Jax fully working for ll. Now investigating the training of a circuit

tomsch420 · Sep 26, 2024 · f57b444 · f57b444
1 parent 114f3a1
commit f57b444
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 24 deletions.
diff --git a/src/probabilistic_model/probabilistic_circuit/jax/inner_layer.py b/src/probabilistic_model/probabilistic_circuit/jax/inner_layer.py
@@ -285,7 +285,7 @@ class ProductLayer(InnerLayer):
     units with the same scope.
     """
 
-    edges: Int[BCOO, "len(child_layers), number_of_nodes"]
+    edges: Int[BCOO, "len(child_layers), number_of_nodes"] = eqx.field(static=True)
     """
     The edges consist of a sparse matrix containing integers.
     The first dimension describes the edges for each child layer.

diff --git a/src/probabilistic_model/probabilistic_circuit/jax/input_layer.py b/src/probabilistic_model/probabilistic_circuit/jax/input_layer.py
@@ -7,6 +7,7 @@
 
 from .inner_layer import InputLayer, NXConverterLayer
 from ..nx.distributions import DiracDeltaDistribution
+import equinox as eqx
 
 
 class ContinuousLayer(InputLayer, ABC):
@@ -20,7 +21,7 @@ class ContinuousLayerWithFiniteSupport(ContinuousLayer, ABC):
     Abstract class for continuous univariate input units with finite support.
     """
 
-    interval: jax.Array
+    interval: jax.Array = eqx.field(static=True)
     """
     The interval of the distribution as a array of shape (num_nodes, 2).
     The first column contains the lower bounds and the second column the upper bounds.

diff --git a/test/test_jax/test_probabilistic_circuit.py b/test/test_jax/test_probabilistic_circuit.py
@@ -1,30 +1,21 @@
 import unittest
 
-from probabilistic_model.probabilistic_circuit.jax import SumLayer
-from probabilistic_model.probabilistic_circuit.nx.probabilistic_circuit import (SumUnit, ProductUnit,
-                                                                                ProbabilisticCircuit as NXProbabilisticCircuit)
-from probabilistic_model.probabilistic_circuit.nx.distributions.distributions import DiracDeltaDistribution
-from random_events.variable import Continuous
-
-from probabilistic_model.probabilistic_circuit.jax.probabilistic_circuit import ProbabilisticCircuit
 import jax.numpy as jnp
-import plotly.graph_objects as go
-from random_events.product_algebra import VariableMap, SimpleEvent
+import numpy as np
+from random_events.variable import Continuous
 
 from probabilistic_model.learning.jpt.jpt import JPT
 from probabilistic_model.learning.jpt.variables import infer_variables_from_dataframe
+from probabilistic_model.probabilistic_circuit.jax import SumLayer
 from probabilistic_model.probabilistic_circuit.jax.probabilistic_circuit import ProbabilisticCircuit
-import numpy as np
-import torch
-import jax.numpy as jnp
-import tqdm
-
-from probabilistic_model.probabilistic_model import ProbabilisticModel
-from probabilistic_model.utils import timeit
+from probabilistic_model.probabilistic_circuit.nx.distributions.distributions import DiracDeltaDistribution
+from probabilistic_model.probabilistic_circuit.nx.probabilistic_circuit import (SumUnit, ProductUnit,
+                                                                                ProbabilisticCircuit as NXProbabilisticCircuit)
 
 np.random.seed(69)
 import pandas as pd
 
+
 class SmallCircuitIntegrationTestCase(unittest.TestCase):
     x = Continuous("x")
     y = Continuous("y")
@@ -74,8 +65,6 @@ def test_creation(self):
         sum_layer2 = product_layer.child_layers[1]
         self.assertEqual(sum_layer1.number_of_nodes, 2)
         self.assertEqual(sum_layer2.number_of_nodes, 2)
-        self.assertTrue(jnp.allclose(sum_layer1.variables, jnp.array([1])))
-        self.assertTrue(jnp.allclose(sum_layer2.variables, jnp.array([0])))
 
     def test_ll(self):
         samples = self.nx_model.sample(1000)
@@ -86,13 +75,12 @@ def test_ll(self):
 
 class JPTIntegrationTestCase(unittest.TestCase):
     number_of_variables = 2
-    number_of_samples= 10000
+    number_of_samples = 10000
 
     jpt: NXProbabilisticCircuit
 
     @classmethod
     def setUpClass(cls):
-
         mean = np.full(cls.number_of_variables, 0)
         cov = np.random.uniform(0, 1, (cls.number_of_variables, cls.number_of_variables))
         cov = np.dot(cov, cov.T)
@@ -110,7 +98,5 @@ def test_from_jpt(self):
         self.assertTrue((jax_ll > -jnp.inf).all())
 
 
-
-
 if __name__ == '__main__':
     unittest.main()