Add data_type to hyperparameters (#54)

When we describe a training job the data type of the hyper parameters is
lost because we use a dict[str, str]. This adds a new field to
Hyperparameter so that we can convert the datatypes at runtime.

instead of validating with isinstance(), we cast the hp value to the type it
is meant to be. This enforces a "strongly typed" value. When we
deserialize from the API string responses it becomes easier to deal with
too.

Author: iquintero

src/sagemaker/amazon/amazon_estimator.py

@@ -28,8 +28,8 @@ class AmazonAlgorithmEstimatorBase(EstimatorBase):
     """Base class for Amazon first-party Estimator implementations. This class isn't intended
     to be instantiated directly."""
 
-    feature_dim = hp('feature_dim', (validation.isint, validation.gt(0)))
-    mini_batch_size = hp('mini_batch_size', (validation.isint, validation.gt(0)))
+    feature_dim = hp('feature_dim', validation.gt(0), data_type=int)
+    mini_batch_size = hp('mini_batch_size', validation.gt(0), data_type=int)
 
     def __init__(self, role, train_instance_count, train_instance_type, data_location=None, **kwargs):
         """Initialize an AmazonAlgorithmEstimatorBase.

src/sagemaker/amazon/factorization_machines.py

@@ -13,7 +13,7 @@
 from sagemaker.amazon.amazon_estimator import AmazonAlgorithmEstimatorBase, registry
 from sagemaker.amazon.common import numpy_to_record_serializer, record_deserializer
 from sagemaker.amazon.hyperparameter import Hyperparameter as hp  # noqa
-from sagemaker.amazon.validation import gt, isin, isint, ge, isnumber
+from sagemaker.amazon.validation import gt, isin, ge
 from sagemaker.predictor import RealTimePredictor
 from sagemaker.model import Model
 from sagemaker.session import Session
@@ -23,34 +23,34 @@ class FactorizationMachines(AmazonAlgorithmEstimatorBase):
 
     repo = 'factorization-machines:1'
 
-    num_factors = hp('num_factors', (gt(0), isint), 'An integer greater than zero')
+    num_factors = hp('num_factors', gt(0), 'An integer greater than zero', int)
     predictor_type = hp('predictor_type', isin('binary_classifier', 'regressor'),
-                        'Value "binary_classifier" or "regressor"')
-    epochs = hp('epochs', (gt(0), isint), "An integer greater than 0")
-    clip_gradient = hp('clip_gradient', isnumber, "A float value")
-    eps = hp('eps', isnumber, "A float value")
-    rescale_grad = hp('rescale_grad', isnumber, "A float value")
-    bias_lr = hp('bias_lr', (ge(0), isnumber), "A non-negative float")
-    linear_lr = hp('linear_lr', (ge(0), isnumber), "A non-negative float")
-    factors_lr = hp('factors_lr', (ge(0), isnumber), "A non-negative float")
-    bias_wd = hp('bias_wd', (ge(0), isnumber), "A non-negative float")
-    linear_wd = hp('linear_wd', (ge(0), isnumber), "A non-negative float")
-    factors_wd = hp('factors_wd', (ge(0), isnumber), "A non-negative float")
+                        'Value "binary_classifier" or "regressor"', str)
+    epochs = hp('epochs', gt(0), "An integer greater than 0", int)
+    clip_gradient = hp('clip_gradient', (), "A float value", float)
+    eps = hp('eps', (), "A float value", float)
+    rescale_grad = hp('rescale_grad', (), "A float value", float)
+    bias_lr = hp('bias_lr', ge(0), "A non-negative float", float)
+    linear_lr = hp('linear_lr', ge(0), "A non-negative float", float)
+    factors_lr = hp('factors_lr', ge(0), "A non-negative float", float)
+    bias_wd = hp('bias_wd', ge(0), "A non-negative float", float)
+    linear_wd = hp('linear_wd', ge(0), "A non-negative float", float)
+    factors_wd = hp('factors_wd', ge(0), "A non-negative float", float)
     bias_init_method = hp('bias_init_method', isin('normal', 'uniform', 'constant'),
-                          'Value "normal", "uniform" or "constant"')
-    bias_init_scale = hp('bias_init_scale', (ge(0), isnumber), "A non-negative float")
-    bias_init_sigma = hp('bias_init_sigma', (ge(0), isnumber), "A non-negative float")
-    bias_init_value = hp('bias_init_value', isnumber, "A float value")
+                          'Value "normal", "uniform" or "constant"', str)
+    bias_init_scale = hp('bias_init_scale', ge(0), "A non-negative float", float)
+    bias_init_sigma = hp('bias_init_sigma', ge(0), "A non-negative float", float)
+    bias_init_value = hp('bias_init_value', (), "A float value", float)
     linear_init_method = hp('linear_init_method', isin('normal', 'uniform', 'constant'),
-                            'Value "normal", "uniform" or "constant"')
-    linear_init_scale = hp('linear_init_scale', (ge(0), isnumber), "A non-negative float")
-    linear_init_sigma = hp('linear_init_sigma', (ge(0), isnumber), "A non-negative float")
-    linear_init_value = hp('linear_init_value', isnumber, "A float value")
+                            'Value "normal", "uniform" or "constant"', str)
+    linear_init_scale = hp('linear_init_scale', ge(0), "A non-negative float", float)
+    linear_init_sigma = hp('linear_init_sigma', ge(0), "A non-negative float", float)
+    linear_init_value = hp('linear_init_value', (), "A float value", float)
     factors_init_method = hp('factors_init_method', isin('normal', 'uniform', 'constant'),
-                             'Value "normal", "uniform" or "constant"')
-    factors_init_scale = hp('factors_init_scale', (ge(0), isnumber), "A non-negative float")
-    factors_init_sigma = hp('factors_init_sigma', (ge(0), isnumber), "A non-negative float")
-    factors_init_value = hp('factors_init_value', isnumber, "A float value")
+                             'Value "normal", "uniform" or "constant"', str)
+    factors_init_scale = hp('factors_init_scale', ge(0), "A non-negative float", float)
+    factors_init_sigma = hp('factors_init_sigma', ge(0), "A non-negative float", float)
+    factors_init_value = hp('factors_init_value', (), "A float value", float)
 
     def __init__(self, role, train_instance_count, train_instance_type,
                  num_factors, predictor_type,

src/sagemaker/amazon/hyperparameter.py

@@ -16,7 +16,7 @@ class Hyperparameter(object):
     """An algorithm hyperparameter with optional validation. Implemented as a python
     descriptor object."""
 
-    def __init__(self, name, validate=lambda _: True, validation_message=""):
+    def __init__(self, name, validate=lambda _: True, validation_message="", data_type=str):
         """Args:
             name (str): The name of this hyperparameter
             validate (callable[object]->[bool]): A validation function or list of validation functions.
@@ -27,6 +27,7 @@ def __init__(self, name, validate=lambda _: True, validation_message=""):
         self.validation = validate
         self.validation_message = validation_message
         self.name = name
+        self.data_type = data_type
         try:
             iter(self.validation)
         except TypeError:
@@ -35,9 +36,10 @@ def __init__(self, name, validate=lambda _: True, validation_message=""):
     def validate(self, value):
         if value is None:  # We allow assignment from None, but Nones are not sent to training.
             return
+
         for valid in self.validation:
             if not valid(value):
-                error_message = "Invalid hyperparameter value {}".format(value)
+                error_message = "Invalid hyperparameter value {} for {}".format(value, self.name)
                 if self.validation_message:
                     error_message = error_message + ". Expecting: " + self.validation_message
                 raise ValueError(error_message)
@@ -50,6 +52,7 @@ def __get__(self, obj, objtype):
 
     def __set__(self, obj, value):
         """Validate the supplied value and set this hyperparameter to value"""
+        value = None if value is None else self.data_type(value)
         self.validate(value)
         if '_hyperparameters' not in dir(obj):
             obj._hyperparameters = dict()

src/sagemaker/amazon/kmeans.py

@@ -13,7 +13,7 @@
 from sagemaker.amazon.amazon_estimator import AmazonAlgorithmEstimatorBase, registry
 from sagemaker.amazon.common import numpy_to_record_serializer, record_deserializer
 from sagemaker.amazon.hyperparameter import Hyperparameter as hp  # noqa
-from sagemaker.amazon.validation import gt, isin, isint, ge
+from sagemaker.amazon.validation import gt, isin, ge
 from sagemaker.predictor import RealTimePredictor
 from sagemaker.model import Model
 from sagemaker.session import Session
@@ -23,15 +23,15 @@ class KMeans(AmazonAlgorithmEstimatorBase):
 
     repo = 'kmeans:1'
 
-    k = hp('k', (gt(1), isint), 'An integer greater-than 1')
-    init_method = hp('init_method', isin('random', 'kmeans++'), 'One of "random", "kmeans++"')
-    max_iterations = hp('local_lloyd_max_iterations', (gt(0), isint), 'An integer greater-than 0')
-    tol = hp('local_lloyd_tol', (gt(0), isint), 'An integer greater-than 0')
-    num_trials = hp('local_lloyd_num_trials', (gt(0), isint), 'An integer greater-than 0')
-    local_init_method = hp('local_lloyd_init_method', isin('random', 'kmeans++'), 'One of "random", "kmeans++"')
-    half_life_time_size = hp('half_life_time_size', (ge(0), isint), 'An integer greater-than-or-equal-to 0')
-    epochs = hp('epochs', (gt(0), isint), 'An integer greater-than 0')
-    center_factor = hp('extra_center_factor', (gt(0), isint), 'An integer greater-than 0')
+    k = hp('k', gt(1), 'An integer greater-than 1', int)
+    init_method = hp('init_method', isin('random', 'kmeans++'), 'One of "random", "kmeans++"', str)
+    max_iterations = hp('local_lloyd_max_iterations', gt(0), 'An integer greater-than 0', int)
+    tol = hp('local_lloyd_tol', gt(0), 'An integer greater-than 0', int)
+    num_trials = hp('local_lloyd_num_trials', gt(0), 'An integer greater-than 0', int)
+    local_init_method = hp('local_lloyd_init_method', isin('random', 'kmeans++'), 'One of "random", "kmeans++"', str)
+    half_life_time_size = hp('half_life_time_size', ge(0), 'An integer greater-than-or-equal-to 0', int)
+    epochs = hp('epochs', gt(0), 'An integer greater-than 0', int)
+    center_factor = hp('extra_center_factor', gt(0), 'An integer greater-than 0', int)
 
     def __init__(self, role, train_instance_count, train_instance_type, k, init_method=None,
                  max_iterations=None, tol=None, num_trials=None, local_init_method=None,

src/sagemaker/amazon/linear_learner.py

@@ -13,7 +13,7 @@
 from sagemaker.amazon.amazon_estimator import AmazonAlgorithmEstimatorBase, registry
 from sagemaker.amazon.common import numpy_to_record_serializer, record_deserializer
 from sagemaker.amazon.hyperparameter import Hyperparameter as hp  # noqa
-from sagemaker.amazon.validation import isin, gt, lt, isint, isbool, isnumber
+from sagemaker.amazon.validation import isin, gt, lt
 from sagemaker.predictor import RealTimePredictor
 from sagemaker.model import Model
 from sagemaker.session import Session
@@ -27,40 +27,41 @@ class LinearLearner(AmazonAlgorithmEstimatorBase):
 
     binary_classifier_model_selection_criteria = hp('binary_classifier_model_selection_criteria',
                                                     isin('accuracy', 'f1', 'precision_at_target_recall',
-                                                         'recall_at_target_precision', 'cross_entropy_loss'))
-    target_recall = hp('target_recall', (gt(0), lt(1)), "A float in (0,1)")
-    target_precision = hp('target_precision', (gt(0), lt(1)), "A float in (0,1)")
-    positive_example_weight_mult = hp('positive_example_weight_mult', gt(0), "A float greater than 0")
-    epochs = hp('epochs', (gt(0), isint), "An integer greater-than 0")
+                                                         'recall_at_target_precision', 'cross_entropy_loss'),
+                                                    data_type=str)
+    target_recall = hp('target_recall', (gt(0), lt(1)), "A float in (0,1)", float)
+    target_precision = hp('target_precision', (gt(0), lt(1)), "A float in (0,1)", float)
+    positive_example_weight_mult = hp('positive_example_weight_mult', gt(0), "A float greater than 0", float)
+    epochs = hp('epochs', gt(0), "An integer greater-than 0", int)
     predictor_type = hp('predictor_type', isin('binary_classifier', 'regressor'),
-                        'One of "binary_classifier" or "regressor"')
-    use_bias = hp('use_bias', isbool, "Either True or False")
-    num_models = hp('num_models', (gt(0), isint), "An integer greater-than 0")
-    num_calibration_samples = hp('num_calibration_samples', (gt(0), isint), "An integer greater-than 0")
-    init_method = hp('init_method', isin('uniform', 'normal'), 'One of "uniform" or "normal"')
-    init_scale = hp('init_scale', (gt(-1), lt(1)), 'A float in (-1, 1)')
-    init_sigma = hp('init_sigma', (gt(0), lt(1)), 'A float in (0, 1)')
-    init_bias = hp('init_bias', isnumber, 'A number')
-    optimizer = hp('optimizer', isin('sgd', 'adam', 'auto'), 'One of "sgd", "adam" or "auto')
+                        'One of "binary_classifier" or "regressor"', str)
+    use_bias = hp('use_bias', (), "Either True or False", bool)
+    num_models = hp('num_models', gt(0), "An integer greater-than 0", int)
+    num_calibration_samples = hp('num_calibration_samples', gt(0), "An integer greater-than 0", int)
+    init_method = hp('init_method', isin('uniform', 'normal'), 'One of "uniform" or "normal"', str)
+    init_scale = hp('init_scale', (gt(-1), lt(1)), 'A float in (-1, 1)', float)
+    init_sigma = hp('init_sigma', (gt(0), lt(1)), 'A float in (0, 1)', float)
+    init_bias = hp('init_bias', (), 'A number', float)
+    optimizer = hp('optimizer', isin('sgd', 'adam', 'auto'), 'One of "sgd", "adam" or "auto', str)
     loss = hp('loss', isin('logistic', 'squared_loss', 'absolute_loss', 'auto'),
-              '"logistic", "squared_loss", "absolute_loss" or"auto"')
-    wd = hp('wd', (gt(0), lt(1)), 'A float in (0,1)')
-    l1 = hp('l1', (gt(0), lt(1)), 'A float in (0,1)')
-    momentum = hp('momentum', (gt(0), lt(1)), 'A float in (0,1)')
-    learning_rate = hp('learning_rate', (gt(0), lt(1)), 'A float in (0,1)')
-    beta_1 = hp('beta_1', (gt(0), lt(1)), 'A float in (0,1)')
-    beta_2 = hp('beta_1', (gt(0), lt(1)), 'A float in (0,1)')
-    bias_lr_mult = hp('bias_lr_mult', gt(0), 'A float greater-than 0')
-    bias_wd_mult = hp('bias_wd_mult', gt(0), 'A float greater-than 0')
-    use_lr_scheduler = hp('use_lr_scheduler', isbool, 'A boolean')
-    lr_scheduler_step = hp('lr_scheduler_step', (gt(0), isint), 'An integer greater-than 0')
-    lr_scheduler_factor = hp('lr_scheduler_factor', (gt(0), lt(1)), 'A float in (0,1)')
-    lr_scheduler_minimum_lr = hp('lr_scheduler_minimum_lr', gt(0), 'A float greater-than 0')
-    normalize_data = hp('normalize_data', isbool, 'A boolean')
-    normalize_label = hp('normalize_label', isbool, 'A boolean')
-    unbias_data = hp('unbias_data', isbool, 'A boolean')
-    unbias_label = hp('unbias_label', isbool, 'A boolean')
-    num_point_for_scalar = hp('num_point_for_scalar', (isint, gt(0)), 'An integer greater-than 0')
+              '"logistic", "squared_loss", "absolute_loss" or"auto"', str)
+    wd = hp('wd', (gt(0), lt(1)), 'A float in (0,1)', float)
+    l1 = hp('l1', (gt(0), lt(1)), 'A float in (0,1)', float)
+    momentum = hp('momentum', (gt(0), lt(1)), 'A float in (0,1)', float)
+    learning_rate = hp('learning_rate', (gt(0), lt(1)), 'A float in (0,1)', float)
+    beta_1 = hp('beta_1', (gt(0), lt(1)), 'A float in (0,1)', float)
+    beta_2 = hp('beta_1', (gt(0), lt(1)), 'A float in (0,1)', float)
+    bias_lr_mult = hp('bias_lr_mult', gt(0), 'A float greater-than 0', float)
+    bias_wd_mult = hp('bias_wd_mult', gt(0), 'A float greater-than 0', float)
+    use_lr_scheduler = hp('use_lr_scheduler', (), 'A boolean', bool)
+    lr_scheduler_step = hp('lr_scheduler_step', gt(0), 'An integer greater-than 0', int)
+    lr_scheduler_factor = hp('lr_scheduler_factor', (gt(0), lt(1)), 'A float in (0,1)', float)
+    lr_scheduler_minimum_lr = hp('lr_scheduler_minimum_lr', gt(0), 'A float greater-than 0', float)
+    normalize_data = hp('normalize_data', (), 'A boolean', bool)
+    normalize_label = hp('normalize_label', (), 'A boolean', bool)
+    unbias_data = hp('unbias_data', (), 'A boolean', bool)
+    unbias_label = hp('unbias_label', (), 'A boolean', bool)
+    num_point_for_scalar = hp('num_point_for_scalar', gt(0), 'An integer greater-than 0', int)
 
     def __init__(self, role, train_instance_count, train_instance_type, predictor_type='binary_classifier',
                  binary_classifier_model_selection_criteria=None, target_recall=None, target_precision=None,

src/sagemaker/amazon/pca.py

@@ -24,14 +24,13 @@ class PCA(AmazonAlgorithmEstimatorBase):
 
     DEFAULT_MINI_BATCH_SIZE = 500
 
-    num_components = hp(name='num_components', validate=lambda x: x > 0 and isinstance(x, int),
-                        validation_message='Value must be an integer greater than zero')
+    num_components = hp(name='num_components', validate=lambda x: x > 0,
+                        validation_message='Value must be an integer greater than zero', data_type=int)
     algorithm_mode = hp(name='algorithm_mode', validate=lambda x: x in ['regular', 'stable', 'randomized'],
-                        validation_message='Value must be one of "regular", "stable", "randomized"')
-    subtract_mean = hp(name='subtract_mean', validate=lambda x: isinstance(x, bool),
-                       validation_message='Value must be a boolean')
-    extra_components = hp(name='extra_components', validate=lambda x: x >= 0 and isinstance(x, int),
-                          validation_message="Value must be an integer greater than or equal to 0")
+                        validation_message='Value must be one of "regular", "stable", "randomized"', data_type=str)
+    subtract_mean = hp(name='subtract_mean', validation_message='Value must be a boolean', data_type=bool)
+    extra_components = hp(name='extra_components', validate=lambda x: x >= 0,
+                          validation_message="Value must be an integer greater than or equal to 0", data_type=int)
 
     def __init__(self, role, train_instance_count, train_instance_type, num_components,
                  algorithm_mode=None, subtract_mean=None, extra_components=None, **kwargs):

src/sagemaker/amazon/validation.py

@@ -10,7 +10,6 @@
 # distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
 # ANY KIND, either express or implied. See the License for the specific
 # language governing permissions and limitations under the License.
-import numbers
 
 
 def gt(minimum):
@@ -41,8 +40,3 @@ def istype(expected):
     def validate(value):
         return isinstance(value, expected)
     return validate
-
-
-isint = istype(int)
-isbool = istype(bool)
-isnumber = istype(numbers.Number)  # noqa

tests/unit/test_hyperparameter.py

@@ -16,9 +16,9 @@
 
 class Test(object):
 
-    blank = Hyperparameter(name="some-name")
+    blank = Hyperparameter(name="some-name", data_type=int)
     elizabeth = Hyperparameter(name='elizabeth')
-    validated = Hyperparameter(name="validated", validate=lambda value: value > 55)
+    validated = Hyperparameter(name="validated", validate=lambda value: value > 55, data_type=int)
 
 
 def test_blank_access():
@@ -55,3 +55,20 @@ def test_validated():
     x.validated = 66
     with pytest.raises(ValueError):
         x.validated = 23
+
+
+def test_data_type():
+    x = Test()
+    x.validated = 66
+    assert type(x.validated) == Test.__dict__["validated"].data_type
+
+
+def test_from_string():
+    x = Test()
+    value = 65
+
+    x.validated = value
+    from_api = str(value)
+
+    x.validated = from_api
+    assert x.validated == value