Models

ModelBase

class texar.tf.models.ModelBase(hparams=None)

Base class inherited by all model classes.

A model class implements interfaces that are compatible with TF Estimator. In particular, _build() implements the model_fn interface; and get_input_fn() is for the input_fn interface.

_build(features, labels, params, mode, config=None)

Used for the model_fn argument when constructing tf.estimator.Estimator.

static default_hparams()

Returns a dictionary of hyperparameters with default values.

get_input_fn(*args, **kwargs)

Returns the input_fn function that constructs the input data, used in tf.estimator.Estimator.

hparams

A HParams instance. The hyperparameters of the module.

Seq2seqBase

class texar.tf.models.Seq2seqBase(data_hparams, hparams=None)

Base class inherited by all seq2seq model classes.

_build(features, labels, params, mode, config=None)

Used for the model_fn argument when constructing tf.estimator.Estimator.

static default_hparams()

Returns a dictionary of hyperparameters with default values.

{
    "source_embedder": "WordEmbedder",
    "source_embedder_hparams": {},
    "target_embedder": "WordEmbedder",
    "target_embedder_hparams": {},
    "embedder_share": True,
    "embedder_hparams_share": True,
    "encoder": "UnidirectionalRNNEncoder",
    "encoder_hparams": {},
    "decoder": "BasicRNNDecoder",
    "decoder_hparams": {},
    "decoding_strategy_train": "train_greedy",
    "decoding_strategy_infer": "infer_greedy",
    "beam_search_width": 0,
    "connector": "MLPTransformConnector",
    "connector_hparams": {},
    "optimization": {},
    "name": "seq2seq",
}

Here:

“source_embedder”: str or class or instance

Word embedder for source text. Can be a class, its name or module path, or a class instance.

“source_embedder_hparams”: dict

Hyperparameters for constructing the source embedder. E.g., See default_hparams() for hyperparameters of WordEmbedder. Ignored if “source_embedder” is an instance.

“target_embedder”, “target_embedder_hparams”:

Same as “source_embedder” and “source_embedder_hparams” but for target text embedder.

“embedder_share”: bool

Whether to share the source and target embedder. If True, source embedder will be used to embed target text.

“embedder_hparams_share”: bool

Whether to share the embedder configurations. If True, target embedder will be created with “source_embedder_hparams”. But the two embedders have different set of trainable variables.

“encoder”, “encoder_hparams”:

Same as “source_embedder” and “source_embedder_hparams” but for encoder.

“decoder”, “decoder_hparams”:

Same as “source_embedder” and “source_embedder_hparams” but for decoder.

“decoding_strategy_train”: str

The decoding strategy in training mode. See _build() for details.

“decoding_strategy_infer”: str

The decoding strategy in eval/inference mode.

“beam_search_width”: int

Beam width. If > 1, beam search is used in eval/inference mode.

“connector”, “connector_hparams”:

The connector class and hyperparameters. A connector transforms an encoder final state to a decoder initial state.

“optimization”: dict

Hyperparameters of optimizating the model. See default_optimization_hparams() for details.

“name”: str

Name of the model.

get_loss(decoder_results, features, labels)

Computes the training loss.

embed_source(features, labels, mode)

Embeds the inputs.

embed_target(features, labels, mode)

Embeds the target inputs. Used in training.

encode(features, labels, mode)

Encodes the inputs.

decode(encoder_results, features, labels, mode)

Decodes.

get_input_fn(mode, hparams=None)

Creates an input function input_fn that provides input data for the model in an Estimator. See, e.g., tf.estimator.train_and_evaluate.

Parameters:

• mode – One of members in tf.estimator.ModeKeys.
• hparams – A dict or an HParams instance containing the hyperparameters of PairedTextData. See default_hparams() for the the structure and default values of the hyperparameters.

Returns:

An input function that returns a tuple (features, labels) when called. features contains data fields that are related to source text, and labels contains data fields related to target text. See PairedTextData for all data fields.

hparams

A HParams instance. The hyperparameters of the module.

BasicSeq2seq

class texar.tf.models.BasicSeq2seq(data_hparams, hparams=None)

The basic seq2seq model (without attention).

Example

model = BasicSeq2seq(data_hparams, model_hparams)
exor = tx.run.Executor(
    model=model,
    data_hparams=data_hparams,
    config=run_config)
exor.train_and_evaluate(
    max_train_steps=10000,
    eval_steps=100)

_build(features, labels, params, mode, config=None)

Used for the model_fn argument when constructing tf.estimator.Estimator.

static default_hparams()

Returns a dictionary of hyperparameters with default values.

Same as default_hparams() of Seq2seqBase.

embed_source(features, labels, mode)

Embeds the inputs.

embed_target(features, labels, mode)

Embeds the target inputs. Used in training.

encode(features, labels, mode)

Encodes the inputs.

decode(encoder_results, features, labels, mode)

Decodes.

get_input_fn(mode, hparams=None)

Creates an input function input_fn that provides input data for the model in an Estimator. See, e.g., tf.estimator.train_and_evaluate.

Parameters:

• mode – One of members in tf.estimator.ModeKeys.
• hparams – A dict or an HParams instance containing the hyperparameters of PairedTextData. See default_hparams() for the the structure and default values of the hyperparameters.

Returns:

An input function that returns a tuple (features, labels) when called. features contains data fields that are related to source text, and labels contains data fields related to target text. See PairedTextData for all data fields.

get_loss(decoder_results, features, labels)

Computes the training loss.

hparams

A HParams instance. The hyperparameters of the module.