Keras training output

I am currently training images for classification. Initially I was using a local machine with a decent GPU. The number of training images is around 3561. I used a batch size of 64. The output during training looks like this. Epoch 47/100 3561/3561 [==============================] - 1643s 461ms/sample - loss: 0.7066 - acc: 0.5906 - val_loss: 2.0526 - val_acc: 0.5577 Epoch 48/100 3561/3561 [==============================] - 2221s 624ms/sample - loss: 0.7076 - acc: 0.5883 - val_loss: 2.8131 - val_acc: 0.5653 As you can notice, it shows the number of images trained I use the exact same code within Amazon SageMaker and the output looks like this:

Epoch 1/40 56/56 [==============================] - 21s 378ms/step - loss: 0.9726 - accuracy: 0.5215 - val_loss: 0.9061 - val_accuracy: 0.4549 Epoch 2/40 56/56 [==============================] - 19s 334ms/step - loss: 0.8186 - accuracy: 0.5156 - val_loss: 2.4369 - val_accuracy: 0.4726 Epoch 3/40 56/56 [==============================] - 17s 305ms/step - loss: 0.8039 - accuracy: 0.5288 - val_loss: 2.7245 - val_accuracy: 0.5223

When I do the math ie. Number of images/batch (3561/64) it comes to around 55.64 ~ 56. Is my presumption correct or is there a setting that can report the number of images trained rather than the group?