def div3_bin_seq():
train = read_file('data/train_div3_bin')
dev = read_file('data/dev_div3_bin')
vocab = ['0', '1']
labels = ['0', '1']
w2i = {w: i for i, w in enumerate(vocab)}
l2i = {l: i for i, l in enumerate(labels)}
dynet_model = DynetModel(w2i, l2i, 64, layers=16)
dynet_model.train(train, dev, iter_num=20)
def credit_card_seq():
train = read_file('data/train_credit_card')
dev = read_file('data/dev_credit_card')
vocab = [str(i) for i in range(10)] + ['-']
labels = ['0', '1']
w2i = {w: i for i, w in enumerate(vocab)}
l2i = {l: i for i, l in enumerate(labels)}
dynet_model = DynetModel(w2i, l2i, layers=4)
dynet_model.train(train, dev, iter_num=20)
def anbn_seq():
train = read_file('data/train_anbn')
dev = read_file('data/dev_anbn')
vocab = ['a', 'b']
labels = ['0', '1']
w2i = {w: i for i, w in enumerate(vocab)}
l2i = {l: i for i, l in enumerate(labels)}
dynet_model = DynetModel(w2i, l2i, layers=8)
dynet_model.train(train, dev, iter_num=10)
def pow_2_seq():
train = read_file('data/train_pow')
dev = read_file('data/dev_pow')
vocab = ['a', 'b']
labels = ['0', '1']
w2i = {w: i for i, w in enumerate(vocab)}
l2i = {l: i for i, l in enumerate(labels)}
dynet_model = DynetModel(w2i, l2i)
dynet_model.train(train, dev, iter_num=2)
def even_seq():
train = read_file('data/train_even')
dev = read_file('data/dev_even')
vocab = ['a']
labels = ['0', '1']
w2i = {w: i for i, w in enumerate(vocab)}
l2i = {l: i for i, l in enumerate(labels)}
dynet_model = DynetModel(w2i, l2i, 128, 64, 32)
dynet_model.train(train, dev, iter_num=10)
from part1.DynetModel import DynetModel
def read_file(filename):
f = open(filename, 'r')
lines = f.read().splitlines()
f.close()
return lines
if __name__ == '__main__':
print 'start'
train = read_file('train_set')
test = read_file('test_set')
vocab = map(str, range(1, 10)) + ['a', 'b', 'c', 'd']
labels = ['pos', 'neg']
w2i = {w: i for i, w in enumerate(vocab)} # map letter to index
l2i = {l: i for i, l in enumerate(labels)} # map label to index
net = DynetModel(w2i, l2i)
net.train(train, test)