# HG changeset patch
# User SylvainPL <sylvain.pannetier.lebeuf@umontreal.ca>
# Date 1268923942 14400
# Node ID a0264184684e20dc0f479040a22639b9eda2c4e7
# Parent  716c99f4eb3a1c5eee88d1ebc21bfce57bcec038
ajout de fonctionnalitees pour deux testsets

diff -r 716c99f4eb3a -r a0264184684e deep/stacked_dae/v_sylvain/sgd_optimization.py
--- a/deep/stacked_dae/v_sylvain/sgd_optimization.py	Wed Mar 17 16:41:51 2010 -0400
+++ b/deep/stacked_dae/v_sylvain/sgd_optimization.py	Thu Mar 18 10:52:22 2010 -0400
@@ -9,6 +9,7 @@
 import datetime
 import theano.tensor as T
 import sys
+import pickle
 
 from jobman import DD
 import jobman, jobman.sql
@@ -121,12 +122,20 @@
         
         #To be able to load them later for tests on finetune
         self.parameters_pre=[copy(x.value) for x in self.classifier.params]
+        f = open('params_pretrain.txt', 'w')
+        pickle.dump(self.parameters_pre,f)
+        f.close()
 
 
-    def finetune(self,dataset,num_finetune):
+    def finetune(self,dataset,dataset_test,num_finetune,ind_test):
         print "STARTING FINETUNING, time = ", datetime.datetime.now()
 
         minibatch_size = self.hp.minibatch_size
+        if ind_test == 0:
+            nom_test = "NIST"
+        else:
+            nom_test = "P07"
+
 
         # create a function to compute the mistakes that are made by the model
         # on the validation set, or testing set
@@ -213,6 +222,13 @@
                             iter = itermax(iter, self.max_minibatches)
                         test_losses = [test_model(x,y) for x,y in iter]
                         test_score = numpy.mean(test_losses)
+                        
+                        #test it on the second test set
+                        iter2 = dataset_test.test(minibatch_size)
+                        if self.max_minibatches:
+                            iter2 = itermax(iter2, self.max_minibatches)
+                        test_losses2 = [test_model(x,y) for x,y in iter2]
+                        test_score2 = numpy.mean(test_losses2)
 
                         self.series["test_error"].\
                             append((epoch, minibatch_index), test_score*100.)
@@ -221,6 +237,11 @@
                               'model %f %%') % 
                                      (epoch, minibatch_index+1,
                                       test_score*100.))
+                                    
+                        print(('     epoch %i, minibatch %i, test error on dataset %s of best '
+                              'model %f %%') % 
+                                     (epoch, minibatch_index+1,nom_test,
+                                      test_score2*100.))
 
                     sys.stdout.flush()
 
@@ -243,12 +264,18 @@
         print(('Optimization complete with best validation score of %f %%,'
                'with test performance %f %%') %  
                      (best_validation_loss * 100., test_score*100.))
+        print(('The test score on the %s dataset is %f')%(nom_test,test_score2*100.))
+        
         print ('The finetuning ran for %f minutes' % ((end_time-start_time)/60.))
         
 
     #Set parameters like they where right after pre-train
     def reload_parameters(self):
-
+        
+        #self.parameters_pre=pickle.load('params_pretrain.txt')
+        f = open('params_pretrain.txt')
+        self.parameters_pre=pickle.load(f)
+        f.close()
         for idx,x in enumerate(self.parameters_pre):
             self.classifier.params[idx].value=copy(x)