$ python translation.py ~/git/javascript-concist-chit-chat/data/fra.txt Using TensorFlow backend. Number of samples: 10000 Number of unique input tokens: 69 Number of unique output tokens: 93 Max sequence length for inputs: 16 Max sequence length for outputs: 59 Saved metadata at: /tmp/translation.keras/metadata.json Train on 8000 samples, validate on 2000 samples Epoch 1/20 2019-03-07 23:40:53.879374: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA 8000/8000 [==============================] - 24s 3ms/step - loss: 0.9242 - val_loss: 0.9666 Epoch 2/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.7361 - val_loss: 0.7743 Epoch 3/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.6229 - val_loss: 0.6941 Epoch 4/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.5664 - val_loss: 0.6479 Epoch 5/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.5270 - val_loss: 0.6120 Epoch 6/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.4960 - val_loss: 0.5716 Epoch 7/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.4691 - val_loss: 0.5653 Epoch 8/20 8000/8000 [==============================] - 23s 3ms/step - loss: 0.4458 - val_loss: 0.5349 Epoch 9/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.4261 - val_loss: 0.5236 Epoch 10/20 8000/8000 [==============================] - 23s 3ms/step - loss: 0.4085 - val_loss: 0.5067 Epoch 11/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.3930 - val_loss: 0.4963 Epoch 12/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.3783 - val_loss: 0.4877 Epoch 13/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.3646 - val_loss: 0.4846 Epoch 14/20 8000/8000 [==============================] - 23s 3ms/step - loss: 0.3517 - val_loss: 0.4780 Epoch 15/20 8000/8000 [==============================] - 23s 3ms/step - loss: 0.3396 - val_loss: 0.4721 Epoch 16/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.3278 - val_loss: 0.4631 Epoch 17/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.3168 - val_loss: 0.4683 Epoch 18/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.3061 - val_loss: 0.4607 Epoch 19/20 8000/8000 [==============================] - 24s 3ms/step - loss: 0.2961 - val_loss: 0.4574 Epoch 20/20 8000/8000 [==============================] - 23s 3ms/step - loss: 0.2863 - val_loss: 0.4564 WARNING:tensorflow:Layer lstm_1 was passed non-serializable keyword arguments: {'initial_state': [<tf.Tensor 'lstm/while/Exit_2:0' shape=(?, 256) dtype=float32>, <tf.Tensor 'lstm/while/Exit_3:0' shape=(?, 256) dtype=float32>]}. They will not be included in the serialized model (and thus will be missing at deserialization time).
Input sentence: Go. Target sentence: Va ! Decoded sentence: Continuez à nouveau.
Input sentence: Hi. Target sentence: Salut ! Decoded sentence: Restez aven !
Input sentence: Run! Target sentence: Cours ! Decoded sentence: Sais-tou ?