Module: tfma

Init module for TensorFlow Model Analysis.


addons module: Init module for TensorFlow Model Analysis addons.

config module: Configuration types.

constants module: Constants used in TensorFlow Model Analysis.

evaluators module: Init module for TensorFlow Model Analysis evaluators.

export module: Library for exporting the EvalSavedModel.

exporter module: Exporter class represents different flavors of model export.

extractors module: Init module for TensorFlow Model Analysis extractors.

math_util module: Math utilities.

metrics module: Init module for TensorFlow Model Analysis metrics.

model_agnostic_eval module: Init module for TensorFlow Model Analysis model_agnostic_eval.

model_util module: Utils for working with models.

post_export_metrics module: Library containing helpers for adding post export metrics for evaluation.

types module: Types.

validators module: Init module for TensorFlow Model Analysis validators.

version module: Contains the version string for this release of TFMA.

view module: Initializes TFMA's view rendering api.

writers module: Init module for TensorFlow Model Analysis writers.


class AggregationOptions: A ProtocolMessage

class BinarizationOptions: A ProtocolMessage

class CombineFnWithModels: Abstract class for CombineFns that need the shared models.

class ConfidenceIntervalOptions: A ProtocolMessage

class DoFnWithModels: Abstract class for DoFns that need the shared models.

class EvalConfig: A ProtocolMessage

class EvalResult: The result of a single model analysis run.

class EvalSharedModel: Shared model used during extraction and evaluation.

class FeaturesPredictionsLabels: FeaturesPredictionsLabels(input_ref, features, predictions, labels)

class GenericChangeThreshold: A ProtocolMessage

class GenericValueThreshold: A ProtocolMessage

class MaterializedColumn: MaterializedColumn(name, value)

class MetricConfig: A ProtocolMessage

class MetricThreshold: A ProtocolMessage

class MetricsForSlice: A ProtocolMessage

class MetricsSpec: A ProtocolMessage

class ModelLoader: Model loader is responsible for loading shared model types.

class ModelSpec: A ProtocolMessage

class Options: A ProtocolMessage

class PerSliceMetricThreshold: A ProtocolMessage

class PlotsForSlice: A ProtocolMessage

class SlicingSpec: A ProtocolMessage

class ValidationResult: A ProtocolMessage


BatchedInputsToExtracts(...): Converts Arrow RecordBatch inputs to Extracts.

ExtractAndEvaluate(...): Performs Extractions and Evaluations in provided order.

ExtractEvaluateAndWriteResults(...): PTransform for performing extraction, evaluation, and writing results.

InputsToExtracts(...): Converts serialized inputs (e.g. examples) to Extracts if not already.

Validate(...): Performs validation of alternative evaluations.

WriteResults(...): Writes Evaluation or Validation results using given writers.

analyze_raw_data(...): Runs TensorFlow model analysis on a pandas.DataFrame.

compound_key(...): Returns a compound key based on a list of keys.

create_keys_key(...): Creates secondary key representing the sparse keys associated with key.

create_values_key(...): Creates secondary key representing sparse values associated with key.

default_eval_shared_model(...): Returns default EvalSharedModel.

default_evaluators(...): Returns the default evaluators for use in ExtractAndEvaluate.

default_extractors(...): Returns the default extractors for use in ExtractAndEvaluate.

default_writers(...): Returns the default writers for use in WriteResults.

get_model_type(...): Returns model type for given model spec taking into account defaults.

is_batched_input(...): Returns true if batched input should be used.

load_eval_result(...): Loads EvalResult object for use with the visualization functions.

load_eval_results(...): Loads results for multiple models or multiple data sets.

load_metrics(...): Read and deserialize the MetricsForSlice records.

load_plots(...): Read and deserialize the PlotsForSlice records.

load_validation_result(...): Read and deserialize the ValidationResult.

make_eval_results(...): Run model analysis for a single model on multiple data sets.

model_construct_fn(...): Returns function for constructing shared models.

multiple_data_analysis(...): Run model analysis for a single model on multiple data sets.

multiple_model_analysis(...): Run model analysis for multiple models on the same data set.

run_model_analysis(...): Runs TensorFlow model analysis.

unique_key(...): Returns a unique key given a list of current keys.

update_eval_config_with_defaults(...): Returns a new config with default settings applied.

verify_and_update_eval_shared_models(...): Verifies eval shared models and normnalizes to produce a single list.

verify_eval_config(...): Verifies eval config.

Type Aliases

Extracts: The central part of internal API.

MaybeMultipleEvalSharedModels: The central part of internal API.

TensorType: The central part of internal API.

TensorTypeMaybeDict: The central part of internal API.

ANALYSIS_KEY 'analysis'
ARROW_INPUT_COLUMN '__raw_record__'
ARROW_RECORD_BATCH_KEY 'arrow_record_batch'
ATTRIBUTIONS_KEY 'attributions'
BASELINE_KEY 'baseline'
BASELINE_SCORE_KEY 'baseline_score'
CANDIDATE_KEY 'candidate'
DATA_CENTRIC_MODE 'data_centric_mode'
EXAMPLE_SCORE_KEY 'example_score'
EXAMPLE_WEIGHTS_KEY 'example_weights'
FEATURES_KEY 'features'
INPUT_KEY 'input'
LABELS_KEY 'labels'
METRICS_KEY 'metrics'
MODEL_CENTRIC_MODE 'model_centric_mode'

PLOTS_KEY 'plots'
PREDICTIONS_KEY 'predictions'
SLICE_KEY_TYPES_KEY '_slice_key_types'
TF_ESTIMATOR 'tf_estimator'
TF_GENERIC 'tf_generic'
TF_JS 'tf_js'
TF_KERAS 'tf_keras'
TF_LITE 'tf_lite'
VALIDATIONS_KEY 'validations'