try optimisation

Author: m-philipps

Benchmark-Models/Fiedler_BMCSystBiol2016/reformulated/opt/calibrate.py

@@ -0,0 +1,68 @@
+import petab
+import amici
+import pypesto
+import pypesto.petab
+import pypesto.optimize
+import pypesto.store
+from pathlib import Path
+
+model_name = "Fiedler_BMCSystBiol2016"
+petab_problem = petab.Problem.from_yaml(f"../{model_name}.yaml")
+petab_problem_flat = petab.Problem.from_yaml(f"../{model_name}.yaml")
+petab.flatten_timepoint_specific_output_overrides(petab_problem_flat)
+petab_problem_flat.visualization_df = None
+
+pypesto_importer = pypesto.petab.PetabImporter(petab_problem_flat)
+pypesto_problem = pypesto_importer.create_problem()
+
+result = pypesto.optimize.minimize(
+    pypesto_problem, 
+    optimizer=pypesto.optimize.FidesOptimizer(),
+    n_starts=1000,
+)
+
+# Save result object in hdf5 file
+pypesto.store.save_to_hdf5.write_result(result, 'result.hdf5', overwrite= True)
+
+# Visualization
+import pypesto.visualize as vis
+import matplotlib.pyplot as plt
+
+vis.waterfall(result, size=(8, 4))
+plt.savefig('waterfall.png')
+plt.close()
+
+from pypesto.visualize.model_fit import visualize_optimized_model_fit
+visualize_optimized_model_fit(
+    petab_problem=petab_problem_flat,
+    result=result,
+    pypesto_problem=pypesto_problem,
+    return_dict=False,
+    start_index=n,
+    unflattened_petab_problem=petab_problem,
+)
+plt.savefig("fit.png")
+plt.close()
+
+FITS_PATH = Path(__file__).resolve().parent / "fits"
+FITS_PATH.mkdir(exist_ok=True)
+for n in range(1, 20):
+    visualize_optimized_model_fit(
+        petab_problem=petab_problem_flat,
+        result=result,
+        pypesto_problem=pypesto_problem,
+        return_dict=False,
+        start_index=n,
+        unflattened_petab_problem=petab_problem,
+    )
+    plt.savefig(FITS_PATH / f"fit_{str(n).rjust(2, '0')}.png")
+    plt.close()
+
+
+vis.parameters(
+    result,
+    legends="optimizer",
+    start_indices=[0, 1, 2, 3],
+)
+plt.savefig('parameters.png')
+plt.close()

Benchmark-Models/Fiedler_BMCSystBiol2016/reformulated/opt/sbatch.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+#SBATCH --job-name fied
+#SBATCH --output log/%j.log
+#SBATCH --nodes 1
+#SBATCH --ntasks 1
+#SBATCH --cpus-per-task 48
+#SBATCH --time 07-00:00:00
+#SBATCH --mail-type=BEGIN,END,FAIL
+#SBATCH [email protected]
+
+source home/maren/prerequisites.sh
+source /home/maren/devenv/bin/activate
+python calibrate.py