train_sums.py

#!/usr/bin/env python

"order triplets by the sum of their two elements"

import pickle
import numpy as np

from keras.models import Model
from keras.layers import LSTM, Input
from keras.utils.np_utils import to_categorical

from PointerLSTM import PointerLSTM

#

x_file = 'data/x_sums.csv'
y_file = 'data/y_sums.csv'

split_at = 9000
batch_size = 100

hidden_size = 100
weights_file = 'model_weights/model_weights_sums_{}.hdf5'.format( hidden_size )

n_steps = 3
n_features = 2

#

x = np.loadtxt( x_file, delimiter = ',', dtype = int )
y = np.loadtxt( y_file, delimiter = ',', dtype = int )

x = x.reshape( x.shape[0], n_steps, -1 )
assert( x.shape[-1] == n_features )

YY = []
for y_ in y:
	YY.append( to_categorical( y_ ))
YY = np.asarray(YY)

x_train = x[:split_at]
x_test = x[split_at:]

y_train = y[:split_at]
y_test = y[split_at:]

YY_train = YY[:split_at]
YY_test = YY[split_at:]

#

print( "building model..." )
main_input = Input( shape=( x.shape[1], x.shape[2] ), name='main_input' )

encoder = LSTM(output_dim = hidden_size, return_sequences = True, name="encoder")(main_input)
decoder = PointerLSTM(hidden_size, output_dim=hidden_size, name="decoder")(encoder)

model = Model( input=main_input, output=decoder )

print( "loading weights from {}...".format( weights_file ))
try:
	model.load_weights( weights_file )
except IOError:
	print( "no weights file, starting anew." )

model.compile(optimizer='rmsprop',
			  loss='categorical_crossentropy',
			  metrics=['accuracy'])
			  
print( 'training and saving model weights each epoch...' )

validation_data = ( x_test, YY_test )

while True:

	history = model.fit( x_train, YY_train, nb_epoch = 1, batch_size = batch_size, 
		validation_data = validation_data )

	p = model.predict( x_test )

	for y_, p_ in zip( y_test, p )[:5]:
		print "y_test:", y_
		print "p:     ", p_.argmax( axis = 1 )
		print

	model.save_weights( weights_file )