Skip to content

manishemirani/Seq2Seq_addition

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 

Repository files navigation

Seq2Seq_addition

Trains a LSTM model to do addition task. I did this on pictures as well

With tensorflow

    400/400 [==============================] - 3s 8ms/step - loss: 0.7426 - acc: 0.7251
    400/400 [==============================] - 3s 8ms/step - loss: 0.7115 - acc: 0.7375
    400/400 [==============================] - 3s 8ms/step - loss: 0.6850 - acc: 0.7442
    400/400 [==============================] - 3s 8ms/step - loss: 0.6472 - acc: 0.7573
    56+43  = 99  (correct)
    51+17  = 66  (incorrect) (correct) = 68
    61+84  = 145 (correct)
    81+64  = 145 (correct)
    400/400 [==============================] - 3s 8ms/step - loss: 0.5913 - acc: 0.7787
    400/400 [==============================] - 3s 8ms/step - loss: 0.5275 - acc: 0.8029
    400/400 [==============================] - 3s 8ms/step - loss: 0.4450 - acc: 0.8406
    400/400 [==============================] - 3s 8ms/step - loss: 0.3497 - acc: 0.8903
    26+64  = 99  (incorrect) (correct) = 90
    77+91  = 168 (correct)
    98+28  = 126 (correct)
    66+11  = 78  (incorrect) (correct) = 77

With jax and flax

You can see the original project from here

    train epoch: 50, loss: 0.14621692895889282, accuracy: 89.24999833106995
    train epoch: 51, loss: 0.12401988357305527, accuracy: 91.9374942779541
    train epoch: 52, loss: 0.12541252374649048, accuracy: 90.37500023841858
    train epoch: 53, loss: 0.10723508894443512, accuracy: 93.81250143051147
    train epoch: 54, loss: 0.10625441372394562, accuracy: 92.56249666213989
    train epoch: 55, loss: 0.1099778413772583, accuracy: 91.99999570846558

Releases

No releases published

Packages

No packages published

Languages