neuralogic 0.8.2.dev0__tar.gz → 0.8.4.dev0__tar.gz

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: neuralogic
-Version: 0.8.2.dev0
+Version: 0.8.4.dev0
 Summary: PyNeuraLogic lets you use Python to create Differentiable Logic Programs.
 Home-page: https://github.com/LukasZahradnik/PyNeuraLogic
 Author: Lukáš Zahradník

neuralogic-0.8.4.dev0/neuralogic/__version__.py

@@ -0,0 +1 @@
+__version__ = "0.8.4.dev"

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/neuralogic/core/builder/components.py

@@ -34,14 +34,6 @@ class Atom:
     def arity(self):
         return self._arity
 
-    @property
-    def value(self):
-        return ValueFactory.from_java(self._atom.getRawState().getValue())
-
-    @property
-    def gradient(self):
-        return ValueFactory.from_java(self._atom.getRawState().getGradient())
-
     def node_type(self) -> NeuronType:
         return NeuronType(self._atom.getClass().getSimpleName())
 
@@ -50,7 +42,20 @@ class Atom:
 
 
 class Neuron(Atom):
-    pass
+    def __init__(self, neuron, substitutions: Dict):
+        self.substitutions = substitutions
+        self._atom = neuron
+
+        self._predicate = neuron.getName()
+        self._arity = len(substitutions)
+
+    @property
+    def value(self):
+        return ValueFactory.from_java(self._atom.getRawState().getValue())
+
+    @property
+    def gradient(self):
+        return ValueFactory.from_java(self._atom.getRawState().getGradient())
 
 
 class NeuralSample:

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/neuralogic/core/constructs/java_objects.py

@@ -23,10 +23,11 @@ class ValueFactory:
     @staticmethod
     def from_java(value):
         size = list(value.size())
-
         if len(size) == 0 or size[0] == 0:
             return float(value.get(0))
-        return np.array(memoryview(value.getAsArray())).reshape(size).tolist()
+        if len(size) == 2 and size[1] != 1:
+            return np.array(memoryview(value.getAsArray())).reshape(size).tolist()
+        return np.array(memoryview(value.getAsArray())).tolist()
 
     def get_value(self, weight):
         if isinstance(weight, (float, int)) or np.ndim(weight) == 0:

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/neuralogic/core/neural_module.py

@@ -149,7 +149,7 @@ class NeuralModule:
         for weight in weights:
             if weight.isLearnable:
                 weights_dict[weight.index] = ValueFactory.from_java(weight.value)
-                weight_names[weight.index] = weight.name
+                weight_names[weight.index] = str(weight.name)
         return {
             "weights": weights_dict,
             "weight_names": weight_names,
@@ -188,6 +188,8 @@ class NeuralModule:
         *args,
         **kwargs,
     ):
+        if self._dataset_builder is None or self._settings is None:
+            raise ValueError("template is not built")
         return draw_model(self, filename, show, img_type, value_detail, graphviz_path, *args, **kwargs)
 
     def _initialize_neural_module(self, dataset_builder: DatasetBuilder, settings: SettingsProxy, model, torch: bool):

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/neuralogic/core/settings/settings_proxy.py

@@ -140,7 +140,7 @@ class SettingsProxy:
 
     @error_function.setter
     def error_function(self, error_function: ErrorFunction):
-        self.settings.inferOutputFcns = True
+        self.settings.inferOutputFcns = False
 
         if isinstance(error_function, MSE):
             self.settings.squishLastLayer = False
@@ -157,7 +157,7 @@ class SettingsProxy:
                 self.settings.squishLastLayer = True
                 java_error_function = self.settings_class.ErrorFcn.SOFTENTROPY
             else:
-                self.settings.inferOutputFcns = False
+                self.settings.inferOutputFcns = True
                 self.settings.squishLastLayer = False
                 java_error_function = self.settings_class.ErrorFcn.CROSSENTROPY
         else:

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/neuralogic/core/torch/tensor.py

@@ -45,7 +45,7 @@ class NeuralogicOptTensor(torch.Tensor):
         if value is None:
             return None
 
-        p = -torch.tensor(self._neuralogic_value_factory.from_java(value, self._neuralogic_number_format))
+        p = -torch.tensor(self._neuralogic_value_factory.from_java(value))
         return p
 
     @staticmethod

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/neuralogic.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: neuralogic
-Version: 0.8.2.dev0
+Version: 0.8.4.dev0
 Summary: PyNeuraLogic lets you use Python to create Differentiable Logic Programs.
 Home-page: https://github.com/LukasZahradnik/PyNeuraLogic
 Author: Lukáš Zahradník

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.black]
 line-length = 120
-target-version = ['py39']
+target-version = ['py313']
 
 [tool.mypy]
 ignore_missing_imports = true

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/tests/test_function.py

@@ -31,8 +31,7 @@ def test_transformation_body_function(torch_fun, fun):
     dataset = Dataset([Sample(R.h, R.input[data.tolist()])])
 
     built_dataset = model.build_dataset(dataset)
-
-    results = np.array(model.test(built_dataset)[0]).round(3)
+    results = np.array(model.test(built_dataset)).round(3)
 
     assert np.allclose(torch_result, results, atol=0.0001)
 
@@ -61,7 +60,7 @@ def test_slice_function():
     dataset = Dataset([Sample(R.h, [R.input[data]])])
 
     built_dataset = model.build_dataset(dataset)
-    results = np.array(model.test(built_dataset)[0])
+    results = np.array(model.test(built_dataset))
 
     assert np.allclose(res, results)
 
@@ -72,7 +71,7 @@ def test_slice_function():
     dataset = Dataset(Sample(R.h, [R.input[data]]))
 
     built_dataset = model.build_dataset(dataset)
-    results = np.array(model.test(built_dataset)[0])
+    results = np.array(model.test(built_dataset))
 
     assert np.allclose(res, results)
 
@@ -84,6 +83,6 @@ def test_slice_function():
     dataset = Dataset(Sample(R.h, [R.input[data]]))
 
     built_dataset = model.build_dataset(dataset)
-    results = np.array(model.test(built_dataset)[0])
+    results = np.array(model.test(built_dataset))
 
     assert np.allclose(res, results)

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/tests/test_general_modules.py

@@ -3,6 +3,7 @@ import pytest
 import torch
 
 from neuralogic.core import Template, R, Settings, Transformation
+from neuralogic.core.constructs.function.function import TransformationFunction
 from neuralogic.nn.module import Linear
 from neuralogic.dataset import Dataset, Sample
 
@@ -16,7 +17,7 @@ from neuralogic.dataset import Dataset, Sample
         (3, 3, 1, Transformation.TANH),
     ),
 )
-def test_linear_module(feature_size: int, output_size: int, num_of_inputs: int, activation: Transformation):
+def test_linear_module(feature_size: int, output_size: int, num_of_inputs: int, activation: TransformationFunction):
     """Test that PyNeuraLogic linear layer computes the same as pytorch linear layer"""
     torch.manual_seed(1)
     linear_input = torch.rand((num_of_inputs, feature_size))

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/tests/test_inference_engine.py

@@ -108,7 +108,7 @@ def test_listing_all_queries() -> None:
     template += R.edge(2, 3)
     template += R.edge(3, 1)
 
-    queries = list(template.queries())
+    queries = list(template.derivable_queries())
 
     expected_queries = sorted(
         ["edge(1, 2).", "edge(2, 3).", "edge(3, 1).", "h(2).", "h(3).", "h(1).", "h1(1).", "h1(2).", "h1(3).", "q."]
@@ -118,7 +118,7 @@ def test_listing_all_queries() -> None:
     for a, b in zip(expected_queries, str_queries):
         assert a == b
 
-    queries = list(template.queries([R.edge(1, 4)]))
+    queries = list(template.derivable_queries([R.edge(1, 4)]))
 
     expected_queries = sorted(
         [

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/tests/test_java_evaluation.py

@@ -26,28 +26,8 @@ import pytest
         (*XOR_Vectorized(), [0, 0.7, 0.657, -0.056]),
         (
             *Trains(),
-            [
-                0.693,
-                -0.751,
-                0.744,
-                0.733,
-                0.714,
-                0.75,
-                0.736,
-                -0.726,
-                0.754,
-                0.758,
-                0.403,
-                -0.752,
-                -0.624,
-                -0.742,
-                -0.378,
-                -0.758,
-                -0.756,
-                -0.758,
-                -0.687,
-                -0.383,
-            ],
+            [0.755, 0.75, -0.761, 0.749, 0.76, 0.724, -0.753, 0.754, -0.748, -0.755, -0.757, -0.761, 0.73, 0.742,
+             -0.752, -0.691, 0.748, 0.758, -0.73, -0.75]
         ),
         (
             *Mutagenesis(),
@@ -273,28 +253,8 @@ def test_evaluator_run_on_files(template: Template, dataset: BaseDataset, expect
         (
             naive_trains.template,
             naive_trains.dataset,
-            [
-                0.743,
-                0.745,
-                0.735,
-                -0.761,
-                -0.74,
-                -0.761,
-                0.739,
-                0.761,
-                -0.761,
-                -0.761,
-                -0.761,
-                -0.761,
-                -0.761,
-                -0.761,
-                -0.746,
-                -0.761,
-                -0.761,
-                -0.749,
-                -0.734,
-                0.761,
-            ],
+            [0.76, 0.761, -0.762, 0.761, 0.761, 0.758, -0.76, 0.761, -0.76, -0.761, -0.761, -0.761, 0.757, 0.757,
+             -0.761, -0.754, 0.759, 0.761, -0.757, -0.761],
             0,
         ),
         (

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/tests/test_quick_start.py

@@ -79,7 +79,7 @@ def test_model_evaluation_from_tensor():
     output = template.train(built_dataset, epochs=100)
 
     assert len(output[0]) == 3
-    assert output[1] == 3
+    assert len(output[1]) == 3
 
 
 def test_model_evaluation_from_logic():
@@ -121,4 +121,4 @@ def test_model_evaluation_from_logic():
     output = template.train(built_dataset, epochs=100)
 
     assert len(output[0]) == 3
-    assert output[1] == 3
+    assert len(output[1]) == 3

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/tests/test_recurrent_modules.py

@@ -69,7 +69,7 @@ def test_gru_module(input_size, hidden_size, sequence_len, epochs):
         loss.backward()
         optimizer.step()
 
-        result, _ = model.train(bd)
+        result = model.train(bd)
         assert np.allclose([float(x) for x in output[-1]], [float(x) for x in result[0][1]], atol=10e-5)
 
 
@@ -127,7 +127,7 @@ def test_rnn_module(input_size, hidden_size, sequence_len, epochs):
         loss.backward()
         optimizer.step()
 
-        result, _ = model.train(bd)
+        result = model.train(bd)
         assert np.allclose([float(x) for x in output[-1]], [float(x) for x in result[0][1]], atol=10e-5)
 
 
@@ -194,7 +194,7 @@ def test_lstm_module(input_size, hidden_size, sequence_len, epochs):
         loss.backward()
         optimizer.step()
 
-        result, _ = model.train(bd)
+        result = model.train(bd)
         assert np.allclose([float(x) for x in output[-1]], [float(x) for x in result[0][1]], atol=10e-5)
 
 
@@ -215,7 +215,7 @@ def test_rnn_module_with_pytorch(input_size, hidden_size, sequence_len, epochs):
     template = Template()
     template += RNN(input_size, hidden_size, "h", "f", "h0", arity=0)
 
-    model = template.build(Settings(chain_pruning=False, iso_value_compression=False, error_function=MSE()))
+    model = template.build(Settings(chain_pruning=False, iso_value_compression=False, error_function=MSE()), torch=True)
 
     parameters = model.parameters()
     pyneuralogic_tensor_parameters = model.tensor_parameters()
@@ -322,7 +322,7 @@ def test_rnn_custom(input_size, hidden_size, sequence_len, epochs):
         loss.backward()
         optimizer.step()
 
-        result, _ = model.train(bd, epochs=1)
+        result  = model.train(bd, epochs=1)
         assert np.allclose([float(x) for x in output[-1]], [float(x) for x in result[0][1]], atol=10e-5)
 
 
@@ -417,7 +417,7 @@ def test_lstm_module_simple(input_size, hidden_size, sequence_len, epochs):
         loss.backward()
         optimizer.step()
 
-        result, _ = model.train(bd, 1)
+        result = model.train(bd, 1)
         assert np.allclose([float(x) for x in output[-1]], [float(x) for x in result[0][1]], atol=10e-5)
 
 
@@ -496,5 +496,5 @@ def test_gru_module_simple(input_size, hidden_size, sequence_len, epochs):
         loss.backward()
         optimizer.step()
 
-        result, _ = model.train(bd, 1)
+        result = model.train(bd, 1)
         assert np.allclose([float(x) for x in output[-1]], [float(x) for x in result[0][1]], atol=10e-5)

{neuralogic-0.8.2.dev0 → neuralogic-0.8.4.dev0}/tests/test_special_predicates.py

@@ -1,15 +1,13 @@
 from neuralogic.core import Template, V, R
-from neuralogic.inference import InferenceEngine
 
 
 def test_neq():
     template = Template()
     template += R.head(V.X, V.Y) <= (R.special.neq(V.X, V.Y), R.val(V.X), R.val(V.Y))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1)]
 
-    out = inference_engine.q(R.head(V.X, V.Y), examples)
+    out = template.q(R.head(V.X, V.Y), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 2
@@ -25,10 +23,9 @@ def test_leq():
     template = Template()
     template += R.head(V.X, V.Y) <= (R.special.leq(V.X, V.Y), R.val(V.X), R.val(V.Y))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1)]
 
-    out = inference_engine.q(R.head(V.X, V.Y), examples)
+    out = template.q(R.head(V.X, V.Y), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 3
@@ -47,10 +44,9 @@ def test_geq():
     template = Template()
     template += R.head(V.X, V.Y) <= (R.special.geq(V.X, V.Y), R.val(V.X), R.val(V.Y))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1)]
 
-    out = inference_engine.q(R.head(V.X, V.Y), examples)
+    out = template.q(R.head(V.X, V.Y), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 3
@@ -69,10 +65,9 @@ def test_lt():
     template = Template()
     template += R.head(V.X, V.Y) <= (R.special.lt(V.X, V.Y), R.val(V.X), R.val(V.Y))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1)]
 
-    out = inference_engine.q(R.head(V.X, V.Y), examples)
+    out = template.q(R.head(V.X, V.Y), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 1
@@ -85,10 +80,9 @@ def test_gt():
     template = Template()
     template += R.head(V.X, V.Y) <= (R.special.gt(V.X, V.Y), R.val(V.X), R.val(V.Y))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1)]
 
-    out = inference_engine.q(R.head(V.X, V.Y), examples)
+    out = template.q(R.head(V.X, V.Y), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 1
@@ -101,10 +95,9 @@ def test_eq():
     template = Template()
     template += R.head(V.X, V.Y) <= (R.special.eq(V.X, V.Y), R.val(V.X), R.val(V.Y))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1)]
 
-    out = inference_engine.q(R.head(V.X, V.Y), examples)
+    out = template.q(R.head(V.X, V.Y), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 2
@@ -120,10 +113,9 @@ def test_next():
     template = Template()
     template += R.head(V.X, V.Y) <= (R.special.next(V.X, V.Y), R.val(V.X), R.val(V.Y))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1), R.val(0)]
 
-    out = inference_engine.q(R.head(V.X, V.Y), examples)
+    out = template.q(R.head(V.X, V.Y), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 2
@@ -139,10 +131,9 @@ def test_next_skip():
     template = Template()
     template += R.head(V.X, V.Y) <= (R.special.next(V.X, V.Z), R.special.next(V.Z, V.Y), R.val(V.X), R.val(V.Y))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1), R.val(0)]
 
-    out = inference_engine.q(R.head(V.X, V.Y), examples)
+    out = template.q(R.head(V.X, V.Y), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 1
@@ -155,10 +146,9 @@ def test_alldiff():
     template = Template()
     template += R.head(V.X, V.Y, V.Z) <= (R.special.alldiff(V.X, V.Y, V.Z), R.val(V.X), R.val(V.Y), R.val(V.Z))
 
-    inference_engine = InferenceEngine(template)
     examples = [R.val(1), R.val(-1), R.val(0)]
 
-    out = inference_engine.q(R.head(V.X, V.Y, V.Z), examples)
+    out = template.q(R.head(V.X, V.Y, V.Z), examples)
     out = sorted(list(out), key=lambda a: a["X"] + a["Y"])
 
     assert len(out) == 6

neuralogic-0.8.2.dev0/neuralogic/__version__.py

@@ -1 +0,0 @@
-__version__ = "0.8.2.dev"