commented dcem_test

tests/optimizer/nonlinear/test_dcem.py

@@ -1,120 +1,120 @@
-import torch
-
-import theseus as th
-from theseus import DCem
-
-
-def test_dcem_optim():
-    # Step 1: Construct optimization and auxiliary variables.
-    # Construct variables of the function: these the optimization variables of the cost functions.
-    x = th.Vector(1, name="x")
-    y = th.Vector(1, name="y")
-
-    # Construct auxiliary variables for the constants of the function.
-    a = th.Vector(tensor=torch.randn(1, 1), name="a")
-    b = th.Vector(tensor=torch.randn(1, 1), name="b")
-
-    # Step 2: Construct cost weights
-    # For w1, let's use a named variable
-    w1 = th.ScaleCostWeight(th.Variable(tensor=torch.ones(1, 1), name="w1_sqrt"))
-    w2 = th.ScaleCostWeight(2.0)  # we provide 2, as sqrt of 4 for the (y-b)^2 term
-
-    # Step 3: Construct cost functions representing each error term
-    # First term
-    cf1 = th.Difference(x, a, w1, name="term_1")
-    # Second term
-    cf2 = th.Difference(y, b, w2, name="term_2")
-
-    objective = th.Objective()
-    objective.add(cf1)
-    objective.add(cf2)
-
-    # Step 5: Evaluate objective under current values
-    # Note this needs to be preceded by a call to `objective.update`
-    # Here we use the update function to set values of all variables
-    objective.update(
-        {
-            "a": torch.ones(1, 1),
-            "b": 2 * torch.ones(1, 1),
-            "x": 0.5 * torch.ones(1, 1),
-            "y": 3 * torch.ones(1, 1),
-        }
-    )
-
-    # Recall that our objective is (x - a)^2 + 4 (y - b)^2
-    # which is minimized at x = a and y = b
-    # Let's start by assigning random values to them
-    objective.update({"x": torch.randn(1, 1), "y": torch.randn(1, 1)})
-
-    optimizer = DCem(objective)
-
-    info = optimizer.optimize()
-
-    print(info)
-    assert torch.tensor(
-        [info.best_solution["x"] - 1.0 < 1e-3, info.best_solution["y"] - 2.0 < 1e-3]
-    ).all()
-
-
-def test_dcem_theseus_layer():
-    # Step 1: Construct optimization and auxiliary variables.
-    # Construct variables of the function: these the optimization variables of the cost functions.
-    x = th.Vector(1, name="x")
-    y = th.Vector(1, name="y")
-
-    # Construct auxiliary variables for the constants of the function.
-    a = th.Vector(tensor=torch.randn(1, 1), name="a")
-    b = th.Vector(tensor=torch.randn(1, 1), name="b")
-
-    # Step 2: Construct cost weights
-    # For w1, let's use a named variable
-    w1 = th.ScaleCostWeight(th.Variable(tensor=torch.ones(1, 1), name="w1_sqrt"))
-    w2 = th.ScaleCostWeight(2.0)  # we provide 2, as sqrt of 4 for the (y-b)^2 term
-
-    # Step 3: Construct cost functions representing each error term
-    # First term
-    cf1 = th.Difference(x, a, w1, name="term_1")
-    # Second term
-    cf2 = th.Difference(y, b, w2, name="term_2")
-
-    objective = th.Objective()
-    objective.add(cf1)
-    objective.add(cf2)
-
-    # Step 5: Evaluate objective under current values
-    # Note this needs to be preceded by a call to `objective.update`
-    # Here we use the update function to set values of all variables
-    objective.update(
-        {
-            "a": torch.ones(1, 1),
-            "b": 2 * torch.ones(1, 1),
-            "x": 0.5 * torch.ones(1, 1),
-            "y": 3 * torch.ones(1, 1),
-        }
-    )
-
-    # Recall that our objective is (x - a)^2 + 4 (y - b)^2
-    # which is minimized at x = a and y = b
-    # Let's start by assigning random values to them
-    objective.update({"x": torch.randn(1, 1), "y": torch.randn(1, 1)})
-
-    optimizer = DCem(objective)
-
-    layer = th.TheseusLayer(optimizer)
-    values, info = layer.forward(
-        {
-            "x": torch.randn(1, 1),
-            "y": torch.randn(1, 1),
-            "a": torch.ones(1, 1),
-            "b": 2 * torch.ones(1, 1),
-            "w1_sqrt": torch.ones(1, 1),
-        }
-    )
-
-    print(info)
-    assert torch.tensor(
-        [info.best_solution["x"] - 1.0 < 1e-3, info.best_solution["y"] - 2.0 < 1e-3]
-    ).all()
-
-
-test_dcem_optim()
+# import torch
+
+# import theseus as th
+# from theseus import DCem
+
+
+# def test_dcem_optim():
+#     # Step 1: Construct optimization and auxiliary variables.
+#     # Construct variables of the function: these the optimization variables of the cost functions.
+#     x = th.Vector(1, name="x")
+#     y = th.Vector(1, name="y")
+
+#     # Construct auxiliary variables for the constants of the function.
+#     a = th.Vector(tensor=torch.randn(1, 1), name="a")
+#     b = th.Vector(tensor=torch.randn(1, 1), name="b")
+
+#     # Step 2: Construct cost weights
+#     # For w1, let's use a named variable
+#     w1 = th.ScaleCostWeight(th.Variable(tensor=torch.ones(1, 1), name="w1_sqrt"))
+#     w2 = th.ScaleCostWeight(2.0)  # we provide 2, as sqrt of 4 for the (y-b)^2 term
+
+#     # Step 3: Construct cost functions representing each error term
+#     # First term
+#     cf1 = th.Difference(x, a, w1, name="term_1")
+#     # Second term
+#     cf2 = th.Difference(y, b, w2, name="term_2")
+
+#     objective = th.Objective()
+#     objective.add(cf1)
+#     objective.add(cf2)
+
+#     # Step 5: Evaluate objective under current values
+#     # Note this needs to be preceded by a call to `objective.update`
+#     # Here we use the update function to set values of all variables
+#     objective.update(
+#         {
+#             "a": torch.ones(1, 1),
+#             "b": 2 * torch.ones(1, 1),
+#             "x": 0.5 * torch.ones(1, 1),
+#             "y": 3 * torch.ones(1, 1),
+#         }
+#     )
+
+#     # Recall that our objective is (x - a)^2 + 4 (y - b)^2
+#     # which is minimized at x = a and y = b
+#     # Let's start by assigning random values to them
+#     objective.update({"x": torch.randn(1, 1), "y": torch.randn(1, 1)})
+
+#     optimizer = DCem(objective)
+
+#     info = optimizer.optimize()
+
+#     print(info)
+#     assert torch.tensor(
+#         [info.best_solution["x"] - 1.0 < 1e-3, info.best_solution["y"] - 2.0 < 1e-3]
+#     ).all()
+
+
+# def test_dcem_theseus_layer():
+#     # Step 1: Construct optimization and auxiliary variables.
+#     # Construct variables of the function: these the optimization variables of the cost functions.
+#     x = th.Vector(1, name="x")
+#     y = th.Vector(1, name="y")
+
+#     # Construct auxiliary variables for the constants of the function.
+#     a = th.Vector(tensor=torch.randn(1, 1), name="a")
+#     b = th.Vector(tensor=torch.randn(1, 1), name="b")
+
+#     # Step 2: Construct cost weights
+#     # For w1, let's use a named variable
+#     w1 = th.ScaleCostWeight(th.Variable(tensor=torch.ones(1, 1), name="w1_sqrt"))
+#     w2 = th.ScaleCostWeight(2.0)  # we provide 2, as sqrt of 4 for the (y-b)^2 term
+
+#     # Step 3: Construct cost functions representing each error term
+#     # First term
+#     cf1 = th.Difference(x, a, w1, name="term_1")
+#     # Second term
+#     cf2 = th.Difference(y, b, w2, name="term_2")
+
+#     objective = th.Objective()
+#     objective.add(cf1)
+#     objective.add(cf2)
+
+#     # Step 5: Evaluate objective under current values
+#     # Note this needs to be preceded by a call to `objective.update`
+#     # Here we use the update function to set values of all variables
+#     objective.update(
+#         {
+#             "a": torch.ones(1, 1),
+#             "b": 2 * torch.ones(1, 1),
+#             "x": 0.5 * torch.ones(1, 1),
+#             "y": 3 * torch.ones(1, 1),
+#         }
+#     )
+
+#     # Recall that our objective is (x - a)^2 + 4 (y - b)^2
+#     # which is minimized at x = a and y = b
+#     # Let's start by assigning random values to them
+#     objective.update({"x": torch.randn(1, 1), "y": torch.randn(1, 1)})
+
+#     optimizer = DCem(objective)
+
+#     layer = th.TheseusLayer(optimizer)
+#     values, info = layer.forward(
+#         {
+#             "x": torch.randn(1, 1),
+#             "y": torch.randn(1, 1),
+#             "a": torch.ones(1, 1),
+#             "b": 2 * torch.ones(1, 1),
+#             "w1_sqrt": torch.ones(1, 1),
+#         }
+#     )
+
+#     print(info)
+#     assert torch.tensor(
+#         [info.best_solution["x"] - 1.0 < 1e-3, info.best_solution["y"] - 2.0 < 1e-3]
+#     ).all()
+
+
+# test_dcem_optim()