Support more recent version of numpy

cgp/ea/mu_plus_lambda.py

@@ -214,7 +214,9 @@ def _create_new_offspring_generation(self, pop: Population) -> List[IndividualBa
 
         offsprings: List[IndividualBase] = []
         while len(offsprings) < self.n_offsprings:
-            tournament_pool = pop.rng.permutation(pop.parents)[: self.tournament_size]
+            tournament_pool = pop.rng.permutation(pop.parents)[  # type: ignore
+                : self.tournament_size
+            ]
             best_in_tournament = sorted(tournament_pool, reverse=True)[0]
             offsprings.append(best_in_tournament.clone())
 

cgp/hl_api.py

@@ -64,7 +64,7 @@ def evolve(
         callback(pop)
 
     # perform evolution
-    max_fitness = np.finfo(float).min
+    max_fitness = float(np.finfo(float).min)
     # Main loop: -1 offset since the last loop iteration will still increase generation by one
     while pop.generation < max_generations - 1 and ea.n_objective_calls < max_objective_calls:
 

extra-requirements.txt

@@ -1,21 +1,21 @@
 # extra requirements
-matplotlib ~= 3.4.1
-scipy ~= 1.6.2
-sympy ~= 1.8
-torch ~= 1.8.0
-gym ~= 0.18.3
+matplotlib >= 3.4.1
+scipy >= 1.6.2
+sympy >= 1.8
+torch >= 1.8.0
+gym ~= 0.25.0
 # dev requirements
 pytest ~= 5.4.1
-mypy ~= 0.800
+mypy >= 0.800
 click ~= 8.0.1
 black ~= 19.10b0
 flake8 ~= 3.7.9
 isort ~=5.2.2
 # doc requirements
-jinja2 ~= 3.0.1
-sphinx ~=3.1.2
-recommonmark~=0.6.0
-msmb_theme~=1.2.0
-sphinx-rtd-theme~=0.5.0
-sphinx-gallery~=0.8.0
-pillow~=8.1.1
+jinja2 >= 3.0.1
+sphinx >=3.1.2
+recommonmark >= 0.6.0
+msmb_theme >= 1.2.0
+sphinx-rtd-theme >= 0.5.0
+sphinx-gallery >= 0.8.0
+pillow ~= 8.1.1

requirements.txt

@@ -1,2 +1,2 @@
-numpy ~= 1.19.0
+numpy >= 1.19.0
 docopt-ng~=0.7.2

test/test_node.py

@@ -58,7 +58,9 @@ def _test_to_numpy(genome, x, y_target):
 def _test_to_torch(genome, x, y_target):
     torch = pytest.importorskip("torch")
     graph = cgp.CartesianGraph(genome)
-    assert graph.to_torch()(torch.Tensor(x).reshape(1, -1)) == pytest.approx(y_target)
+    assert graph.to_torch()(torch.Tensor(x).reshape(1, -1)).detach().numpy() == pytest.approx(
+        y_target
+    )
 
 
 def _test_to_sympy(genome, x, y_target):
@@ -461,7 +463,7 @@ class CustomAdd(cgp.OperatorNode):
 
 
 def test_raise_broken_def_numpy_output():
-    with pytest.raises(ValueError):
+    with pytest.raises(TypeError):
 
         class CustomAdd(cgp.OperatorNode):
             _arity = 2