Public API for tf._api.v2.tpu.experimental.embedding namespace

Classes

class Adagrad: Optimization parameters for Adagrad with TPU embeddings.

class AdagradMomentum: Optimization parameters for Adagrad + Momentum with TPU embeddings.

class Adam: Optimization parameters for Adam with TPU embeddings.

class FTRL: Optimization parameters for FTRL with TPU embeddings.

class FeatureConfig: Configuration data for one embedding feature.

class QuantizationConfig: Settings for simulated quantization of the tpu embedding table.

class RowIdInitializer: An initializer that initializes the table with vocabulary ids.

class SGD: Optimization parameters for stochastic gradient descent for TPU embeddings.

class TPUEmbedding: The TPUEmbedding mid level API.

class TPUEmbeddingForServing: The TPUEmbedding mid level API running on CPU for serving.

class TPUEmbeddingV0: The TPUEmbedding mid level API running on TPU without Embedding accelerator.

class TPUEmbeddingV2: The TPUEmbedding mid level API running on TPU with sparse core accelerator.

class TableConfig: Configuration data for one embedding table.

Functions

serving_embedding_lookup(...): Apply standard lookup ops with tf.tpu.experimental.embedding configs.