Stats

phonemes2text: training

Loss function of the model using 50 epochs. Download the weights to model/data/phonemes2text/ and test it using the following:

from model import phonemes2text

def test_phonemes2text():
    # not necessary to train if we have the weights in that directory already
    #phonemes2text.train(summarize=False, data_limit=None)
    phonemes2text.test()

if __name__ == '__main__':
    test_phonemes2text()

Epoch 1/50
39826/39826 [==============================] - 186s - loss: 5.6256     
Epoch 2/50
39826/39826 [==============================] - 193s - loss: 4.2672     
Epoch 3/50
39826/39826 [==============================] - 194s - loss: 3.5249     
Epoch 4/50
39826/39826 [==============================] - 192s - loss: 2.9762     
Epoch 5/50
39826/39826 [==============================] - 3168s - loss: 2.5636    
Epoch 6/50
39826/39826 [==============================] - 4379s - loss: 2.2567     
Epoch 7/50
39826/39826 [==============================] - 203s - loss: 2.0107     
Epoch 8/50
39826/39826 [==============================] - 197s - loss: 1.8158     
Epoch 9/50
39826/39826 [==============================] - 200s - loss: 1.6670     
Epoch 10/50
39826/39826 [==============================] - 1269s - loss: 1.5411    
Epoch 11/50
39826/39826 [==============================] - 303s - loss: 1.4296      
Epoch 12/50
39826/39826 [==============================] - 211s - loss: 1.3505     
Epoch 13/50
39826/39826 [==============================] - 229s - loss: 1.2731     
Epoch 14/50
39826/39826 [==============================] - 238s - loss: 1.2083     
Epoch 15/50
39826/39826 [==============================] - 218s - loss: 1.1551     
Epoch 16/50
39826/39826 [==============================] - 214s - loss: 1.1066     
Epoch 17/50
39826/39826 [==============================] - 214s - loss: 1.0636     
Epoch 18/50
39826/39826 [==============================] - 210s - loss: 1.0216     
Epoch 19/50
39826/39826 [==============================] - 213s - loss: 0.9908     
Epoch 20/50
39826/39826 [==============================] - 209s - loss: 0.9633     
Epoch 21/50
39826/39826 [==============================] - 209s - loss: 0.9363     
Epoch 22/50
39826/39826 [==============================] - 208s - loss: 0.9230     
Epoch 23/50
39826/39826 [==============================] - 191s - loss: 0.8858     
Epoch 24/50
39826/39826 [==============================] - 190s - loss: 0.8745     
Epoch 25/50
39826/39826 [==============================] - 192s - loss: 0.8552     
Epoch 26/50
39826/39826 [==============================] - 199s - loss: 0.8363     
Epoch 27/50
39826/39826 [==============================] - 222s - loss: 0.8238     
Epoch 28/50
39826/39826 [==============================] - 228s - loss: 0.8081     
Epoch 29/50
39826/39826 [==============================] - 223s - loss: 0.7951     
Epoch 30/50
39826/39826 [==============================] - 228s - loss: 0.7904     
Epoch 31/50
39826/39826 [==============================] - 232s - loss: 0.7699     
Epoch 32/50
39826/39826 [==============================] - 222s - loss: 0.7597     
Epoch 33/50
39826/39826 [==============================] - 213s - loss: 0.7495     
Epoch 34/50
39826/39826 [==============================] - 225s - loss: 0.7397     
Epoch 35/50
39826/39826 [==============================] - 220s - loss: 0.7285     
Epoch 36/50
39826/39826 [==============================] - 215s - loss: 0.7301     
Epoch 37/50
39826/39826 [==============================] - 215s - loss: 0.7176     
Epoch 38/50
39826/39826 [==============================] - 232s - loss: 0.7038     
Epoch 39/50
39826/39826 [==============================] - 281s - loss: 0.6973     
Epoch 40/50
39826/39826 [==============================] - 260s - loss: 0.6944     
Epoch 41/50
39826/39826 [==============================] - 254s - loss: 0.6886     
Epoch 42/50
39826/39826 [==============================] - 248s - loss: 0.6856     
Epoch 43/50
39826/39826 [==============================] - 258s - loss: 0.6727     
Epoch 44/50
39826/39826 [==============================] - 248s - loss: 0.6743     
Epoch 45/50
39826/39826 [==============================] - 253s - loss: 0.6618     
Epoch 46/50
39826/39826 [==============================] - 251s - loss: 0.6541     
Epoch 47/50
39826/39826 [==============================] - 259s - loss: 0.6533     
Epoch 48/50
39826/39826 [==============================] - 260s - loss: 0.6454     
Epoch 49/50
39826/39826 [==============================] - 262s - loss: 0.6437     
Epoch 50/50
39826/39826 [==============================] - 264s - loss: 0.6436

Its accuracy when testing against 14552 samples is 60.7%.

Loading testing data...
Found .npy files for X_test and y_test. Loading...
Accuracy using 14552 testing samples: 0.607064