Save best gdat/scan files

docs/config_keys.rst

@@ -245,6 +245,14 @@ Output Options
 
   Example:
     * ``output_every = 1000``
+
+**save_best_data**
+  If 1, run an extra simulation at the end of fitting using the best-fit parameters, and save the best-fit .gdat and .scan files to the Results directory. 
+
+  Default: 0
+
+  Example:
+    * ``save_best_data = 1``
 
 **verbosity**
   An integer value that specifies the amount of information output to the terminal.

pybnf/algorithms.py

@@ -14,7 +14,7 @@
 from .pset import PSet
 from .pset import Trajectory
 
-from .pset import NetModel, BNGLModel
+from .pset import NetModel, BNGLModel, SbmlModelNoTimeout
 from .pset import OutOfBoundsException
 from .pset import FailedSimulationError
 from .printing import print0, print1, print2, PybnfError
@@ -29,6 +29,7 @@
 import sys
 import traceback
 import pickle
+from glob import glob
 
 
 logger = logging.getLogger(__name__)
@@ -849,6 +850,30 @@ def run(self, log_prefix, scheduler_node=None, resume=None, debug=False):
                         logger.error('Cannot find files corresponding to best fit parameter set')
                         print0('Could not find your best fit gdat file. This could happen if all of the simulations\n'
                                ' in your run failed, or if that gdat file was somehow deleted during the run.')
+        if self.config.config['delete_old_files'] > 0 and self.config.config['save_best_data']:
+            # Rerun the best fit parameter set so the gdat file(s) are saved in the Results folder.
+            logger.info('Rerunning best fit parameter set to save data files.')
+            # Enable saving files for SBML models
+            for m in self.model_list:
+                if isinstance(m, SbmlModelNoTimeout):
+                    m.save_files = True
+            finaljob = Job(self.model_list, best_pset, 'bestfit',
+                           self.sim_dir, self.config.config['wall_time_sim'], None,
+                           self.config.config['normalization'], self.config.postprocessing,
+                           False)
+            try:
+                run_job(finaljob)
+            except Exception:
+                logger.exception('Failed to rerun best fit parameter set')
+                print1('Failed to rerun best fit parameter set. See log for details')
+            else:
+                # Copy all gdat and scan to Results
+                for fname in glob(self.sim_dir+'/bestfit/*.gdat') + glob(self.sim_dir+'/bestfit/*.scan'):
+                    shutil.copy(fname, self.res_dir)
+            # Disable saving files for SBML models (in case there is future bootstrapping or refinement)
+            for m in self.model_list:
+                if isinstance(m, SbmlModelNoTimeout):
+                    m.save_files = False
 
         try:
             os.rename('%s/alg_backup.bp' % self.config.config['output_dir'],

pybnf/config.py

@@ -137,7 +137,7 @@ def default_config():
             'output_every': 20, 'initialization': 'lh', 'refine': 0, 'bng_command': bng_command, 'smoothing': 1,
             'backup_every': 1, 'time_course': (), 'param_scan': (), 'min_objective': -np.inf, 'bootstrap': 0,
             'bootstrap_max_obj': None, 'ind_var_rounding': 0, 'local_objective_eval': 0, 'constraint_scale': 1.0,
-            'sbml_integrator': 'cvode', 'parallel_count': None,
+            'sbml_integrator': 'cvode', 'parallel_count': None, 'save_best_data': 0,
 
             'mutation_rate': 0.5, 'mutation_factor': 0.5, 'islands': 1, 'migrate_every': 20, 'num_to_migrate': 3,
             'stop_tolerance': 0.002, 'de_strategy': 'rand1',

pybnf/parse.py

@@ -21,7 +21,7 @@
                'local_min_limit', 'reserve_size', 'burn_in', 'sample_every', 'output_hist_every',
                'hist_bins', 'refine', 'simplex_max_iterations', 'wall_time_sim', 'wall_time_gen', 'verbosity',
                'exchange_every', 'backup_every', 'bootstrap', 'crossover_number', 'ind_var_rounding',
-               'local_objective_eval', 'reps_per_beta']
+               'local_objective_eval', 'reps_per_beta', 'save_best_data']
 numkeys_float = ['extra_weight', 'swap_rate', 'min_objective', 'cognitive', 'social', 'particle_weight',
                  'particle_weight_final', 'adaptive_n_max', 'adaptive_n_stop', 'adaptive_abs_tol', 'adaptive_rel_tol',
                  'mutation_rate', 'mutation_factor', 'stop_tolerance', 'step_size', 'simplex_step', 'simplex_log_step',