Fix batch accumulation

[lukeyeager]

Fix #1240

[kostinalexey]

You need to modify following lines to fix test and display parameters for caffe

542-544, 
-            solver.test_iter.append(int(math.ceil(float(self.dataset.get_entry_count(constants.VAL_DB)) / val_data_layer.data_param.batch_size)))
-            solver.test_iter.append(int(math.ceil(float(self.dataset.get_entry_count(constants.VAL_DB)) / (val_data_layer.data_param.batch_size \* self.batch_accumulation))))

603, 
-                int(math.ceil(5000.0 / train_data_layer.data_param.batch_size))
-                int(math.ceil(5000.0 / (train_data_layer.data_param.batch_size \* self.batch_accumulation)))

772-773
-            solver.test_iter.append(int(math.ceil(float(self.dataset.get_entry_count(constants.VAL_DB)) / val_image_data_layer.data_param.batch_size)))
-            solver.test_iter.append(int(math.ceil(float(self.dataset.get_entry_count(constants.VAL_DB)) / (val_image_data_layer.data_param.batch_size \* self.batch_accumulation))))

828
-                int(math.ceil(5000.0 / train_image_data_layer.data_param.batch_size))
-                int(math.ceil(5000.0 / (train_image_data_layer.data_param.batch_size \* self.batch_accumulation)))

[lukeyeager]

@kostinalexey as far as I can tell, batch accumulation has no effect during the TEST phase. Do you have proof to the contrary?

You're right about the display bug though - thanks! Let me fix that...

[kostinalexey]

You are right, test phase is not effected by batch_accumulation. My mistake.

[lukeyeager]

No problem at all - thanks for double-checking my work!