gradient checkpointing

Author: raschka

Diff for: README.md

@@ -315,10 +315,15 @@ A collection of various deep learning architectures, models, and tips for Tensor
    [PyTorch: [GitHub](pytorch_ipynb/rnn/rnn_lstm_packed_own_csv_imdb.ipynb) | [Nbviewer](https://nbviewer.jupyter.org/github/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_lstm_packed_own_csv_imdb.ipynb)]
 
 
+#### Improving Memory Efficiency
+
+- Gradient Checkpointing Demo (Network-in-Network trained on CIFAR-10)
+   [PyTorch: [GitHub](pytorch_ipynb/cnn/cnn-vgg16-celeba-data-parallel.ipynb) | [Nbviewer](https://nbviewer.jupyter.org/github/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-vgg16-celeba-data-parallel.ipynb)]
+
 #### Parallel Computing
 
 - Using Multiple GPUs with DataParallel -- VGG-16 Gender Classifier on CelebA   
-   [PyTorch: [GitHub](pytorch_ipynb/cnn/cnn-vgg16-celeba-data-parallel.ipynb) | [Nbviewer](https://nbviewer.jupyter.org/github/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-vgg16-celeba-data-parallel.ipynb)]
+   [PyTorch: [GitHub](pytorch_ipynb/mechanics/gradient-checkpointing-nin.ipynb) | [Nbviewer](https://nbviewer.jupyter.org/github/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/gradient-checkpointing-nin.ipynb)]
 
 #### Other 
 

Diff for: pytorch_ipynb/helper_train.py

@@ -62,15 +62,15 @@ def train_classifier_simple_v1(num_epochs, model, optimizer, device,
 
                 train_acc = compute_accuracy(model, train_loader, device)
                 train_loss = compute_epoch_loss(model, train_loader, device)
-                print('***Epoch: %03d/%03d | Train. Acc.: %.3f%% |  Loss: %.3f' % (
+                print('***Epoch: %03d/%03d | Train. Acc.: %.3f%% | Loss: %.3f' % (
                       epoch+1, num_epochs, train_acc, train_loss))
                 log_dict['train_loss_per_epoch'].append(train_loss.item())
                 log_dict['train_loss_per_epoch'].append(train_acc.item())
 
                 if valid_loader is not None:
                     valid_acc = compute_accuracy(model, valid_loader, device)
                     valid_loss = compute_epoch_loss(model, valid_loader, device)
-                    print('***Epoch: %03d/%03d | Valid. Acc.: %.3f%% |  Loss: %.3f' % (
+                    print('***Epoch: %03d/%03d | Valid. Acc.: %.3f%% | Loss: %.3f' % (
                           epoch+1, num_epochs, valid_acc, valid_loss))
                     log_dict['valid_loss_per_epoch'].append(valid_loss.item())
                     log_dict['valid_loss_per_epoch'].append(valid_acc.item())