introduced memlimit for fanova as cli-argument for spysmac_analyze, and spysmac_run for the default config now

creates the output directory if it doesn't existsy

Author: sfalkner

SpySMAC_analyze.py

@@ -107,6 +107,8 @@ def parse_args(argv):
     opt_params.add_argument("-d", "--disable_fanova", action="store_true", default=False,
                             help="disables fANOVA")
 
+    opt_params.add_argument("-m", "--memlimit_fanova", default=512,
+                            help="sets memory limit in MB for fANOVA")
 
     # Process arguments
     args = parser.parse_args(argv[1:])
@@ -149,8 +151,8 @@ def analyze_simulations(args):
               "Most likely you did not run SpySMAC_run with --seed 0.")
 
     tmp = obj.data[0]['test_performances']
-    baseline_train = np.array([tmp[j] for j in train_indices]).flatten()
-    baseline_test  = np.array([tmp[j] for j in test_indices]).flatten()
+    baseline_train = np.array([list(tmp[j]) for j in train_indices]).flatten()
+    baseline_test  = np.array([list(tmp[j]) for j in test_indices]).flatten()
 
 
     # import the data for the configuration runs
@@ -163,14 +165,13 @@ def analyze_simulations(args):
     # get performance for each run
     for i in list(obj.data.keys()):
         tmp = obj.data[i]['test_performances']
-        tmp_train = np.array([tmp[j] for j in train_indices]).flatten()
-        tmp_test  = np.array([tmp[j] for j in test_indices]).flatten()
 
+        tmp_train = np.array([list(tmp[j]) for j in train_indices]).flatten()
+        tmp_test  = np.array([list(tmp[j]) for j in test_indices]).flatten()
         run_ids.append(i)
         train_performances.append(tmp_train)
         test_performances.append(tmp_test)
 
-    
     # each ROW contains the measured performance for the instances
     train_performances = np.array(train_performances)
     test_performances  = np.array(test_performances)
@@ -216,15 +217,15 @@ def analyze_simulations(args):
 
         # fANOVA        
         try:
-            p_not_imps, fanova_not_plots = get_fanova(obj.get_pyfanova_obj(check_scenario_files = False, improvement_over="NOTHING"), 
+            p_not_imps, fanova_not_plots = get_fanova(obj.get_pyfanova_obj(check_scenario_files = False, improvement_over="NOTHING", heap_size = options['memlimit_fanova']), 
                                       cs, options['outputdir'], improvement_over="NOTHING")
         except:
             traceback.print_exc()
             logging.warn("fANOVA (without capping) failed")
             p_not_imps, fanova_not_plots = [],[]
 
         try:
-            p_def_imps, fanova_def_plots = get_fanova(obj.get_pyfanova_obj(check_scenario_files = False, improvement_over="DEFAULT"), 
+            p_def_imps, fanova_def_plots = get_fanova(obj.get_pyfanova_obj(check_scenario_files = False, improvement_over="DEFAULT", heap_size = options['memlimit_fanova']), 
                                       cs, options['outputdir'], improvement_over="DEFAULT")
         except:
             traceback.print_exc()

SpySMAC_run.py

@@ -17,6 +17,7 @@
 import signal
 import resource
 import imp
+import errno
 
 import logging
 import random
@@ -305,7 +306,14 @@ def run_simulations(args):
 
     # for the special seed 0, 
     if options['seed'] == 0:
-        # store meta information in a file for the report
+
+        # make sure that the output directory exists
+        try:
+            os.makedirs(options['outputdir'])
+        except OSError as exception:
+            if exception.errno != errno.EEXIST:
+                raise
+        # store meta information in a file for the report        
         with open(os.path.join(options['outputdir'], 'spysmac.meta'),'w') as fh:
 
             # os information