* Store activations in Doc when `store_activations` is enabled
This change adds the new `activations` attribute to `Doc`. This
attribute can be used by trainable pipes to store their activations,
probabilities, and guesses for downstream users.
As an example, this change modifies the `tagger` and `senter` pipes to
add an `store_activations` option. When this option is enabled, the
probabilities and guesses are stored in `set_annotations`.
* Change type of `store_activations` to `Union[bool, List[str]]`
When the value is:
- A bool: all activations are stored when set to `True`.
- A List[str]: the activations named in the list are stored
* Formatting fixes in Tagger
* Support store_activations in spancat and morphologizer
* Make Doc.activations type visible to MyPy
* textcat/textcat_multilabel: add store_activations option
* trainable_lemmatizer/entity_linker: add store_activations option
* parser/ner: do not currently support returning activations
* Extend tagger and senter tests
So that they, like the other tests, also check that we get no
activations if no activations were requested.
* Document `Doc.activations` and `store_activations` in the relevant pipes
* Start errors/warnings at higher numbers to avoid merge conflicts
Between the master and v4 branches.
* Add `store_activations` to docstrings.
* Replace store_activations setter by set_store_activations method
Setters that take a different type than what the getter returns are still
problematic for MyPy. Replace the setter by a method, so that type inference
works everywhere.
* Use dict comprehension suggested by @svlandeg
* Revert "Use dict comprehension suggested by @svlandeg"
This reverts commit 6e7b958f70.
* EntityLinker: add type annotations to _add_activations
* _store_activations: make kwarg-only, remove doc_scores_lens arg
* set_annotations: add type annotations
* Apply suggestions from code review
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* TextCat.predict: return dict
* Make the `TrainablePipe.store_activations` property a bool
This means that we can also bring back `store_activations` setter.
* Remove `TrainablePipe.activations`
We do not need to enumerate the activations anymore since `store_activations` is
`bool`.
* Add type annotations for activations in predict/set_annotations
* Rename `TrainablePipe.store_activations` to `save_activations`
* Error E1400 is not used anymore
This error was used when activations were still `Union[bool, List[str]]`.
* Change wording in API docs after store -> save change
* docs: tag (save_)activations as new in spaCy 4.0
* Fix copied line in morphologizer activations test
* Don't train in any test_save_activations test
* Rename activations
- "probs" -> "probabilities"
- "guesses" -> "label_ids", except in the edit tree lemmatizer, where
"guesses" -> "tree_ids".
* Remove unused W400 warning.
This warning was used when we still allowed the user to specify
which activations to save.
* Formatting fixes
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Replace "kb_ids" by a constant
* spancat: replace a cast by an assertion
* Fix EOF spacing
* Fix comments in test_save_activations tests
* Do not set RNG seed in activation saving tests
* Revert "spancat: replace a cast by an assertion"
This reverts commit 0bd5730d16.
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Tagger: use unnormalized probabilities for inference
Using unnormalized softmax avoids use of the relatively expensive exp function,
which can significantly speed up non-transformer models (e.g. I got a speedup
of 27% on a German tagging + parsing pipeline).
* Add spacy.Tagger.v2 with configurable normalization
Normalization of probabilities is disabled by default to improve
performance.
* Update documentation, models, and tests to spacy.Tagger.v2
* Move Tagger.v1 to spacy-legacy
* docs/architectures: run prettier
* Unnormalized softmax is now a Softmax_v2 option
* Require thinc 8.0.14 and spacy-legacy 3.0.9
* morphologizer: avoid recreating label tuple for each token
The `labels` property converts the dictionary key set to a tuple. This
property was used for every annotated token, recreating the tuple over
and over again.
Construct the tuple once in the set_annotations function and reuse it.
On a Finnish pipeline that I was experimenting with, this results in a
speedup of ~15% (~13000 -> ~15000 WPS).
* tagger: avoid recreating label tuple for each token
* Add overwrite settings for more components
For pipeline components where it's relevant and not already implemented,
add an explicit `overwrite` setting that controls whether
`set_annotations` overwrites existing annotation.
For the `morphologizer`, add an additional setting `extend`, which
controls whether the existing features are preserved.
* +overwrite, +extend: overwrite values of existing features, add any new
features
* +overwrite, -extend: overwrite completely, removing any existing
features
* -overwrite, +extend: keep values of existing features, add any new
features
* -overwrite, -extend: do not modify the existing value if set
In all cases an unset value will be set by `set_annotations`.
Preserve current overwrite defaults:
* True: morphologizer, entity linker
* False: tagger, sentencizer, senter
* Add backwards compat overwrite settings
* Put empty line back
Removed by accident in last commit
* Set backwards-compatible defaults in __init__
Because the `TrainablePipe` serialization methods update `cfg`, there's
no straightforward way to detect whether models serialized with a
previous version are missing the overwrite settings.
It would be possible in the sentencizer due to its separate
serialization methods, however to keep the changes parallel, this also
sets the default in `__init__`.
* Remove traces
Co-authored-by: Paul O'Leary McCann <polm@dampfkraft.com>
* Add scorer option to components
Add an optional `scorer` parameter to all pipeline components. If a
scoring function is provided, it overrides the default scoring method
for that component.
* Add registered scorers for all components
* Add `scorers` registry
* Move all scoring methods outside of components as independent
functions and register
* Use the registered scoring methods as defaults in configs and inits
Additional:
* The scoring methods no longer have access to the full component, so
use settings from `cfg` as default scorer options to handle settings
such as `labels`, `threshold`, and `positive_label`
* The `attribute_ruler` scoring method no longer has access to the
patterns, so all scoring methods are called
* Bug fix: `spancat` scoring method is updated to set `allow_overlap` to
score overlapping spans correctly
* Update Russian lemmatizer to use direct score method
* Check type of cfg in Pipe.score
* Fix check
* Update spacy/pipeline/sentencizer.pyx
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Remove validate_examples from scoring functions
* Use Pipe.labels instead of Pipe.cfg["labels"]
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Replace negative rows with 0 in StaticVectors
Replace negative row indices with 0-vectors in `StaticVectors`.
* Increase versions related to StaticVectors
* Increase versions of all architctures and layers related to
`StaticVectors`
* Improve efficiency of 0-vector operations
Parallel `spacy-legacy` PR: https://github.com/explosion/spacy-legacy/pull/5
* Update config defaults to new versions
* Update docs
* Add util method for check
* Add new languages to list with lexeme norm tables
* Add check to all relevant components
* Add config details to warning message
Note that we're not actually inspecting the model config to see if
`NORM` is used as an attribute, so it may warn in cases where it's not
relevant.
* Handle unset token.morph in Morphologizer
Handle unset `token.morph` in `Morphologizer.initialize` and
`Morphologizer.get_loss`. If both `token.morph` and `token.pos` are
unset, treat the annotation as missing rather than empty.
* Add token.has_morph()
* fix TorchBiLSTMEncoder documentation
* ensure the types of the encoding Tok2vec layers are correct
* update references from v1 to v2 for the new architectures
Fix bug where `Morphologizer.get_loss` treated misaligned annotation as
`EMPTY_MORPH` rather than ignoring it. Remove unneeded default `EMPTY_MORPH`
mappings.
* Handle missing reference values in scorer
Handle missing values in reference doc during scoring where it is
possible to detect an unset state for the attribute. If no reference
docs contain annotation, `None` is returned instead of a score. `spacy
evaluate` displays `-` for missing scores and the missing scores are
saved as `None`/`null` in the metrics.
Attributes without unset states:
* `token.head`: relies on `token.dep` to recognize unset values
* `doc.cats`: unable to handle missing annotation
Additional changes:
* add optional `has_annotation` check to `score_scans` to replace
`doc.sents` hack
* update `score_token_attr_per_feat` to handle missing and empty morph
representations
* fix bug in `Doc.has_annotation` for normalization of `IS_SENT_START`
vs. `SENT_START`
* Fix import
* Update return types
* rename Pipe to TrainablePipe
* split functionality between Pipe and TrainablePipe
* remove unnecessary methods from certain components
* cleanup
* hasattr(component, "pipe") should be sufficient again
* remove serialization and vocab/cfg from Pipe
* unify _ensure_examples and validate_examples
* small fixes
* hasattr checks for self.cfg and self.vocab
* make is_resizable and is_trainable properties
* serialize strings.json instead of vocab
* fix KB IO + tests
* fix typos
* more typos
* _added_strings as a set
* few more tests specifically for _added_strings field
* bump to 3.0.0a36
* Refactor Token morph setting
* Remove `Token.morph_`
* Add `Token.set_morph()`
* `0` resets `token.c.morph` to unset
* Any other values are passed to `Morphology.add`
* Add token.morph setter to set from MorphAnalysis
* Refactor Docs.is_ flags
* Add derived `Doc.has_annotation` method
* `Doc.has_annotation(attr)` returns `True` for partial annotation
* `Doc.has_annotation(attr, require_complete=True)` returns `True` for
complete annotation
* Add deprecation warnings to `is_tagged`, `is_parsed`, `is_sentenced`
and `is_nered`
* Add `Doc._get_array_attrs()`, which returns a full list of `Doc` attrs
for use with `Doc.to_array`, `Doc.to_bytes` and `Doc.from_docs`. The
list is the `DocBin` attributes list plus `SPACY` and `LENGTH`.
Notes on `Doc.has_annotation`:
* `HEAD` is converted to `DEP` because heads don't have an unset state
* Accept `IS_SENT_START` as a synonym of `SENT_START`
Additional changes:
* Add `NORM`, `ENT_ID` and `SENT_START` to default attributes for
`DocBin`
* In `Doc.from_array()` the presence of `DEP` causes `HEAD` to override
`SENT_START`
* In `Doc.from_array()` using `attrs` other than
`Doc._get_array_attrs()` (i.e., a user's custom list rather than our
default internal list) with both `HEAD` and `SENT_START` shows a warning
that `HEAD` will override `SENT_START`
* `set_children_from_heads` does not require dependency labels to set
sentence boundaries and sets `sent_start` for all non-sentence starts to
`-1`
* Fix call to set_children_form_heads
Co-authored-by: Matthew Honnibal <honnibal+gh@gmail.com>
* ensure Language passes on valid examples for initialization
* fix tagger model initialization
* check for valid get_examples across components
* assume labels were added before begin_training
* fix senter initialization
* fix morphologizer initialization
* use methods to check arguments
* test textcat init, requires thinc>=8.0.0a31
* fix tok2vec init
* fix entity linker init
* use islice
* fix simple NER
* cleanup debug model
* fix assert statements
* fix tests
* throw error when adding a label if the output layer can't be resized anymore
* fix test
* add failing test for simple_ner
* UX improvements
* morphologizer UX
* assume begin_training gets a representative set and processes the labels
* remove assumptions for output of untrained NER model
* restore test for original purpose
Add and update `score` methods, provided `scores`, and default weights
`default_score_weights` for pipeline components.
* `scores` provides all top-level keys returned by `score` (merely informative, similar to `assigns`).
* `default_score_weights` provides the default weights for a default config.
* The keys from `default_score_weights` determine which values will be
shown in the `spacy train` output, so keys with weight `0.0` will be
displayed but not counted toward the overall score.
* 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
