🩹 Fixed integration benchmark

The unitest benchmarks are still partially broken since we decided that they are not very informative since they mostly reflect setup time

benchmark/benchmarks/integration/ex_two_datasets/benchmark.py

@@ -1,5 +1,7 @@
 from pathlib import Path
 
+from glotaran import __version__
+
 try:
     # 0.4.0 -0.5.1
     from glotaran.analysis.optimize import optimize
@@ -28,15 +30,20 @@ class IntegrationTwoDatasets:
     def setup(self):
         dataset1 = load_dataset(SCRIPT_DIR / "data/data1.ascii")
         dataset2 = load_dataset(SCRIPT_DIR / "data/data2.ascii")
-        model = load_model(str(SCRIPT_DIR / "models/model.yml"))
         parameters = load_parameters(str(SCRIPT_DIR / "models/parameters.yml"))
+        addition_kwargs = {}
+        if int(__version__.split(".")[1]) < 7:
+            model = load_model(str(SCRIPT_DIR / "models/model_lt_0.7.0.yml"))
+            addition_kwargs["non_negative_least_squares"] = True
+        else:
+            model = load_model(str(SCRIPT_DIR / "models/model.yml"))
         self.scheme = Scheme(
             model,
             parameters,
             {"dataset1": dataset1, "dataset2": dataset2},
             maximum_number_function_evaluations=11,
-            non_negative_least_squares=True,
             optimization_method="TrustRegionReflection",
+            **addition_kwargs,
         )
 
     def time_optimize(self):

benchmark/benchmarks/integration/ex_two_datasets/models/model.yml

@@ -1,4 +1,4 @@
-type: kinetic-spectrum
+default_megacomplex: decay
 
 dataset:
   dataset1:
@@ -33,16 +33,16 @@ initial_concentration:
 
 irf:
   irf1:
-    type: spectral-multi-gaussian
-    center: [irf.center]
-    width: [irf.width]
+    type: gaussian
+    center: irf.center
+    width: irf.width
   irf1_no_dispersion:
-    type: spectral-multi-gaussian
-    center: [irf.center]
-    width: [irf.width]
+    type: gaussian
+    center: irf.center
+    width: irf.width
 
 # It works without equal_area_penalties but then the inputs cannot be estimated
-equal_area_penalties:
+clp_penalties:
   - type: equal_area
     source: s1
     source_intervals: [[300, 3000]]
@@ -57,7 +57,6 @@ equal_area_penalties:
     target_intervals: [[300, 3000]]
     parameter: area.1
     weight: 0.1
-
 # Example of weight application:
 # weights:
 #   - datasets: [dataset1, dataset2]

benchmark/benchmarks/integration/ex_two_datasets/models/model_lt_0.7.0.yml

@@ -0,0 +1,65 @@
+type: kinetic-spectrum
+
+dataset:
+  dataset1:
+    megacomplex: [complex1]
+    initial_concentration: input1
+    irf: irf1
+    scale: scale.1
+  dataset2:
+    megacomplex: [complex1]
+    initial_concentration: input2
+    irf: irf1
+    scale: scale.2
+
+megacomplex:
+  complex1:
+    k_matrix: [km1]
+
+k_matrix:
+  km1:
+    matrix:
+      (s1, s1): "rates.k1"
+      (s2, s2): "rates.k2"
+      (s3, s3): "rates.k3"
+
+initial_concentration:
+  input1:
+    compartments: [s1, s2, s3]
+    parameters: [inputs.1, inputs.2, inputs.3]
+  input2:
+    compartments: [s1, s2, s3]
+    parameters: [inputs.1, inputs.7, inputs.8]
+
+irf:
+  irf1:
+    type: spectral-multi-gaussian
+    center: [irf.center]
+    width: [irf.width]
+  irf1_no_dispersion:
+    type: spectral-multi-gaussian
+    center: [irf.center]
+    width: [irf.width]
+
+# It works without equal_area_penalties but then the inputs cannot be estimated
+equal_area_penalties:
+  - type: equal_area
+    source: s1
+    source_intervals: [[300, 3000]]
+    target: s2
+    target_intervals: [[300, 3000]]
+    parameter: area.1
+    weight: 0.1
+  - type: equal_area
+    source: s1
+    source_intervals: [[300, 3000]]
+    target: s3
+    target_intervals: [[300, 3000]]
+    parameter: area.1
+    weight: 0.1
+# Example of weight application:
+# weights:
+#   - datasets: [dataset1, dataset2]
+#     global_interval: [100, 102]
+#     model_interval: [301, 502]
+#     value: 0.95