רכיב הטיונר TFX Pipeline

קל לארגן דפים בעזרת אוספים אפשר לשמור ולסווג תוכן על סמך ההעדפות שלך.

...
from keras_tuner.engine import base_tuner

TunerFnResult = NamedTuple('TunerFnResult', [('tuner', base_tuner.BaseTuner),
                                             ('fit_kwargs', Dict[Text, Any])])

def tuner_fn(fn_args: FnArgs) -> TunerFnResult:
  """Build the tuner using the KerasTuner API.
  Args:
    fn_args: Holds args as name/value pairs.
      - working_dir: working dir for tuning.
      - train_files: List of file paths containing training tf.Example data.
      - eval_files: List of file paths containing eval tf.Example data.
      - train_steps: number of train steps.
      - eval_steps: number of eval steps.
      - schema_path: optional schema of the input data.
      - transform_graph_path: optional transform graph produced by TFT.
  Returns:
    A namedtuple contains the following:
      - tuner: A BaseTuner that will be used for tuning.
      - fit_kwargs: Args to pass to tuner's run_trial function for fitting the
                    model , e.g., the training and validation dataset. Required
                    args depend on the above tuner's implementation.
  """
  ...

...
tuner = Tuner(
    module_file=module_file,  # Contains `tuner_fn`.
    examples=transform.outputs['transformed_examples'],
    transform_graph=transform.outputs['transform_graph'],
    train_args=trainer_pb2.TrainArgs(num_steps=20),
    eval_args=trainer_pb2.EvalArgs(num_steps=5))

trainer = Trainer(
    module_file=module_file,  # Contains `run_fn`.
    examples=transform.outputs['transformed_examples'],
    transform_graph=transform.outputs['transform_graph'],
    schema=schema_gen.outputs['schema'],
    # This will be passed to `run_fn`.
    hyperparameters=tuner.outputs['best_hyperparameters'],
    train_args=trainer_pb2.TrainArgs(num_steps=100),
    eval_args=trainer_pb2.EvalArgs(num_steps=5))
...

hparams_importer = Importer(
    # This can be Tuner's output file or manually edited file. The file contains
    # text format of hyperparameters (keras_tuner.HyperParameters.get_config())
    source_uri='path/to/best_hyperparameters.txt',
    artifact_type=HyperParameters,
).with_id('import_hparams')

trainer = Trainer(
    ...
    # An alternative is directly use the tuned hyperparameters in Trainer's user
    # module code and set hyperparameters to None here.
    hyperparameters = hparams_importer.outputs['result'])

...
from tensorflow_cloud import CloudTuner

...
def tuner_fn(fn_args: FnArgs) -> TunerFnResult:
  """An implementation of tuner_fn that instantiates CloudTuner."""

  ...
  tuner = CloudTuner(
      _build_model,
      hyperparameters=...,
      ...
      project_id=...,       # GCP Project ID
      region=...,           # GCP Region where Vizier service is run.
  )

  ...
  return TuneFnResult(
      tuner=tuner,
      fit_kwargs={...}
  )

tuner = google_cloud_ai_platform.Tuner(
    ...   # Same kwargs as the above stock Tuner component.
    tune_args=proto.TuneArgs(num_parallel_trials=3),  # 3-worker parallel
    custom_config={
        # Configures Cloud AI Platform-specific configs . For for details, see
        # https://cloud.google.com/ai-platform/training/docs/reference/rest/v1/projects.jobs#traininginput.
        TUNING_ARGS_KEY:
            {
                'project': ...,
                'region': ...,
                # Configuration of machines for each master/worker in the flock.
                'masterConfig': ...,
                'workerConfig': ...,
                ...
            }
    })
...