Merge pull request #3453 from MTDzi/master

Corrected number of shots in research/learning_to_remember_rare_events/train.py (method Trainer.run)

research/learning_to_remember_rare_events/train.py

@@ -197,7 +197,8 @@ class Trainer(object):
 
         # validation
         correct = []
-        correct_by_shot = dict((k, []) for k in xrange(self.episode_width + 1))
+        num_shots = episode_length // episode_width
+        correct_by_shot = dict((k, []) for k in xrange(num_shots))
         for _ in xrange(FLAGS.validation_length):
           x, y = self.sample_episode_batch(
               valid_data, episode_length, episode_width, 1)
@@ -213,16 +214,16 @@ class Trainer(object):
             # loop over batch examples
             for k, (yyy, yyy_preds) in enumerate(zip(yy, yy_preds)):
               yyy, yyy_preds = int(yyy), int(yyy_preds)
-              count = seen_counts[k][yyy % self.episode_width]
+              count = seen_counts[k][yyy % episode_width]
               if count in correct_by_shot:
                 correct_by_shot[count].append(
                     self.individual_compute_correct(yyy, yyy_preds))
-              seen_counts[k][yyy % self.episode_width] = count + 1
+              seen_counts[k][yyy % episode_width] = count + 1
 
         logging.info('validation overall accuracy %f', np.mean(correct))
-        logging.info('%d-shot: %.3f, ' * (self.episode_width + 1),
+        logging.info('%d-shot: %.3f, ' * num_shots,
                      *sum([[k, np.mean(correct_by_shot[k])]
-                           for k in xrange(self.episode_width + 1)], []))
+                           for k in xrange(num_shots)], []))
 
         if saver and FLAGS.save_dir:
           saved_file = saver.save(sess,