Updated tests for experiments

Author: Yaroslav-Muravev

.github/workflows/discovery.yml

@@ -33,7 +33,7 @@ jobs:
       - name: Run discovery tests
         run: |
           pytest tests/functional/ -m functional \
-            --discovery --report \
+            --discovery --use-solver --report \
             --operators ${{ matrix.operator }} \
             --ignore=tests/functional/tests/test_ns.py \
             --timeout=3600

tests/functional/conftest.py

@@ -27,6 +27,13 @@ def pytest_addoption(parser):
         help="Comma-separated list of operators to test",
     )
 
+    parser.addoption(
+        "--use-solver",
+        action="store_true",
+        default=False,
+        help="Enable solver-based fitness (DeepXDE) during discovery",
+    )
+
 @pytest.fixture
 def runtime_options(request):
     operators = request.config.getoption("--operators")
@@ -36,6 +43,7 @@ def runtime_options(request):
         "report": request.config.getoption("--report"),
         "report_dir": Path(request.config.getoption("--report-dir")),
         "operators": operator_list,
+        "use_solver": request.config.getoption("--use-solver"),
     }
 
 def pytest_configure(config):

tests/functional/scenarios/ac/ac.py

@@ -50,12 +50,12 @@ def noise_data(data, noise_level):
     def make_additional_tokens(self):
         return []
 
-    def make_search(self):
+    def make_search(self, use_solver=False):
         grid, data = self.ac_data()
         data_nn = self.load_pretrained_PINN()
 
         epde_search_obj = EpdeSearch(
-            use_solver=False,
+            use_solver=use_solver,
             use_pic=True,
             boundary=(5, 12),
             coordinate_tensors=(grid[0], grid[1]),

tests/functional/scenarios/burgers/burgers.py

@@ -121,14 +121,15 @@ def make_additional_tokens_discovery(self, grid):
             params_ranges={"power": (1, 1)},
             params_equality_ranges=None,
         )
-        trig_tokens = TrigonometricTokens(dimensionality=dimensionality, freq=(0.999, 1.001))
+        trig_tokens = TrigonometricTokens(dimensionality=self.dimensionality, freq=(0.999, 1.001))
         return [custom_grid_tokens]
 
-    def make_search(self):
+    def make_search(self, use_solver=False):
         grid, data = self.load_burgers_sindy_data(os.path.join(self.foldername, "burgers.mat"))
+        self.dimensionality = data.ndim - 1
 
         epde_search_obj = EpdeSearch(
-            use_solver=False,
+            use_solver=use_solver,
             use_pic=True,
             boundary=10,
             coordinate_tensors=(grid[0], grid[1]),

tests/functional/scenarios/kdv/kdv.py

@@ -118,11 +118,11 @@ def eq_to_text(eq):
             encoding="utf-8",
         )
 
-    def make_search(self):
+    def make_search(self, use_solver=False):
         grid, data = self.load_data()
 
         epde_search_obj = EpdeSearch(
-            use_solver=False,
+            use_solver=use_solver,
             use_pic=True,
             boundary=10,
             coordinate_tensors=(grid[0], grid[1]),

tests/functional/scenarios/ks/ks.py

@@ -86,11 +86,11 @@ def eq_to_text(eq):
             encoding="utf-8",
         )
 
-    def make_search(self):
+    def make_search(self, use_solver=False):
         grid, data = self.load_ks_data(os.path.join(self.foldername, "kuramoto_sivishinky.mat"))
 
         epde_search_obj = EpdeSearch(
-            use_solver=False,
+            use_solver=use_solver,
             use_pic=True,
             boundary=(50, 400),
             coordinate_tensors=grid,

tests/functional/scenarios/lorenz/lorenz.py

@@ -135,7 +135,7 @@ def lorenz_data(self):
         return t, data
 
     @pytest.mark.slow
-    def run_discovery(self, search_obj, report_dir=None, operator_name="unknown"):
+    def run_discovery(self, search_obj, report_dir=None, operator_name="unknown", use_solver=False):
         t, data = self.lorenz_data()
         noised_data = self.noise_data(data, self.noise_level)
         end = 1000
@@ -144,7 +144,7 @@ def run_discovery(self, search_obj, report_dir=None, operator_name="unknown"):
         y = noised_data[:end, 1]
         z = noised_data[:end, 2]
 
-        epde_search_obj = EpdeSearch(use_solver=False, multiobjective_mode=True, use_pic=True, boundary=(100),
+        epde_search_obj = EpdeSearch(use_solver=use_solver, multiobjective_mode=True, use_pic=True, boundary=(100),
                                      coordinate_tensors=[t, ], verbose_params={'show_iter_idx': True},
                                      device='cuda' if torch.cuda.is_available() else 'cpu')
 

tests/functional/scenarios/lv/lv.py

@@ -116,7 +116,7 @@ def make_search(self):
         return epde_search_obj
 
     @pytest.mark.slow
-    def run_discovery(self, search_obj, report_dir=None, operator_name="unknown"):
+    def run_discovery(self, search_obj, report_dir=None, operator_name="unknown", use_solver=False):
         t, data = self.lv_data()
         noised_data = self.noise_data(data, self.noise_level)
         end = 150
@@ -125,7 +125,7 @@ def run_discovery(self, search_obj, report_dir=None, operator_name="unknown"):
         v = noised_data[:end, 1]
 
         epde_search_obj = EpdeSearch(
-            use_solver=False,
+            use_solver=use_solver,
             multiobjective_mode=True,
             use_pic=True,
             boundary=15,

tests/functional/scenarios/ns/ns.py

@@ -124,14 +124,14 @@ def make_search(self):
         return epde_search_obj
 
     @pytest.mark.slow
-    def run_discovery(self, search_obj, report_dir=None, operator_name="unknown"):
+    def run_discovery(self, search_obj, report_dir=None, operator_name="unknown", use_solver=False):
         """Режим поиска уравнений (discovery)."""
         grids, data = self.ns_data()
         noised_data = self.noise_data(data, self.noise_level)
         t, y, x = grids
 
         epde_search_obj = EpdeSearch(
-            use_solver=False,
+            use_solver=use_solver,
             multiobjective_mode=True,
             use_pic=True,
             boundary=[21,21,46],

tests/functional/scenarios/vdp/vdp.py

@@ -47,13 +47,13 @@ def make_additional_tokens():
     def noise_data(data, noise_level):
         return noise_level * 0.01 * np.std(data) * np.random.normal(size=data.shape) + data
 
-    def make_search(self):
+    def make_search(self, use_solver=False):
         step = 0.05
         steps_num = 320
         t = np.arange(0., step * steps_num, step)
 
         epde_search_obj = EpdeSearch(
-            use_solver=False,
+            use_solver=use_solver,
             use_pic=True,
             boundary=2,
             coordinate_tensors=[t],