tf.lookup.experimental.DenseHashTable

A mutable hash table with faster lookups and higher memory usage.

Inherits From: TrackableResource

tf.lookup.experimental.DenseHashTable(
    key_dtype,
    value_dtype,
    default_value,
    empty_key,
    deleted_key,
    initial_num_buckets=None,
    name='MutableDenseHashTable',
    checkpoint=True,
    experimental_is_anonymous=False
)

Data can be inserted by calling the insert method and removed by calling the remove method. It does not support initialization via the init method.

Compared to MutableHashTable, DenseHashTable offers generally faster insert, remove and lookup operations, in exchange for a higher overall memory footprint.

It uses "open addressing" with quadratic reprobing to resolve collisions. This requires specifying two keys in the key space, empty_key and deleted_key, that can never inserted into the table.

Unlike MutableHashTable, DenseHashTable does not require additional memory for temporary tensors created during checkpointing and restore operations.

Example usage:

table = tf.lookup.experimental.DenseHashTable(
    key_dtype=tf.string,
    value_dtype=tf.int64,
    default_value=-1,
    empty_key='',
    deleted_key='$')
keys = tf.constant(['a', 'b', 'c'])
values = tf.constant([0, 1, 2], dtype=tf.int64)
table.insert(keys, values)
table.remove(tf.constant(['c']))
table.lookup(tf.constant(['a', 'b', 'c','d'])).numpy()
array([ 0,  1, -1, -1])

key_dtype the type of the key tensors.
value_dtype the type of the value tensors.
default_value The value to use if a key is missing in the table.
empty_key the key to use to represent empty buckets internally. Must not be used in insert, remove or lookup operations.
deleted_key the key to use to represent deleted buckets internally. Must not be used in insert, remove or lookup operations and be different from the empty_key.
initial_num_buckets the initial number of buckets (optional, default to 2^17=131072). Note that the default value is relatively large (~1MB), so if you are going to create many tables (likely the case when experimental_is_anonymous is True), you should set initial_num_buckets to a smaller value to reduce memory usage.
name A name for the operation (optional).
checkpoint if True, the contents of the table are saved to and restored from checkpoints. If shared_name is empty for a checkpointed table, it is shared using the table node name.
experimental_is_anonymous Whether to use anonymous mode for the table (default is False). In anonymous mode, the table resource can only be accessed via a resource handle. It can't be looked up by a name. When all resource handles pointing to that resource are gone, the resource will be deleted automatically.

ValueError If checkpoint is True and no name was specified.

key_dtype The table key dtype.
name The name of the table.
resource_handle Returns the resource handle associated with this Resource.
value_dtype The table value dtype.

Methods

erase

View source

erase(
    keys, name=None
)

Removes keys and its associated values from the table.

If a key is not present in the table, it is silently ignored.

Args
keys Keys to remove. Can be a tensor of any shape. Must match the table's key type.
name A name for the operation (optional).

Returns
The created Operation.

Raises
TypeError when keys do not match the table data types.

export

View source

export(
    name=None
)

Returns tensors of all keys and values in the table.

Args
name A name for the operation (optional).

Returns
A pair of tensors with the first tensor containing all keys and the second tensors containing all values in the table.

insert

View source

insert(
    keys, values, name=None
)

Associates keys with values.

Args
keys Keys to insert. Can be a tensor of any shape. Must match the table's key type.
values Values to be associated with keys. Must be a tensor of the same shape as keys and match the table's value type.
name A name for the operation (optional).

Returns
The created Operation.

Raises
TypeError when keys or values doesn't match the table data types.

insert_or_assign

View source

insert_or_assign(
    keys, values, name=None
)

Associates keys with values.

Args
keys Keys to insert. Can be a tensor of any shape. Must match the table's key type.
values Values to be associated with keys. Must be a tensor of the same shape as keys and match the table's value type.
name A name for the operation (optional).

Returns
The created Operation.

Raises
TypeError when keys or values doesn't match the table data types.

lookup

View source

lookup(
    keys, name=None
)

Looks up keys in a table, outputs the corresponding values.

The default_value is used for keys not present in the table.

Args
keys Keys to look up. Can be a tensor of any shape. Must match the table's key_dtype.
name A name for the operation (optional).

Returns
A tensor containing the values in the same shape as keys using the table's value type.

Raises
TypeError when keys do not match the table data types.

remove

View source

remove(
    keys, name=None
)

Removes keys and its associated values from the table.

If a key is not present in the table, it is silently ignored.

Args
keys Keys to remove. Can be a tensor of any shape. Must match the table's key type.
name A name for the operation (optional).

Returns
The created Operation.

Raises
TypeError when keys do not match the table data types.

size

View source

size(
    name=None
)

Compute the number of elements in this table.

Args
name A name for the operation (optional).

Returns
A scalar tensor containing the number of elements in this table.

__getitem__

View source

__getitem__(
    keys
)

Looks up keys in a table, outputs the corresponding values.