* add span key option for CLI evaluation
* Rephrase CLI help to refer to Doc.spans instead of spancat
* Rephrase docs to refer to Doc.spans instead of spancat
---------
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* fix construction example
* shorten task-specific factory list
* small edits to HF models
* small edit to API models
* typo
* fix space
Co-authored-by: Raphael Mitsch <r.mitsch@outlook.com>
---------
Co-authored-by: Raphael Mitsch <r.mitsch@outlook.com>
* initial
* initial documentation run
* fix typo
* Remove mentions of Torchscript and quantization
Both are disabled in the initial release of `spacy-curated-transformers`.
* Fix `piece_encoder` entries
* Remove `spacy-transformers`-specific warning
* Fix duplicate entries in tables
* Doc fixes
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Remove type aliases
* Fix copy-paste typo
* Change `debug pieces` version tag to `3.7`
* Set curated transformers API version to `3.7`
* Fix transformer listener naming
* Add docs for `init fill-config-transformer`
* Update CLI command invocation syntax
* Update intro section of the pipeline component docs
* Fix source URL
* Add a note to the architectures section about the `init fill-config-transformer` CLI command
* Apply suggestions from code review
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Update CLI command name, args
* Remove hyphen from the `curated-transformers.mdx` filename
* Fix links
* Remove placeholder text
* Add text to the model/tokenizer loader sections
* Fill in the `DocTransformerOutput` section
* Formatting fixes
* Add curated transformer page to API docs sidebar
* More formatting fixes
* Remove TODO comment
* Remove outdated info about default config
* Apply suggestions from code review
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Add link to HF model hub
* `prettier`
---------
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
SpaCy's HashEmbedCNN layer performs convolutions over tokens to produce
contextualized embeddings using a `MaxoutWindowEncoder` layer. These
convolutions are implemented using Thinc's `expand_window` layer, which
concatenates `window_size` neighboring sequence items on either side of
the sequence item being processed. This is repeated across `depth`
convolutional layers.
For example, consider the sequence "ABCDE" and a `MaxoutWindowEncoder`
layer with a context window of 1 and a depth of 2. We'll focus on the
token "C". We can visually represent the contextual embedding produced
for "C" as:
```mermaid
flowchart LR
A0(A<sub>0</sub>)
B0(B<sub>0</sub>)
C0(C<sub>0</sub>)
D0(D<sub>0</sub>)
E0(E<sub>0</sub>)
B1(B<sub>1</sub>)
C1(C<sub>1</sub>)
D1(D<sub>1</sub>)
C2(C<sub>2</sub>)
A0 --> B1
B0 --> B1
C0 --> B1
B0 --> C1
C0 --> C1
D0 --> C1
C0 --> D1
D0 --> D1
E0 --> D1
B1 --> C2
C1 --> C2
D1 --> C2
```
Described in words, this graph shows that before the first layer of the
convolution, the "receptive field" centered at each token consists only
of that same token. That is to say, that we have a receptive field of 1.
The first layer of the convolution adds one neighboring token on either
side to the receptive field. Since this is done on both sides, the
receptive field increases by 2, giving the first layer a receptive field
of 3. The second layer of the convolutions adds an _additional_
neighboring token on either side to the receptive field, giving a final
receptive field of 5.
However, this doesn't match the formula currently given in the docs,
which read:
> The receptive field of the CNN will be
> `depth * (window_size * 2 + 1)`, so a 4-layer network with a window
> size of `2` will be sensitive to 20 words at a time.
Substituting in our depth of 2 and window size of 1, this formula gives
us a receptive field of:
```
depth * (window_size * 2 + 1)
= 2 * (1 * 2 + 1)
= 2 * (2 + 1)
= 2 * 3
= 6
```
This not only doesn't match our computations from above, it's also an
even number! This is suspicious, since the receptive field is supposed
to be centered on a token, and not between tokens. Generally, this
formula results in an even number for any even value of `depth`.
The error in this formula is that the adjustment for the center token
is multiplied by the depth, when it should occur only once. The
corrected formula, `depth * window_size * 2 + 1`, gives the correct
value for our small example from above:
```
depth * window_size * 2 + 1
= 2 * 1 * 2 + 1
= 4 + 1
= 5
```
These changes update the docs to correct the receptive field formula and
the example receptive field size.
There was a mistake in the regex pattern which caused not matching all the desired tokens. The problem was that when we use r string literal prefix to suppose a raw text, we should not use two backslashes to demonstrate a backslash.
* Support registered vectors
* Format
* Auto-fill [nlp] on load from config and from bytes/disk
* Only auto-fill [nlp]
* Undo all changes to Language.from_disk
* Expand BaseVectors
These methods are needed in various places for training and vector
similarity.
* isort
* More linting
* Only fill [nlp.vectors]
* Update spacy/vocab.pyx
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Revert changes to test related to auto-filling [nlp]
* Add vectors registry
* Rephrase error about vocab methods for vectors
* Switch to dummy implementation for BaseVectors.to_ops
* Add initial draft of docs
* Remove example from BaseVectors docs
* Apply suggestions from code review
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Update website/docs/api/basevectors.mdx
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Fix type and lint bpemb example
* Update website/docs/api/basevectors.mdx
---------
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Add support for multiple code files to all relevant commands
Prior to this, only the package command supported multiple code files.
* Update docs
* Add debug data test, plus generic fixtures
One tricky thing here: it's tempting to create the config by creating a
pipeline in code, but that requires declaring the custom components
here. However the CliRunner appears to be run in the same process or
otherwise have access to our registry, so it works even without any
code arguments. So it's necessary to avoid declaring the components in
the tests.
* Add debug config test and restructure
The code argument imports the provided file. If it adds item to the
registry, that affects global state, which CliRunner doesn't isolate.
Since there's no standard way to remove things from the registry, this
instead uses subprocess.run to run commands.
* Use a more generic, parametrized test
* Add output arg for assemble and pretrain
Assemble and pretrain require an output argument. This commit adds
assemble testing, but not pretrain, as that requires an actual trainable
component, which is not currently in the test config.
* Add evaluate test and some cleanup
* Mark tests as slow
* Revert argument name change
* Apply suggestions from code review
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Format API CLI docs
* isort
* Fix imports in tests
* isort
* Undo changes to package CLI help
* Fix python executable and lang code in test
* Fix executable in another test
---------
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
Co-authored-by: Raphael Mitsch <r.mitsch@outlook.com>
* Recommend lookups tables from URLs or other loaders
Shift away from the `lookups` extra (which isn't removed, just no longer
mentioned) and recommend loading data from the `spacy-lookups-data` repo
or other sources rather than the `spacy-lookups-data` package.
If the tables can't be loaded from the `lookups` registry in the
lemmatizer, show how to specify the tables in `[initialize]` rather than
recommending the `spacy-lookups-data` package.
* Add tests for some rule-based lemmatizers
* Apply suggestions from code review
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
---------
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Add data structures to docs
* Adjusted descriptions for more consistency
* Add _optional_ flag to parameters
* Add tests and adjust optional title key in doc
* Add title to dep visualizations
* fix typo
---------
Co-authored-by: thomashacker <EdwardSchmuhl@web.de>
* Add cli for finding locations of registered func
* fixes: naming and typing
* isort
* update naming
* remove to find-function
* remove file:// bit
* use registry name if given and exit gracefully if a registry was not found
* clean up failure msg
* specify registry_name options
* mypy fixes
* return location for internal usage
* add documentation
* more mypy fixes
* clean up example
* add section to menu
* add tests
---------
Co-authored-by: svlandeg <svlandeg@github.com>
* modified: spacy/language.py
- corrected typo in docstring for :method:`Language.replace_listeners`
- added noqa comment on unused local variable assignment in :method:`Language.from_config` as I wasn't sure if it should be unassigned
modified: website/docs/api/language.mdx
- corrected typo in `Language.replace_listeners` markdown
* modified: spacy/language.py
- removed noqa comment
---------
Co-authored-by: Ian Thompson <ian.thompson@hrblock.com>
* Literal True for first/last options
* add test case
* update docs
* remove old redundant test case
* black formatting
* use Optional typing in docstrings
Co-authored-by: Raphael Mitsch <r.mitsch@outlook.com>
---------
Co-authored-by: Raphael Mitsch <r.mitsch@outlook.com>
* Support custom token/lexeme attribute for vectors
* Fix imports
* Back off to ORTH without Vectors.attr
* Fallback if vectors.attr doesn't exist
* Update docs
* span finder integrated into spacy from experimental
* black
* isort
* black
* default spankey constant
* black
* Update spacy/pipeline/spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* rename
* rename
* max_length and min_length as Optional[int] and strict checking
* black
* mypy fix for integer type infinity
* revert line order
* implement all comparison operators for inf int
* avoid two for loops over all docs by not precomputing
* interleave thresholding with span creation
* black
* revert to not interleaving (relized its faster)
* black
* Update spacy/errors.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* update dosctring
* enforce that the gold and predicted documents have the same text
* new error for ensuring reference and predicted texts are the same
* remove todo
* adjust test
* black
* handle misaligned tokenization
* return correct variable
* failing overfit test
* only use a single spans_key like in spancat
* black
* remove debug lines
* typo
* remove comment
* remove near duplicate reduntant method
* use the 'spans_key' variable name everywhere
* Update spacy/pipeline/span_finder.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* flaky test fix suggestion, hand set bias terms
* only test suggester and test result exhaustively
* make it clear that the span_finder_suggester is more general (not specific to span_finder)
* Update spacy/tests/pipeline/test_span_finder.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Apply suggestions from code review
* remove question comment
* move preset_spans_suggester test to spancat tests
* Add docs and unify default configs for spancat and span finder
* Add `allow_overlap=True` to span finder scorer
* Fix offset bug in set_annotations
* Ignore labels in span finder scorer
* Format
* Add span_finder to quickstart template
* Move settings to self.cfg, store min/max unset as None
* Remove debugging
* Update docstrings and docs
* Update spacy/pipeline/span_finder.py
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Fix imports
---------
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Add scorer option to return per-component scores
Add `per_component` option to `Language.evaluate` and `Scorer.score` to
return scores keyed by `tokenizer` (hard-coded) or by component name.
Add option to `evaluate` CLI to score by component. Per-component scores
can only be saved to JSON.
* Update help text and messages
* Add distillation initialization and loop
* Fix up configuration keys
* Add docstring
* Type annotations
* init_nlp_distill -> init_nlp_student
* Do not resolve dot name distill corpus in initialization
(Since we don't use it.)
* student: do not request use of optimizer in student pipe
We apply finish up the updates once in the training loop instead.
Also add the necessary logic to `Language.distill` to mirror
`Language.update`.
* Correctly determine sort key in subdivide_batch
* Fix _distill_loop docstring wrt. stopping condition
* _distill_loop: fix distill_data docstring
Make similar changes in train_while_improving, since it also had
incorrect types and missing type annotations.
* Move `set_{gpu_allocator,seed}_from_config` to spacy.util
* Update Language.update docs for the sgd argument
* Type annotation
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
---------
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
* Add default to MorphAnalysis.get
Similar to `dict`, allow a `default` option for `MorphAnalysis.get` for
the user to provide a default return value if the field is not found.
The default return value remains `[]`, which is not the same as
`dict.get`, but is already established as this method's default return
value with the return type `List[str]`. However the new `default` option
does not enforce that the user-provided default is actually `List[str]`.
* Restore test case
Replace `progress_bar = "all_steps"` with `progress_bar = "eval"`, which is consistent with the default behavior for `spacy.ConsoleLogger.v1` and `spacy.ConsoleLogger.v2`.
* Avoid `TrainablePipe.finish_update` getting called twice during training
PR #12136 fixed an issue where the tok2vec pipe was updated before
gradient were accumulated. However, it introduced a new bug that cause
`finish_update` to be called twice when using the training loop. This
causes a fairly large slowdown.
The `Language.update` method accepts the `sgd` argument for passing an
optimizer. This argument has three possible values:
- `Optimizer`: use the given optimizer to finish pipe updates.
- `None`: use a default optimizer to finish pipe updates.
- `False`: do not finish pipe updates.
However, the latter option was not documented and not valid with the
existing type of `sgd`. I assumed that this was a remnant of earlier
spaCy versions and removed handling of `False`.
However, with that change, we are passing `None` to `Language.update`.
As a result, we were calling `finish_update` in both `Language.update`
and in the training loop after all subbatches are processed.
This change restores proper handling/use of `False`. Moreover, the role
of `False` is now documented and added to the type to avoid future
accidents.
* Fix typo
* Document defaults for `Language.update`
* Convert Candidate from Cython to Python class.
* Format.
* Fix .entity_ typo in _add_activations() usage.
* Change type for mentions to look up entity candidates for to SpanGroup from Iterable[Span].
* Update docs.
* Update spacy/kb/candidate.py
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Update doc string of BaseCandidate.__init__().
* Update spacy/kb/candidate.py
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Rename Candidate to InMemoryCandidate, BaseCandidate to Candidate.
* Adjust Candidate to support and mandate numerical entity IDs.
* Format.
* Fix docstring and docs.
* Update website/docs/api/kb.mdx
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Rename alias -> mention.
* Refactor Candidate attribute names. Update docs and tests accordingly.
* Refacor Candidate attributes and their usage.
* Format.
* Fix mypy error.
* Update error code in line with v4 convention.
* Reverse erroneous changes during merge.
* Update return type in EL tests.
* Re-add Candidate to setup.py.
* Format updated docs.
---------
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Clean up Vocab constructor
* Change effective type of `strings` from `Iterable[str]` to `Optional[StringStore]`
* Don't automatically add strings to vocab
* Change default values to `None`
* Remove `**deprecated_kwargs`
* Format
* [wip] Update
* [wip] Update
* Add initial port
* [wip] Update
* Fix all imports
* Add spancat_exclusive to pipeline
* [WIP] Update
* [ci skip] Add breakpoint for debugging
* Use spacy.SpanCategorizer.v1 as default archi
* Update spacy/pipeline/spancat_exclusive.py
Co-authored-by: kadarakos <kadar.akos@gmail.com>
* [ci skip] Small updates
* Use Softmax v2 directly from thinc
* Cache the label map
* Fix mypy errors
However, I ignored line 370 because it opened up a bunch of type errors
that might be trickier to solve and might lead to a more complicated
codebase.
* avoid multiplication with 1.0
Co-authored-by: kadarakos <kadar.akos@gmail.com>
* Update spacy/pipeline/spancat_exclusive.py
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Update component versions to v2
* Add scorer to docstring
* Add _n_labels property to SpanCategorizer
Instead of using len(self.labels) in initialize() I am using a private
property self._n_labels. This achieves implementation parity and allows
me to delete the whole initialize() method for spancat_exclusive (since
it's now the same with spancat).
* Inherit from SpanCat instead of TrainablePipe
This commit changes the inheritance structure of Exclusive_Spancat,
now it's inheriting from SpanCategorizer than TrainablePipe. This
allows me to remove duplicate methods that are already present in
the parent function.
* Revert documentation link to spancat
* Fix init call for exclusive spancat
* Update spacy/pipeline/spancat_exclusive.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Import Suggester from spancat
* Include zero_init.v1 for spancat
* Implement _allow_extra_label to use _n_labels
To ensure that spancat / spancat_exclusive cannot be resized after
initialization, I inherited the _allow_extra_label() method from
spacy/pipeline/trainable_pipe.pyx and used self._n_labels instead
of len(self.labels) for checking.
I think that changing it locally is a better solution rather than
forcing each class that inherits TrainablePipe to use the self._n_labels
attribute.
Also note that I turned-off black formatting in this block of code
because it reads better without the overhang.
* Extend existing tests to spancat_exclusive
In this commit, I extended the existing tests for spancat to include
spancat_exclusive. I parametrized the test functions with 'name'
(similar var name with textcat and textcat_multilabel) for each
applicable test.
TODO: Add overfitting tests for spancat_exclusive
* Update documentation for spancat
* Turn on formatting for allow_extra_label
* Remove initializers in default config
* Use DEFAULT_EXCL_SPANCAT_MODEL
I also renamed spancat_exclusive_default_config into
spancat_excl_default_config because black does some not pretty
formatting changes.
* Update documentation
Update grammar and usage
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Clarify docstring for Exclusive_SpanCategorizer
* Remove mypy ignore and typecast labels to list
* Fix documentation API
* Use a single variable for tests
* Update defaults for number of rows
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Put back initializers in spancat config
Whenever I remove model.scorer.init_w and model.scorer.init_b,
I encounter an error in the test:
SystemError: <method '__getitem__' of 'dict' objects> returned a result
with an error set.
My Thinc version is 8.1.5, but I can't seem to check what's causing the
error.
* Update spancat_exclusive docstring
* Remove init_W and init_B parameters
This commit is expected to fail until the new Thinc release.
* Require thinc>=8.1.6 for serializable Softmax defaults
* Handle zero suggestions to make tests pass
I'm not sure if this is the most elegant solution. But what should
happen is that the _make_span_group function MUST return an empty
SpanGroup if there are no suggestions.
The error happens when the 'scores' variable is empty. We cannot
get the 'predicted' and other downstream vars.
* Better approach for handling zero suggestions
* Update website/docs/api/spancategorizer.md
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update spancategorizer headers
* Apply suggestions from code review
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Add default value in negative_weight in docs
* Add default value in allow_overlap in docs
* Update how spancat_exclusive is constructed
In this commit, I added the following:
- Put the default values of negative_weight and allow_overlap
in the default_config dictionary.
- Rename make_spancat -> make_exclusive_spancat
* Run prettier on spancategorizer.mdx
* Change exactly one -> at most one
* Add suggester documentation in Exclusive_SpanCategorizer
* Add suggester to spancat docstrings
* merge multilabel and singlelabel spancat
* rename spancat_exclusive to singlelable
* wire up different make_spangroups for single and multilabel
* black
* black
* add docstrings
* more docstring and fix negative_label
* don't rely on default arguments
* black
* remove spancat exclusive
* replace single_label with add_negative_label and adjust inference
* mypy
* logical bug in configuration check
* add spans.attrs[scores]
* single label make_spangroup test
* bugfix
* black
* tests for make_span_group with negative labels
* refactor make_span_group
* black
* Update spacy/tests/pipeline/test_spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* remove duplicate declaration
* Update spacy/pipeline/spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* raise error instead of just print
* make label mapper private
* update docs
* run prettier
* Update website/docs/api/spancategorizer.mdx
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update website/docs/api/spancategorizer.mdx
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update spacy/pipeline/spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update spacy/pipeline/spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update spacy/pipeline/spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update spacy/pipeline/spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* don't keep recomputing self._label_map for each span
* typo in docs
* Intervals to private and document 'name' param
* Update spacy/pipeline/spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update spacy/pipeline/spancat.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* add Tag to new features
* replace tags
* revert
* revert
* revert
* revert
* Update website/docs/api/spancategorizer.mdx
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update website/docs/api/spancategorizer.mdx
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* prettier
* Fix merge
* Update website/docs/api/spancategorizer.mdx
* remove references to 'single_label'
* remove old paragraph
* Add spancat_singlelabel to config template
* Format
* Extend init config tests
---------
Co-authored-by: kadarakos <kadar.akos@gmail.com>
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Make empty_kb() configurable.
* Format.
* Update docs.
* Be more specific in KB serialization test.
* Update KB serialization tests. Update docs.
* Remove doc update for batched candidate generation.
* Fix serialization of subclassed KB in tests.
* Format.
* Update docstring.
* Update docstring.
* Switch from pickle to json for custom field serialization.
* Add immediate left/right child/parent dependency relations
* Add tests for new REL_OPs: `>+`, `>-`, `<+`, and `<-`.
---------
Co-authored-by: Tan Long <tanloong@foxmail.com>
* Fix FUZZY operator definition
The default length of the FUZZY operator is 2 and not 3.
* adjust edit distance in matcher usage docs too
---------
Co-authored-by: svlandeg <svlandeg@github.com>
* Remove backwards-compatible overwrite from Entity Linker
This also adds a docstring about overwrite, since it wasn't present.
* Fix docstring
* Remove backward compat settings in Morphologizer
This also needed a docstring added.
For this component it's less clear what the right overwrite settings
are.
* Remove backward compat from sentencizer
This was simple
* Remove backward compat from senter
Another simple one
* Remove backward compat setting from tagger
* Add docstrings
* Update spacy/pipeline/morphologizer.pyx
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Update docs
---------
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Init
* fix tests
* Update spacy/errors.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Fix test_blank_languages
* Rename xx to mul in docs
* Format _util with black
* prettier formatting
---------
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Add `Language.distill`
This method is the distillation counterpart of `Language.update`. It
takes a teacher `Language` instance and distills the student pipes on
the teacher pipes.
* Apply suggestions from code review
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
* Clarify that how Example is used in distillation
* Update transition parser distill docstring for examples argument
* Pass optimizer to `TrainablePipe.distill`
* Annotate pipe before update
As discussed internally, we want to let a pipe annotate before doing an
update with gold/silver data. Otherwise, the output may be (too)
informed by the gold/silver data.
* Rename `component_map` to `student_to_teacher`
* Better synopsis in `Language.distill` docstring
* `name` -> `student_name`
* Fix labels type in docstring
* Mark distill test as slow
* Fix `student_to_teacher` type in docs
---------
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
* Add span_id to Span.char_span, update Doc/Span.char_span docs
`Span.char_span(id=)` should be removed in the future.
* Also use Union[int, str] in Doc docstring
* Add `spacy.PlainTextCorpusReader.v1`
This is a corpus reader that reads plain text corpora with the following
format:
- UTF-8 encoding
- One line per document.
- Blank lines are ignored.
It is useful for applications where we deal with very large corpora,
such as distillation, and don't want to deal with the space overhead of
serialized formats. Additionally, many large corpora already use such
a text format, keeping the necessary preprocessing to a minimum.
* Update spacy/training/corpus.py
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* docs: add version to `PlainTextCorpus`
* Add docstring to registry function
* Add plain text corpus tests
* Only strip newline/carriage return
* Add return type _string_to_tmp_file helper
* Use a temporary directory in place of file name
Different OS auto delete/sharing semantics are just wonky.
* This will be new in 3.5.1 (rather than 4)
* Test improvements from code review
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Don't pass mem pool to new lexeme function
* Remove unused mem from function args
Two methods calling _new_lexeme, get and get_by_orth, took mem arguments
just to call the internal method. That's no longer necessary, so this
cleans it up.
* prettier formatting
* Remove more unused mem args
* skeleton
* Fill in non-CLI details from release notes draft
* Add TODO for fuzzy matching
* Website updates for v3-5 draft
* Fill in usage examples
* Add fuzzy matching to intro
* Fix fuzzy examples
* Shell example formatting
* Fix typo
* Format
* Remove trailing periods in internal list
* Update
* Fix spacing for nested lists
* Update InMemoryLookupKB link
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
Co-authored-by: Ines Montani <ines@ines.io>
* API docs: Rename kb_in_memory to inmemorylookupkb, add to sidebar
* adjust to mdx
* linkout to InMemoryLookupKB at first occurrence in kb.mdx
* fix links to docs
* revert Azure trigger setting (I'll make a separate PR)
Co-authored-by: svlandeg <svlandeg@github.com>
* Fix batching regression
Some time ago, the spaCy v4 branch switched to the new Thinc v9
schedule. However, this introduced an error in how batching is handed.
In the PR, the batchers were changed to keep track of their step,
so that the step can be passed to the schedule. However, the issue
is that the training loop repeatedly calls the batching functions
(rather than using an infinite generator/iterator). So, the step and
therefore the schedule would be reset each epoch. Before the schedule
switch we didn't have this issue, because the old schedules were
stateful.
This PR fixes this issue by reverting the batching functions to use
a (stateful) generator. Their registry functions do accept a `Schedule`
and we convert `Schedule`s to generators.
* Update batcher docs
* Docstring fixes
* Make minibatch take iterables again as well
* Bump thinc requirement to 9.0.0.dev2
* Use type declaration
* Convert another comment into a proper type declaration
* Try to fix doc.copy
* Set dev version
* Make vocab always own lexemes
* Change version
* Add SpanGroups.copy method
* Fix set_annotations during Parser.update
* Fix dict proxy copy
* Upd version
* Fix copying SpanGroups
* Fix set_annotations in parser.update
* Fix parser set_annotations during update
* Revert "Fix parser set_annotations during update"
This reverts commit eb138c89ed.
* Revert "Fix set_annotations in parser.update"
This reverts commit c6df0eafd0.
* Fix set_annotations during parser update
* Inc version
* Handle final states in get_oracle_sequence
* Inc version
* Try to fix parser training
* Inc version
* Fix
* Inc version
* Fix parser oracle
* Inc version
* Inc version
* Fix transition has_gold
* Inc version
* Try to use real histories, not oracle
* Inc version
* Upd parser
* Inc version
* WIP on rewrite parser
* WIP refactor parser
* New progress on parser model refactor
* Prepare to remove parser_model.pyx
* Convert parser from cdef class
* Delete spacy.ml.parser_model
* Delete _precomputable_affine module
* Wire up tb_framework to new parser model
* Wire up parser model
* Uncython ner.pyx and dep_parser.pyx
* Uncython
* Work on parser model
* Support unseen_classes in parser model
* Support unseen classes in parser
* Cleaner handling of unseen classes
* Work through tests
* Keep working through errors
* Keep working through errors
* Work on parser. 15 tests failing
* Xfail beam stuff. 9 failures
* More xfail. 7 failures
* Xfail. 6 failures
* cleanup
* formatting
* fixes
* pass nO through
* Fix empty doc in update
* Hackishly fix resizing. 3 failures
* Fix redundant test. 2 failures
* Add reference version
* black formatting
* Get tests passing with reference implementation
* Fix missing prints
* Add missing file
* Improve indexing on reference implementation
* Get non-reference forward func working
* Start rigging beam back up
* removing redundant tests, cf #8106
* black formatting
* temporarily xfailing issue 4314
* make flake8 happy again
* mypy fixes
* ensure labels are added upon predict
* cleanup remnants from merge conflicts
* Improve unseen label masking
Two changes to speed up masking by ~10%:
- Use a bool array rather than an array of float32.
- Let the mask indicate whether a label was seen, rather than
unseen. The mask is most frequently used to index scores for
seen labels. However, since the mask marked unseen labels,
this required computing an intermittent flipped mask.
* Write moves costs directly into numpy array (#10163)
This avoids elementwise indexing and the allocation of an additional
array.
Gives a ~15% speed improvement when using batch_by_sequence with size
32.
* Temporarily disable ner and rehearse tests
Until rehearse is implemented again in the refactored parser.
* Fix loss serialization issue (#10600)
* Fix loss serialization issue
Serialization of a model fails with:
TypeError: array(738.3855, dtype=float32) is not JSON serializable
Fix this using float conversion.
* Disable CI steps that require spacy.TransitionBasedParser.v2
After finishing the refactor, TransitionBasedParser.v2 should be
provided for backwards compat.
* Add back support for beam parsing to the refactored parser (#10633)
* Add back support for beam parsing
Beam parsing was already implemented as part of the `BeamBatch` class.
This change makes its counterpart `GreedyBatch`. Both classes are hooked
up in `TransitionModel`, selecting `GreedyBatch` when the beam size is
one, or `BeamBatch` otherwise.
* Use kwarg for beam width
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Avoid implicit default for beam_width and beam_density
* Parser.{beam,greedy}_parse: ensure labels are added
* Remove 'deprecated' comments
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Parser `StateC` optimizations (#10746)
* `StateC`: Optimizations
Avoid GIL acquisition in `__init__`
Increase default buffer capacities on init
Reduce C++ exception overhead
* Fix typo
* Replace `set::count` with `set::find`
* Add exception attribute to c'tor
* Remove unused import
* Use a power-of-two value for initial capacity
Use default-insert to init `_heads` and `_unshiftable`
* Merge `cdef` variable declarations and assignments
* Vectorize `example.get_aligned_parses` (#10789)
* `example`: Vectorize `get_aligned_parse`
Rename `numpy` import
* Convert aligned array to lists before returning
* Revert import renaming
* Elide slice arguments when selecting the entire range
* Tagger/morphologizer alignment performance optimizations (#10798)
* `example`: Unwrap `numpy` scalar arrays before passing them to `StringStore.__getitem__`
* `AlignmentArray`: Use native list as staging buffer for offset calculation
* `example`: Vectorize `get_aligned`
* Hoist inner functions out of `get_aligned`
* Replace inline `if..else` clause in assignment statement
* `AlignmentArray`: Use raw indexing into offset and data `numpy` arrays
* `example`: Replace array unique value check with `groupby`
* `example`: Correctly exclude tokens with no alignment in `_get_aligned_vectorized`
Simplify `_get_aligned_non_vectorized`
* `util`: Update `all_equal` docstring
* Explicitly use `int32_t*`
* Restore C CPU inference in the refactored parser (#10747)
* Bring back the C parsing model
The C parsing model is used for CPU inference and is still faster for
CPU inference than the forward pass of the Thinc model.
* Use C sgemm provided by the Ops implementation
* Make tb_framework module Cython, merge in C forward implementation
* TransitionModel: raise in backprop returned from forward_cpu
* Re-enable greedy parse test
* Return transition scores when forward_cpu is used
* Apply suggestions from code review
Import `Model` from `thinc.api`
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Use relative imports in tb_framework
* Don't assume a default for beam_width
* We don't have a direct dependency on BLIS anymore
* Rename forwards to _forward_{fallback,greedy_cpu}
* Require thinc >=8.1.0,<8.2.0
* tb_framework: clean up imports
* Fix return type of _get_seen_mask
* Move up _forward_greedy_cpu
* Style fixes.
* Lower thinc lowerbound to 8.1.0.dev0
* Formatting fix
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
* Reimplement parser rehearsal function (#10878)
* Reimplement parser rehearsal function
Before the parser refactor, rehearsal was driven by a loop in the
`rehearse` method itself. For each parsing step, the loops would:
1. Get the predictions of the teacher.
2. Get the predictions and backprop function of the student.
3. Compute the loss and backprop into the student.
4. Move the teacher and student forward with the predictions of
the student.
In the refactored parser, we cannot perform search stepwise rehearsal
anymore, since the model now predicts all parsing steps at once.
Therefore, rehearsal is performed in the following steps:
1. Get the predictions of all parsing steps from the student, along
with its backprop function.
2. Get the predictions from the teacher, but use the predictions of
the student to advance the parser while doing so.
3. Compute the loss and backprop into the student.
To support the second step a new method, `advance_with_actions` is
added to `GreedyBatch`, which performs the provided parsing steps.
* tb_framework: wrap upper_W and upper_b in Linear
Thinc's Optimizer cannot handle resizing of existing parameters. Until
it does, we work around this by wrapping the weights/biases of the upper
layer of the parser model in Linear. When the upper layer is resized, we
copy over the existing parameters into a new Linear instance. This does
not trigger an error in Optimizer, because it sees the resized layer as
a new set of parameters.
* Add test for TransitionSystem.apply_actions
* Better FIXME marker
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
* Fixes from Madeesh
* Apply suggestions from Sofie
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Remove useless assignment
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
* Rename some identifiers in the parser refactor (#10935)
* Rename _parseC to _parse_batch
* tb_framework: prefix many auxiliary functions with underscore
To clearly state the intent that they are private.
* Rename `lower` to `hidden`, `upper` to `output`
* Parser slow test fixup
We don't have TransitionBasedParser.{v1,v2} until we bring it back as a
legacy option.
* Remove last vestiges of PrecomputableAffine
This does not exist anymore as a separate layer.
* ner: re-enable sentence boundary checks
* Re-enable test that works now.
* test_ner: make loss test more strict again
* Remove commented line
* Re-enable some more beam parser tests
* Remove unused _forward_reference function
* Update for CBlas changes in Thinc 8.1.0.dev2
Bump thinc dependency to 8.1.0.dev3.
* Remove references to spacy.TransitionBasedParser.{v1,v2}
Since they will not be offered starting with spaCy v4.
* `tb_framework`: Replace references to `thinc.backends.linalg` with `CBlas`
* dont use get_array_module (#11056) (#11293)
Co-authored-by: kadarakos <kadar.akos@gmail.com>
* Move `thinc.extra.search` to `spacy.pipeline._parser_internals` (#11317)
* `search`: Move from `thinc.extra.search`
Fix NPE in `Beam.__dealloc__`
* `pytest`: Add support for executing Cython tests
Move `search` tests from thinc and patch them to run with `pytest`
* `mypy` fix
* Update comment
* `conftest`: Expose `register_cython_tests`
* Remove unused import
* Move `argmax` impls to new `_parser_utils` Cython module (#11410)
* Parser does not have to be a cdef class anymore
This also fixes validation of the initialization schema.
* Add back spacy.TransitionBasedParser.v2
* Fix a rename that was missed in #10878.
So that rehearsal tests pass.
* Remove module from setup.py that got added during the merge
* Bring back support for `update_with_oracle_cut_size` (#12086)
* Bring back support for `update_with_oracle_cut_size`
This option was available in the pre-refactor parser, but was never
implemented in the refactored parser. This option cuts transition
sequences that are longer than `update_with_oracle_cut` size into
separate sequences that have at most `update_with_oracle_cut`
transitions. The oracle (gold standard) transition sequence is used to
determine the cuts and the initial states for the additional sequences.
Applying this cut makes the batches more homogeneous in the transition
sequence lengths, making forward passes (and as a consequence training)
much faster.
Training time 1000 steps on de_core_news_lg:
- Before this change: 149s
- After this change: 68s
- Pre-refactor parser: 81s
* Fix a rename that was missed in #10878.
So that rehearsal tests pass.
* Apply suggestions from @shadeMe
* Use chained conditional
* Test with update_with_oracle_cut_size={0, 1, 5, 100}
And fix a git that occurs with a cut size of 1.
* Fix up some merge fall out
* Update parser distillation for the refactor
In the old parser, we'd iterate over the transitions in the distill
function and compute the loss/gradients on the go. In the refactored
parser, we first let the student model parse the inputs. Then we'll let
the teacher compute the transition probabilities of the states in the
student's transition sequence. We can then compute the gradients of the
student given the teacher.
* Add back spacy.TransitionBasedParser.v1 references
- Accordion in the architecture docs.
- Test in test_parse, but disabled until we have a spacy-legacy release.
Co-authored-by: Matthew Honnibal <honnibal+gh@gmail.com>
Co-authored-by: svlandeg <svlandeg@github.com>
Co-authored-by: Sofie Van Landeghem <svlandeg@users.noreply.github.com>
Co-authored-by: Madeesh Kannan <shadeMe@users.noreply.github.com>
Co-authored-by: Adriane Boyd <adrianeboyd@gmail.com>
Co-authored-by: kadarakos <kadar.akos@gmail.com>
