View source on GitHub
|
Export dynamic_embedding APIs.
keras module
math module: math operations.
shadow_ops module: Dynamic Embedding is designed for Large-scale Sparse Weights Training.
class CuckooHashTable: A generic mutable hash table implementation.
class CuckooHashTableCreator: A generic KV table creator.
class FrequencyRestrictPolicy: A derived policy to eliminate features in variable follow the
class GraphKeys: (Deprecated) extended standard names related to
class ModelMode: The global config of model modes.
class HkvHashTable: A generic mutable hash table implementation.
class HkvHashTableConfig: HkvHashTableConfig config init_capacity, max_capacity, max_hbm_for_values of HkvHashTable
class HkvHashTableCreator: A generic KV table creator.
class RedisTable: A generic mutable hash table implementation.
class RedisTableConfig: RedisTableConfig config json file for connecting Redis service and
class RedisTableCreator: RedisTableCreator will create a object to pass itself to the others classes
class RestrictPolicy: Base class of restrict policies. Never use this class directly, but
class TimestampRestrictPolicy: A derived policy to eliminate features in variable follow the
class TrainableWrapper: This class is a trainable wrapper of Dynamic Embedding,
class Variable: A Distributed version of HashTable(reference from lookup_ops.MutableHashTable)
DynamicEmbeddingOptimizer(...): An optimizer wrapper to make any TensorFlow optimizer capable of training
embedding_lookup(...): Provides a dynamic version of embedding_lookup
embedding_lookup_sparse(...): Provides a dynamic version of embedding_lookup_sparse
embedding_lookup_unique(...): Version of embedding_lookup that avoids duplicate lookups.
enable_inference_mode(...): set inference mode.
enable_train_mode(...): enable train mode.
get_model_mode(...): get model mode.
get_variable(...): Gets an Variable object with this name if it exists,
safe_embedding_lookup_sparse(...): Provides a dynamic version of tf.nn.safe_embedding_lookup_sparse.