Google I/O is a wrap! Catch up on TensorFlow sessions View sessions


Creates an empty hash table that uses tensors as the backing store.

It uses "open addressing" with quadratic reprobing to resolve collisions.

This op creates a mutable hash table, specifying the type of its keys and values. Each value must be a scalar. Data can be inserted into the table using the insert operations. It does not support the initialization operation.

empty_key A Tensor. The key used to represent empty key buckets internally. Must not be used in insert or lookup operations.
deleted_key A Tensor. Must have the same type as empty_key.
value_dtype A tf.DType. Type of the table values.
container An optional string. Defaults to "". If non-empty, this table is placed in the given container. Otherwise, a default container is used.
shared_name An optional string. Defaults to "". If non-empty, this table is shared under the given name across multiple sessions.
use_node_name_sharing An optional bool. Defaults to False.
value_shape An optional tf.TensorShape or list of ints. Defaults to []. The shape of each value.
initial_num_buckets An optional int. Defaults to 131072. The initial number of hash table buckets. Must be a power to 2.
max_load_factor An optional float. Defaults to 0.8. The maximum ratio between number of entries and number of buckets before growing the table. Must be between 0 and 1.
name A name for the operation (optional).

A Tensor of type resource.