[Improvement] Raise RuntimeError if predict is used with no parameters.

[Improvement] None can now be used as a value for Discrete Hyper parameters.

[Bugfix] None as a value for discrete hyper parameters caused crashes, but not caught in tests.

pyshac/config/hyperparameters.py

@@ -15,6 +15,18 @@
 _CUSTOM_PARAMETERS = OrderedDict()
 
 
+class _NoneTypeWrapper(object):
+    """
+    A wrapper to handle cases when `None` is passed as a possible parameter
+    value to the engine.
+    """
+    def __init__(self):
+        pass
+
+    def __call__(self, *args, **kwargs):
+        return args[0]
+
+
 class AbstractHyperParameter(ABC):
     """
     Abstract Hyper Parameter that defines the methods that all hyperparameters
@@ -143,8 +155,12 @@ def _build_maps(self, values):
                 self.id2param[i] = v
 
                 # prepare a type map from string to its type, for fast checks
-                self.param2type[v] = type(v)
-                self.param2type[str(v)] = type(v)
+                if v is not None:
+                    self.param2type[v] = type(v)
+                    self.param2type[str(v)] = type(v)
+                else:
+                    self.param2type[v] = _NoneTypeWrapper()
+                    self.param2type[str(v)] = _NoneTypeWrapper()
 
     def __repr__(self):
         s = self.name + " : "

pyshac/core/engine.py

@@ -532,10 +532,17 @@ def predict(self, num_samples=None, num_batches=None, num_workers_per_batch=None
         # Raises:
             ValueError: If `max_classifiers` is larger than the number of available
                 classifiers.
+            RuntimeError: If `classifiers` are not available, either due to not being
+                trained or not being loaded into the engine.
 
         # Returns:
             batches of samples in the form of an OrderedDict
         """
+        if self.parameters is None:
+            raise RuntimeError("Unable to find any parameters. Please make sure "
+                               "to set the parameters for the engine first, or "
+                               "load them into the engine prior to calling `predict`")
+
         if max_classfiers is not None and max_classfiers > len(self.classifiers):
             raise ValueError("Maximum number of classifiers (%d) must be less than the number of "
                              "classifiers (%d)" % (max_classfiers, len(self.classifiers)))

tests/config/test_hyperparameters.py

@@ -86,16 +86,27 @@ def test_discrete_encode_decode():
     decoded = h1.decode(encoded)
     assert decoded == values[encoded]
 
+    # Test for None input
+    values = [None, 1, 2, 3]
+    h2 = hp.DiscreteHyperParameter('h1', values)
+    sample = None
+
+    encoded = h2.encode(sample)
+    assert encoded == 0
+
+    decoded = h2.decode(encoded)
+    assert decoded == values[encoded]
+
 
 def test_discrete_serialization_deserialization():
-    h1 = hp.DiscreteHyperParameter('h1', [0, 1])
+    h1 = hp.DiscreteHyperParameter('h1', [0, 1, None])
 
     config = h1.get_config()
     assert 'name' in config
     assert 'values' in config
 
     values = config['values']
-    assert len(values) == 2
+    assert len(values) == 3
 
     h2 = hp.DiscreteHyperParameter.load_from_config(config)
     config = h2.get_config()
@@ -104,7 +115,7 @@ def test_discrete_serialization_deserialization():
     assert 'values' in config
 
     values = config['values']
-    assert len(values) == 2
+    assert len(values) == 3
 
 
 def test_multi_discrete():
@@ -156,17 +167,29 @@ def test_multi_discrete_encode_decode():
     for i in range(len(decoded)):
         assert decoded[i] == values[encoded[i]]
 
+    # Test for None input
+    values = [None, 1, 2, 3]
+    h2 = hp.MultiDiscreteHyperParameter('h1', values, sample_count=10)
+    sample = h2.sample()
+
+    encoded = h2.encode(sample)
+    assert encoded == [3, 1, 3, 1, 2, 0, 3, 2, 0, 0]
+
+    decoded = h2.decode(encoded)
+    for i in range(len(decoded)):
+        assert decoded[i] == values[encoded[i]]
+
 
 def test_multi_discrete_serialization_deserialization():
-    h1 = hp.MultiDiscreteHyperParameter('h1', [0, 1], sample_count=5)
+    h1 = hp.MultiDiscreteHyperParameter('h1', [0, 1, None], sample_count=5)
 
     config = h1.get_config()
     assert 'name' in config
     assert 'values' in config
     assert 'sample_count' in config
 
     values = config['values']
-    assert len(values) == 2
+    assert len(values) == 3
     assert config['sample_count'] == 5
 
     h2 = hp.MultiDiscreteHyperParameter.load_from_config(config)
@@ -177,7 +200,7 @@ def test_multi_discrete_serialization_deserialization():
     assert 'sample_count' in config
 
     values = config['values']
-    assert len(values) == 2
+    assert len(values) == 3
     assert config['sample_count'] == 5
 
 

tests/core/test_engine.py

@@ -230,6 +230,13 @@ def test_shac_initialization():
     with pytest.raises(ValueError):
         shac.evaluator_backend = 'random'
 
+    shac = engine.SHAC(None, total_budget=total_budget,
+                       num_batches=batch_size, objective=objective)
+
+    # No parameters
+    with pytest.raises(RuntimeError):
+        shac.predict()
+
 
 @optimizer_wrapper
 def test_shac_simple():