..

python/tests/test_events.py

@@ -139,9 +139,12 @@ def model_definition_events_plus_heavisides():
         "k2": 0.01,
         "k3": 5,
         "alpha": 2,
+        # FIXME: adjoint sensitivities w.r.t. beta are slightly off
         "beta": 3,
         "gamma": 2,
         "delta": 3,
+        # FIXME: adjoint sensitivities w.r.t. eta are slightly off
+        #  changing eta to e.g. 2.5 "fixes" python/tests/test_events.py::test_models[events_plus_heavisides]
         "eta": 1,
         "zeta": 5,
     }
@@ -281,6 +284,7 @@ def model_definition_nested_events():
         "k2": 0,
         "inflow_1": 4,
         "decay_1": 2,
+        # FIXME adjoint sensitivities w.r.t. decay_2 are slightly off
         "decay_2": 5,
         "bolus": 0,  # for bolus != 0, nested event sensitivities are off!
     }

python/tests/util.py

@@ -199,8 +199,8 @@ def check_trajectories_with_adjoint_sensitivities(amici_model: AmiciModel):
     solver.setRelativeTolerance(1e-14)
     solver.setAbsoluteToleranceB(1e-16)
     solver.setRelativeToleranceB(1e-15)
-    solver.setAbsoluteToleranceQuadratures(1e-16)
-    solver.setRelativeToleranceQuadratures(1e-10)
+    solver.setAbsoluteToleranceQuadratures(1e-14)
+    solver.setRelativeToleranceQuadratures(1e-8)
     rdata_asa = runAmiciSimulation(amici_model, solver=solver, edata=edata)
 
     assert_allclose(rdata_fsa.x, rdata_asa.x, atol=1e-14, rtol=1e-10)
@@ -215,7 +215,6 @@ def check_trajectories_with_adjoint_sensitivities(amici_model: AmiciModel):
     )
     df["abs_diff"] = df["fsa"] - df["asa"]
     df["rel_diff"] = df["abs_diff"] / df["fsa"]
-    print(df)
 
     # Also test against finite differences
     parameters = amici_model.getUnscaledParameters()
@@ -237,6 +236,7 @@ def check_trajectories_with_adjoint_sensitivities(amici_model: AmiciModel):
         rdata_m = runAmiciSimulation(amici_model, solver=solver, edata=edata)
         sllh_fd.append((rdata_p["llh"] - rdata_m["llh"]) / (2 * eps))
     df["fd"] = sllh_fd
+    print(df)
 
     # test less strict in terms of absolute error, as the gradient are
     # typically in the order of 1e3