The second Capstone project as part of the Artificial Intelligence Nanodegree, and focuses on Machine Translation Project from english to french. It foucses on using multiple variants of RNNs implemented via Keras!
All 3!
Final model and training:
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_9 (InputLayer) (None, 21) 0
_________________________________________________________________
embedding_4 (Embedding) (None, 21, 256) 50944
_________________________________________________________________
bidirectional_7 (Bidirection (None, 21, 512) 787968
_________________________________________________________________
bidirectional_8 (Bidirection (None, 21, 512) 1181184
_________________________________________________________________
bidirectional_9 (Bidirection (None, 21, 512) 1181184
_________________________________________________________________
time_distributed_9 (TimeDist (None, 21, 344) 176472
=================================================================
Total params: 3,377,752
Trainable params: 3,377,752
Non-trainable params: 0
_________________________________________________________________
Train on 110288 samples, validate on 27573 samples
Epoch 1/10
110288/110288 [==============================] - 70s 634us/step - loss: 2.4100 - acc: 0.5148 - val_loss: nan - val_acc: 0.6534
Epoch 2/10
110288/110288 [==============================] - 68s 621us/step - loss: 0.9249 - acc: 0.7439 - val_loss: nan - val_acc: 0.8198
Epoch 3/10
110288/110288 [==============================] - 69s 622us/step - loss: 0.4829 - acc: 0.8567 - val_loss: nan - val_acc: 0.8862
Epoch 4/10
110288/110288 [==============================] - 69s 622us/step - loss: 0.3154 - acc: 0.9034 - val_loss: nan - val_acc: 0.9186
Epoch 5/10
110288/110288 [==============================] - 69s 621us/step - loss: 0.2395 - acc: 0.9257 - val_loss: nan - val_acc: 0.9315
Epoch 6/10
110288/110288 [==============================] - 69s 622us/step - loss: 0.1977 - acc: 0.9380 - val_loss: nan - val_acc: 0.9458
Epoch 7/10
110288/110288 [==============================] - 69s 622us/step - loss: 0.1645 - acc: 0.9490 - val_loss: nan - val_acc: 0.9522
Epoch 8/10
110288/110288 [==============================] - 69s 622us/step - loss: 0.1409 - acc: 0.9566 - val_loss: nan - val_acc: 0.9528
Epoch 9/10
110288/110288 [==============================] - 69s 622us/step - loss: 0.1273 - acc: 0.9610 - val_loss: nan - val_acc: 0.9591
Epoch 10/10
110288/110288 [==============================] - 69s 622us/step - loss: 0.1176 - acc: 0.9643 - val_loss: nan - val_acc: 0.9670
Sample 1:
il a vu un vieux camion jaune <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>
Il a vu un vieux camion jaune
Sample 2:
new jersey est parfois calme au l' automne et il est il en <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>
new jersey est parfois calme pendant l' automne et il est neigeux en avril <PAD> <PAD> <PAD> <PAD> <PAD> <PAD> <PAD>
97% Accuracy on unseen data! Pas mal, non? :D