spaCy/vectors.md at 7d50804644eccb3dabd421ce413d1d2f748814a5

19 KiB

Raw Blame History

title	teaser	tag	source	new
Vectors	Store, save and load word vectors	class	spacy/vectors.pyx	2

Vectors data is kept in the Vectors.data attribute, which should be an instance of numpy.ndarray (for CPU vectors) or cupy.ndarray (for GPU vectors).

As of spaCy v3.2, Vectors supports two types of vector tables:

default: A standard vector table (as in spaCy v3.1 and earlier) where each key is mapped to one row in the vector table. Multiple keys can be mapped to the same vector, and not all of the rows in the table need to be assigned – so vectors.n_keys may be greater or smaller than vectors.shape[0].
floret: Only supports vectors trained with floret, an extended version of fastText that produces compact vector tables by combining fastText's subword ngrams with Bloom embeddings. The compact tables are similar to the HashEmbed embeddings already used in many spaCy components. Each word is represented as the sum of one or more rows as determined by the settings related to character ngrams and the hash table.

Vectors.init

Create a new vector store. With the default mode, you can set the vector values and keys directly on initialization, or supply a shape keyword argument to create an empty table you can add vectors to later. In floret mode, the complete vector data and settings must be provided on initialization and cannot be modified later.

Example

from spacy.vectors import Vectors

empty_vectors = Vectors(shape=(10000, 300))

data = numpy.zeros((3, 300), dtype='f')
keys = ["cat", "dog", "rat"]
vectors = Vectors(data=data, keys=keys)

Name	Description
keyword-only
`strings`	The string store. A new string store is created if one is not provided. Defaults to `None`. ~~Optional[StringStore]~~
`shape`	Size of the table as `(n_entries, n_columns)`, the number of entries and number of columns. Not required if you're initializing the object with `data` and `keys`. ~~Tuple[int, int]~~
`data`	The vector data. ~~numpy.ndarray[ndim=1, dtype=float32]~~
`keys`	A sequence of keys aligned with the data. ~~Iterable[Union[str, int]]~~
`name`	A name to identify the vectors table. ~~str~~
`mode` 3.2	Vectors mode: `"default"` or `"floret"` (default: `"default"`). ~~str~~
`minn` 3.2	The floret char ngram minn (default: `0`). ~~int~~
`maxn` 3.2	The floret char ngram maxn (default: `0`). ~~int~~
`hash_count` 3.2	The floret hash count. Supported values: 1--4 (default: `1`). ~~int~~
`hash_seed` 3.2	The floret hash seed (default: `0`). ~~int~~
`bow` 3.2	The floret BOW string (default: `"<"`). ~~str~~
`eow` 3.2	The floret EOW string (default: `">"`). ~~str~~

Vectors.getitem

Get a vector by key. If the key is not found in the table, a KeyError is raised.

Example

cat_id = nlp.vocab.strings["cat"]
cat_vector = nlp.vocab.vectors[cat_id]
assert cat_vector == nlp.vocab["cat"].vector

Name	Description
`key`	The key to get the vector for. ~~Union[int, str]~~
RETURNS	The vector for the key. ~~numpy.ndarray[ndim=1, dtype=float32]~~

Vectors.setitem

Set a vector for the given key. Not supported for floret mode.

Example

cat_id = nlp.vocab.strings["cat"]
vector = numpy.random.uniform(-1, 1, (300,))
nlp.vocab.vectors[cat_id] = vector

Name	Description
`key`	The key to set the vector for. ~~int~~
`vector`	The vector to set. ~~numpy.ndarray[ndim=1, dtype=float32]~~

Vectors.iter

Iterate over the keys in the table. In floret mode, the keys table is not used.

Example

for key in nlp.vocab.vectors:
   print(key, nlp.vocab.strings[key])

Name	Description
YIELDS	A key in the table. ~~int~~

Vectors.len

Return the number of vectors in the table.

Example

vectors = Vectors(shape=(3, 300))
assert len(vectors) == 3

Name	Description
RETURNS	The number of vectors in the table. ~~int~~

Vectors.contains

Check whether a key has been mapped to a vector entry in the table. In floret mode, returns True for all keys.

Example

cat_id = nlp.vocab.strings["cat"]
nlp.vocab.vectors.add(cat_id, numpy.random.uniform(-1, 1, (300,)))
assert cat_id in vectors

Name	Description
`key`	The key to check. ~~int~~
RETURNS	Whether the key has a vector entry. ~~bool~~

Vectors.add

Add a key to the table, optionally setting a vector value as well. Keys can be mapped to an existing vector by setting row, or a new vector can be added. Not supported for floret mode.

Example

vector = numpy.random.uniform(-1, 1, (300,))
cat_id = nlp.vocab.strings["cat"]
nlp.vocab.vectors.add(cat_id, vector=vector)
nlp.vocab.vectors.add("dog", row=0)

Name	Description
`key`	The key to add. ~~Union[str, int]~~
keyword-only
`vector`	An optional vector to add for the key. ~~numpy.ndarray[ndim=1, dtype=float32]~~
`row`	An optional row number of a vector to map the key to. ~~int~~
RETURNS	The row the vector was added to. ~~int~~

Vectors.resize

Resize the underlying vectors array. If inplace=True, the memory is reallocated. This may cause other references to the data to become invalid, so only use inplace=True if you're sure that's what you want. If the number of vectors is reduced, keys mapped to rows that have been deleted are removed. These removed items are returned as a list of (key, row) tuples. Not supported for floret mode.

Example

removed = nlp.vocab.vectors.resize((10000, 300))

Name	Description
`shape`	A `(rows, dims)` tuple describing the number of rows and dimensions. ~~Tuple[int, int]~~
`inplace`	Reallocate the memory. ~~bool~~
RETURNS	The removed items as a list of `(key, row)` tuples. ~~List[Tuple[int, int]]~~

Vectors.keys

A sequence of the keys in the table. In floret mode, the keys table is not used.

Example

for key in nlp.vocab.vectors.keys():
    print(key, nlp.vocab.strings[key])

Name	Description
RETURNS	The keys. ~~Iterable[int]~~

Vectors.values

Iterate over vectors that have been assigned to at least one key. Note that some vectors may be unassigned, so the number of vectors returned may be less than the length of the vectors table. In floret mode, the keys table is not used.

Example

for vector in nlp.vocab.vectors.values():
    print(vector)

Name	Description
YIELDS	A vector in the table. ~~numpy.ndarray[ndim=1, dtype=float32]~~

Vectors.items

Iterate over (key, vector) pairs, in order. In floret mode, the keys table is empty.

Example

for key, vector in nlp.vocab.vectors.items():
   print(key, nlp.vocab.strings[key], vector)

Name	Description
YIELDS	`(key, vector)` pairs, in order. ~~Tuple[int, numpy.ndarray[ndim=1, dtype=float32]]~~

Vectors.find

Look up one or more keys by row, or vice versa. Not supported for floret mode.

Example

row = nlp.vocab.vectors.find(key="cat")
rows = nlp.vocab.vectors.find(keys=["cat", "dog"])
key = nlp.vocab.vectors.find(row=256)
keys = nlp.vocab.vectors.find(rows=[18, 256, 985])

Name	Description
keyword-only
`key`	Find the row that the given key points to. Returns int, `-1` if missing. ~~Union[str, int]~~
`keys`	Find rows that the keys point to. Returns `numpy.ndarray`. ~~Iterable[Union[str, int]]~~
`row`	Find the first key that points to the row. Returns integer. ~~int~~
`rows`	Find the keys that point to the rows. Returns `numpy.ndarray`. ~~Iterable[int]~~
RETURNS	The requested key, keys, row or rows. ~~Union[int, numpy.ndarray[ndim=1, dtype=float32]]~~

Vectors.shape

Get (rows, dims) tuples of number of rows and number of dimensions in the vector table.

Example

vectors = Vectors(shape(1, 300))
vectors.add("cat", numpy.random.uniform(-1, 1, (300,)))
rows, dims = vectors.shape
assert rows == 1
assert dims == 300

Name	Description
RETURNS	A `(rows, dims)` pair. ~~Tuple[int, int]~~

Vectors.size

The vector size, i.e. rows * dims.

Example

vectors = Vectors(shape=(500, 300))
assert vectors.size == 150000

Name	Description
RETURNS	The vector size. ~~int~~

Vectors.is_full

Whether the vectors table is full and has no slots are available for new keys. If a table is full, it can be resized using Vectors.resize. In floret mode, the table is always full and cannot be resized.

Example

vectors = Vectors(shape=(1, 300))
vectors.add("cat", numpy.random.uniform(-1, 1, (300,)))
assert vectors.is_full

Name	Description
RETURNS	Whether the vectors table is full. ~~bool~~

Vectors.n_keys

Get the number of keys in the table. Note that this is the number of all keys, not just unique vectors. If several keys are mapped to the same vectors, they will be counted individually. In floret mode, the keys table is not used.

Example

vectors = Vectors(shape=(10, 300))
assert len(vectors) == 10
assert vectors.n_keys == 0

Name	Description
RETURNS	The number of all keys in the table. ~~int~~

Vectors.most_similar

For each of the given vectors, find the n most similar entries to it by cosine. Queries are by vector. Results are returned as a (keys, best_rows, scores) tuple. If queries is large, the calculations are performed in chunks to avoid consuming too much memory. You can set the batch_size to control the size/space trade-off during the calculations. Not supported for floret mode.

Example

queries = numpy.asarray([numpy.random.uniform(-1, 1, (300,))])
most_similar = nlp.vocab.vectors.most_similar(queries, n=10)

Name	Description
`queries`	An array with one or more vectors. ~~numpy.ndarray~~
keyword-only
`batch_size`	The batch size to use. Default to `1024`. ~~int~~
`n`	The number of entries to return for each query. Defaults to `1`. ~~int~~
`sort`	Whether to sort the entries returned by score. Defaults to `True`. ~~bool~~
RETURNS	tuple

Vectors.get_batch

Get the vectors for the provided keys efficiently as a batch.

Example

words = ["cat", "dog"]
vectors = nlp.vocab.vectors.get_batch(words)

Name	Description
`keys`	The keys. ~~Iterable[Union[int, str]]~~

Vectors.to_disk

Save the current state to a directory.

Example

vectors.to_disk("/path/to/vectors")

Name	Description
`path`	A path to a directory, which will be created if it doesn't exist. Paths may be either strings or `Path`-like objects. ~~Union[str, Path]~~

Vectors.from_disk

Loads state from a directory. Modifies the object in place and returns it.

Example

vectors = Vectors(StringStore())
vectors.from_disk("/path/to/vectors")

Name	Description
`path`	A path to a directory. Paths may be either strings or `Path`-like objects. ~~Union[str, Path]~~
RETURNS	The modified `Vectors` object. ~~Vectors~~

Vectors.to_bytes

Serialize the current state to a binary string.

Example

vectors_bytes = vectors.to_bytes()

Name	Description
RETURNS	The serialized form of the `Vectors` object. ~~bytes~~

Vectors.from_bytes

Load state from a binary string.

Example

fron spacy.vectors import Vectors
vectors_bytes = vectors.to_bytes()
new_vectors = Vectors(StringStore())
new_vectors.from_bytes(vectors_bytes)

Name	Description
`data`	The data to load from. ~~bytes~~
RETURNS	The `Vectors` object. ~~Vectors~~

Attributes

Name	Description
`data`	Stored vectors data. `numpy` is used for CPU vectors, `cupy` for GPU vectors. ~~Union[numpy.ndarray[ndim=1, dtype=float32], cupy.ndarray[ndim=1, dtype=float32]]~~
`key2row`	Dictionary mapping word hashes to rows in the `Vectors.data` table. ~~Dict[int, int]~~
`keys`	Array keeping the keys in order, such that `keys[vectors.key2row[key]] == key`. ~~Union[numpy.ndarray[ndim=1, dtype=float32], cupy.ndarray[ndim=1, dtype=float32]]~~

19 KiB Raw Blame History Unescape Escape

Vectors.__init__

Example

Vectors.__getitem__

Example

Vectors.__setitem__

Example

Vectors.__iter__

Example

Vectors.__len__

Example

Vectors.__contains__

Example

Vectors.add

Example

Vectors.resize

Example

Vectors.keys

Example

Vectors.values

Example

Vectors.items

Example

Vectors.find

Example

Vectors.shape

Example

Vectors.size

Example

Vectors.is_full

Example

Vectors.n_keys

Example

Vectors.most_similar

Example

Vectors.get_batch

Example

Vectors.to_disk

Example

Vectors.from_disk

Example

Vectors.to_bytes

Example

Vectors.from_bytes

Example

Attributes

19 KiB

Raw Blame History

Vectors.init

Vectors.getitem

Vectors.setitem

Vectors.iter

Vectors.len

Vectors.contains