
|
# from self.vocab.vectors, so set the name directly
|
|
|
|
|
self.vocab.vectors.name = data.get("vectors", {}).get("name")
|
|
|
|
|
|
2019-12-22 03:53:56 +03:00
|
|
|
|
deserializers = {}
|
2020-08-27 17:44:36 +03:00
|
|
|
|
deserializers["config.cfg"] = lambda b: self.config.from_bytes(
|
|
|
|
|
b, interpolate=False
|
|
|
|
|
)
|
2020-05-27 15:48:54 +03:00
|
|
|
|
deserializers["meta.json"] = deserialize_meta
|
2020-08-25 15:37:45 +03:00
|
|
|
|
deserializers["vocab"] = self.vocab.from_bytes
|
2019-08-01 18:13:01 +03:00
|
|
|
|
deserializers["tokenizer"] = lambda b: self.tokenizer.from_bytes(
|
|
|
|
|
b, exclude=["vocab"]
|
|
|
|
|
)
|
2020-08-29 16:20:11 +03:00
|
|
|
|
for name, proc in self._components:
|
2019-03-10 21:16:45 +03:00
|
|
|
|
if name in exclude:
|
2017-05-29 12:45:45 +03:00
|
|
|
|
continue
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
if not hasattr(proc, "from_bytes"):
|
2017-05-29 12:45:45 +03:00
|
|
|
|
continue
|
2019-08-01 18:13:01 +03:00
|
|
|
|
deserializers[name] = lambda b, proc=proc: proc.from_bytes(
|
|
|
|
|
b, exclude=["vocab"]
|
|
|
|
|
)
|
2019-03-10 21:16:45 +03:00
|
|
|
|
util.from_bytes(bytes_data, deserializers, exclude)
|
2020-01-29 19:06:46 +03:00
|
|
|
|
self._link_components()
|
2017-05-17 13:04:50 +03:00
|
|
|
|
return self
|
2017-05-22 02:43:31 +03:00
|
|
|
|
|
2017-05-29 12:45:45 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
@dataclass
|
|
|
|
|
class FactoryMeta:
|
2020-07-29 00:12:42 +03:00
|
|
|
|
"""Dataclass containing information about a component and its defaults
|
|
|
|
|
provided by the @Language.component or @Language.factory decorator. It's
|
|
|
|
|
created whenever a component is defined and stored on the Language class for
|
|
|
|
|
each component instance and factory instance.
|
|
|
|
|
"""
|
2020-07-29 16:14:07 +03:00
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
factory: str
|
|
|
|
|
default_config: Optional[Dict[str, Any]] = None # noqa: E704
|
|
|
|
|
assigns: Iterable[str] = tuple()
|
|
|
|
|
requires: Iterable[str] = tuple()
|
|
|
|
|
retokenizes: bool = False
|
2020-07-26 14:18:43 +03:00
|
|
|
|
scores: Iterable[str] = tuple()
|
2020-07-28 12:22:24 +03:00
|
|
|
|
default_score_weights: Optional[Dict[str, float]] = None # noqa: E704
|
2019-10-27 15:35:49 +03:00
|
|
|
|
|
|
|
|
|
|
2017-10-25 14:46:41 +03:00
|
|
|
|
class DisabledPipes(list):
|
2017-10-27 15:40:14 +03:00
|
|
|
|
"""Manager for temporary pipeline disabling."""
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
|
2020-07-29 00:12:42 +03:00
|
|
|
|
def __init__(self, nlp: Language, names: List[str]) -> None:
|
2017-10-25 14:46:41 +03:00
|
|
|
|
self.nlp = nlp
|
|
|
|
|
self.names = names
|
2020-08-28 16:20:14 +03:00
|
|
|
|
for name in self.names:
|
|
|
|
|
self.nlp.disable_pipe(name)
|
2017-10-25 14:46:41 +03:00
|
|
|
|
list.__init__(self)
|
2020-08-28 16:20:14 +03:00
|
|
|
|
self.extend(self.names)
|
2017-10-25 14:46:41 +03:00
|
|
|
|
|
|
|
|
|
def __enter__(self):
|
2017-10-25 15:56:16 +03:00
|
|
|
|
return self
|
2017-10-25 14:46:41 +03:00
|
|
|
|
|
|
|
|
|
def __exit__(self, *args):
|
|
|
|
|
self.restore()
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def restore(self) -> None:
|
💫 Tidy up and auto-format .py files (#2983)
<!--- Provide a general summary of your changes in the title. -->
## Description
- [x] Use [`black`](https://github.com/ambv/black) to auto-format all `.py` files.
- [x] Update flake8 config to exclude very large files (lemmatization tables etc.)
- [x] Update code to be compatible with flake8 rules
- [x] Fix various small bugs, inconsistencies and messy stuff in the language data
- [x] Update docs to explain new code style (`black`, `flake8`, when to use `# fmt: off` and `# fmt: on` and what `# noqa` means)
Once #2932 is merged, which auto-formats and tidies up the CLI, we'll be able to run `flake8 spacy` actually get meaningful results.
At the moment, the code style and linting isn't applied automatically, but I'm hoping that the new [GitHub Actions](https://github.com/features/actions) will let us auto-format pull requests and post comments with relevant linting information.
### Types of change
enhancement, code style
## Checklist
<!--- Before you submit the PR, go over this checklist and make sure you can
tick off all the boxes. [] -> [x] -->
- [x] I have submitted the spaCy Contributor Agreement.
- [x] I ran the tests, and all new and existing tests passed.
- [x] My changes don't require a change to the documentation, or if they do, I've added all required information.
2018-11-30 19:03:03 +03:00
|
|
|
|
"""Restore the pipeline to its state when DisabledPipes was created."""
|
2020-08-28 16:20:14 +03:00
|
|
|
|
for name in self.names:
|
2020-08-28 22:04:02 +03:00
|
|
|
|
if name not in self.nlp.component_names:
|
2020-08-28 21:35:26 +03:00
|
|
|
|
raise ValueError(Errors.E008.format(name=name))
|
2020-08-28 16:20:14 +03:00
|
|
|
|
self.nlp.enable_pipe(name)
|
2017-10-25 14:46:41 +03:00
|
|
|
|
self[:] = []
|
|
|
|
|
|
|
|
|
|
|
2021-01-19 18:47:44 +03:00
|
|
|
|
def _copy_examples(examples: Iterable[Example]) -> List[Example]:
|
|
|
|
|
"""Make a copy of a batch of examples, copying the predicted Doc as well.
|
|
|
|
|
This is used in contexts where we need to take ownership of the examples
|
|
|
|
|
so that they can be mutated, for instance during Language.evaluate and
|
|
|
|
|
Language.update.
|
|
|
|
|
"""
|
|
|
|
|
return [Example(eg.x.copy(), eg.y) for eg in examples]
|
|
|
|
|
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def _apply_pipes(
|
|
|
|
|
make_doc: Callable[[str], Doc],
|
|
|
|
|
pipes: Iterable[Callable[[Doc], Doc]],
|
|
|
|
|
receiver,
|
|
|
|
|
sender,
|
|
|
|
|
underscore_state: Tuple[dict, dict, dict],
|
|
|
|
|
) -> None:
|
2019-10-08 13:20:55 +03:00
|
|
|
|
"""Worker for Language.pipe
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
make_doc (Callable[[str,] Doc]): Function to create Doc from text.
|
|
|
|
|
pipes (Iterable[Callable[[Doc], Doc]]): The components to apply.
|
2019-10-18 12:33:38 +03:00
|
|
|
|
receiver (multiprocessing.Connection): Pipe to receive text. Usually
|
|
|
|
|
created by `multiprocessing.Pipe()`
|
|
|
|
|
sender (multiprocessing.Connection): Pipe to send doc. Usually created by
|
|
|
|
|
`multiprocessing.Pipe()`
|
2020-07-22 14:42:59 +03:00
|
|
|
|
underscore_state (Tuple[dict, dict, dict]): The data in the Underscore class
|
|
|
|
|
of the parent.
|
2019-10-08 13:20:55 +03:00
|
|
|
|
"""
|
2020-02-12 13:50:42 +03:00
|
|
|
|
Underscore.load_state(underscore_state)
|
2019-10-08 13:20:55 +03:00
|
|
|
|
while True:
|
2020-02-12 13:50:42 +03:00
|
|
|
|
texts = receiver.get()
|
2019-10-08 13:20:55 +03:00
|
|
|
|
docs = (make_doc(text) for text in texts)
|
|
|
|
|
for pipe in pipes:
|
|
|
|
|
docs = pipe(docs)
|
|
|
|
|
# Connection does not accept unpickable objects, so send list.
|
|
|
|
|
sender.send([doc.to_bytes() for doc in docs])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class _Sender:
|
|
|
|
|
"""Util for sending data to multiprocessing workers in Language.pipe"""
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def __init__(
|
|
|
|
|
self, data: Iterable[Any], queues: List[mp.Queue], chunk_size: int
|
|
|
|
|
) -> None:
|
2019-10-08 13:20:55 +03:00
|
|
|
|
self.data = iter(data)
|
|
|
|
|
self.queues = iter(cycle(queues))
|
|
|
|
|
self.chunk_size = chunk_size
|
|
|
|
|
self.count = 0
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def send(self) -> None:
|
2019-10-08 13:20:55 +03:00
|
|
|
|
"""Send chunk_size items from self.data to channels."""
|
|
|
|
|
for item, q in itertools.islice(
|
|
|
|
|
zip(self.data, cycle(self.queues)), self.chunk_size
|
|
|
|
|
):
|
|
|
|
|
# cycle channels so that distribute the texts evenly
|
|
|
|
|
q.put(item)
|
|
|
|
|
|
2020-07-22 14:42:59 +03:00
|
|
|
|
def step(self) -> None:
|
|
|
|
|
"""Tell sender that comsumed one item. Data is sent to the workers after
|
|
|
|
|
every chunk_size calls.
|
|
|
|
|
"""
|
2019-10-08 13:20:55 +03:00
|
|
|
|
self.count += 1
|
|
|
|
|
if self.count >= self.chunk_size:
|
|
|
|
|
self.count = 0
|
|
|
|
|
self.send()
|