def run() -> None:
"""
Parse input from command line and run script
"""
parameters = load_params()
pi = "pi" in sys.argv
if "repeat" in sys.argv:
to_text = "text" in sys.argv
while True:
try:
to_text = pull_and_show(parameters, text=to_text, pi=pi)
except Exception as e:
msg = f'\n{time.strftime("%m/%d %H:%M:%S", time.localtime())}: Cannot run this cycle: {e}'
write_to_log(msg)
pass
time.sleep(parameters["wait_between_steps_secs"])
else:
pull_and_show(parameters, pi=pi)
def main():
model_name = os.path.basename(os.path.dirname(os.path.realpath(__file__)))
model = '../../models/{}.npz'.format(model_name)
valid_datasets = ['../../data/sequence_and_features/premise_snli_1.0_dev_token.txt',
'../../data/sequence_and_features/hypothesis_snli_1.0_dev_token.txt',
'../../data/sequence_and_features/premise_snli_1.0_dev_lemma.txt',
'../../data/sequence_and_features/hypothesis_snli_1.0_dev_lemma.txt',
'../../data/sequence_and_features/label_snli_1.0_dev.txt']
test_datasets = ['../../data/sequence_and_features/premise_snli_1.0_test_token.txt',
'../../data/sequence_and_features/hypothesis_snli_1.0_test_token.txt',
'../../data/sequence_and_features/premise_snli_1.0_test_lemma.txt',
'../../data/sequence_and_features/hypothesis_snli_1.0_test_lemma.txt',
'../../data/sequence_and_features/label_snli_1.0_test.txt']
dictionary = ['../../data/sequence_and_features/vocab_cased.pkl',
'../../data/sequence_and_features/vocab_cased_lemma.pkl']
# load model model_options
with open('%s.pkl' % model, 'rb') as f:
options = pkl.load(f)
print options
# load dictionary and invert
with open(dictionary[0], 'rb') as f:
word_dict = pkl.load(f)
print 'Loading knowledge base ...'
kb_dicts = options['kb_dicts']
with open(kb_dicts[0], 'rb') as f:
kb_dict = pkl.load(f)
n_words = options['n_words']
valid_batch_size = options['valid_batch_size']
valid = TextIterator(valid_datasets[0], valid_datasets[1], valid_datasets[2], valid_datasets[3], valid_datasets[4],
dictionary[0], dictionary[1],
n_words=n_words,
batch_size=valid_batch_size,
shuffle=False)
test = TextIterator(test_datasets[0], test_datasets[1], test_datasets[2], test_datasets[3], test_datasets[4],
dictionary[0], dictionary[1],
n_words=n_words,
batch_size=valid_batch_size,
shuffle=False)
# allocate model parameters
params = init_params(options, word_dict)
# load model parameters and set theano shared variables
params = load_params(model, params)
tparams = init_tparams(params)
trng, use_noise, \
x1, x1_mask, x1_kb, x2, x2_mask, x2_kb, kb_att, y, \
opt_ret, \
cost, \
f_pred, \
f_probs = \
build_model(tparams, options)
use_noise.set_value(0.)
valid_acc = pred_acc(f_pred, prepare_data, options, valid, kb_dict)
test_acc = pred_acc(f_pred, prepare_data, options, test, kb_dict)
print 'valid accuracy', valid_acc
print 'test accuracy', test_acc
predict_labels_valid = pred_label(f_pred, prepare_data, options, valid, kb_dict)
predict_labels_test = pred_label(f_pred, prepare_data, options, test, kb_dict)
with open('predict_gold_samples_valid.txt', 'w') as fw:
with open(valid_datasets[0], 'r') as f1:
with open(valid_datasets[1], 'r') as f2:
with open(valid_datasets[-1], 'r') as f3:
for a, b, c, d in zip(predict_labels_valid, f3, f1, f2):
fw.write(str(a) + '\t' + b.rstrip() + '\t' + c.rstrip() + '\t' + d.rstrip() + '\n')
with open('predict_gold_samples_test.txt', 'w') as fw:
with open(test_datasets[0], 'r') as f1:
with open(test_datasets[1], 'r') as f2:
with open(test_datasets[-1], 'r') as f3:
for a, b, c, d in zip(predict_labels_test, f3, f1, f2):
fw.write(str(a) + '\t' + b.rstrip() + '\t' + c.rstrip() + '\t' + d.rstrip() + '\n')
print 'Done'
def main():
model_name = os.path.basename(os.path.dirname(os.path.realpath(__file__)))
model = '../../models/{}.npz'.format(model_name)
dev_datasets = [
'../../data/word_sequence/premise_snli_1.0_dev.txt',
'../../data/word_sequence/hypothesis_snli_1.0_dev.txt',
'../../data/word_sequence/label_snli_1.0_dev.txt'
]
test_datasets = [
'../../data/word_sequence/premise_snli_1.0_test.txt',
'../../data/word_sequence/hypothesis_snli_1.0_test.txt',
'../../data/word_sequence/label_snli_1.0_test.txt'
]
dictionary = '../../data/word_sequence/vocab_cased.pkl'
# load model model_options
with open('%s.pkl' % model, 'rb') as f:
options = pkl.load(f)
print options
# load dictionary and invert
with open(dictionary, 'rb') as f:
word_dict = pkl.load(f)
word_idict = dict()
for kk, vv in word_dict.iteritems():
word_idict[vv] = kk
dev = TextIterator(dev_datasets[0],
dev_datasets[1],
dev_datasets[2],
dictionary,
n_words=options['n_words'],
batch_size=options['valid_batch_size'],
shuffle=False)
test = TextIterator(test_datasets[0],
test_datasets[1],
test_datasets[2],
dictionary,
n_words=options['n_words'],
batch_size=options['valid_batch_size'],
shuffle=False)
# allocate model parameters
params = init_params(options, word_dict)
# load model parameters and set theano shared variables
params = load_params(model, params)
tparams = init_tparams(params)
trng, use_noise, \
x1, x1_mask, char_x1, char_x1_mask, x2, x2_mask, char_x2, char_x2_mask, y, \
opt_ret, \
cost, \
f_pred, f_prods = \
build_model(tparams, options)
use_noise.set_value(0.)
dev_acc = pred_acc(f_pred, prepare_data, options, dev, word_idict)
test_acc = pred_acc(f_pred, prepare_data, options, test, word_idict)
print 'dev accuracy', dev_acc
print 'test accuracy', test_acc
predict_labels_dev = pred_label(f_prods, prepare_data, options, dev,
word_idict)
predict_labels_test = pred_label(f_prods, prepare_data, options, test,
word_idict)
with open('predict_gold_samples_dev.txt', 'w') as fw:
with open(dev_datasets[0], 'r') as f1:
with open(dev_datasets[1], 'r') as f2:
with open(dev_datasets[-1], 'r') as f3:
for a, b, c, d in zip(predict_labels_dev, f3, f1, f2):
fw.write(
str(a) + '\t' + b.rstrip() + '\t' + c.rstrip() +
'\t' + d.rstrip() + '\n')
with open('predict_gold_samples_test.txt', 'w') as fw:
with open(test_datasets[0], 'r') as f1:
with open(test_datasets[1], 'r') as f2:
with open(test_datasets[-1], 'r') as f3:
for a, b, c, d in zip(predict_labels_test, f3, f1, f2):
fw.write(
str(a) + '\t' + b.rstrip() + '\t' + c.rstrip() +
'\t' + d.rstrip() + '\n')
print 'Done'
def main():
model_name = os.path.basename(os.path.dirname(os.path.realpath(__file__)))
model = "{}.npz".format(model_name)
datasets = [
"../../data/word_sequence/premise_snli_1.0_train.txt",
"../../data/word_sequence/hypothesis_snli_1.0_train.txt",
"../../data/word_sequence/label_snli_1.0_train.txt",
]
valid_datasets = [
"../../data/word_sequence/premise_snli_1.0_dev.txt",
"../../data/word_sequence/hypothesis_snli_1.0_dev.txt",
"../../data/word_sequence/label_snli_1.0_dev.txt",
]
test_datasets = [
"../../data/word_sequence/premise_snli_1.0_test.txt",
"../../data/word_sequence/hypothesis_snli_1.0_test.txt",
"../../data/word_sequence/label_snli_1.0_test.txt",
]
dictionary = "../../data/word_sequence/vocab_cased.pkl"
# load model model_options
with open("%s.pkl" % model, "rb") as f:
options = pkl.load(f)
print(options)
# load dictionary and invert
with open(dictionary, "rb") as f:
word_dict = pkl.load(f)
n_words = options["n_words"]
valid_batch_size = options["valid_batch_size"]
valid = TextIterator(
valid_datasets[0],
valid_datasets[1],
valid_datasets[2],
dictionary,
n_words=n_words,
batch_size=valid_batch_size,
shuffle=False,
)
test = TextIterator(
test_datasets[0],
test_datasets[1],
test_datasets[2],
dictionary,
n_words=n_words,
batch_size=valid_batch_size,
shuffle=False,
)
# allocate model parameters
params = init_params(options, word_dict)
# load model parameters and set theano shared variables
params = load_params(model, params)
tparams = init_tparams(params)
trng, use_noise, x1, x1_mask, x2, x2_mask, y, opt_ret, cost, f_pred = build_model(
tparams, options)
use_noise.set_value(0.0)
valid_acc = pred_acc(f_pred, prepare_data, options, valid)
test_acc = pred_acc(f_pred, prepare_data, options, test)
print("valid accuracy", valid_acc)
print("test accuracy", test_acc)
predict_labels_valid = pred_label(f_pred, prepare_data, valid)
predict_labels_test = pred_label(f_pred, prepare_data, test)
with open("predict_gold_samples_valid.txt", "w") as fw:
with open(valid_datasets[0], "r") as f1:
with open(valid_datasets[1], "r") as f2:
with open(valid_datasets[-1], "r") as f3:
for a, b, c, d in zip(predict_labels_valid, f3, f1, f2):
fw.write(
str(a) + "\t" + b.rstrip() + "\t" + c.rstrip() +
"\t" + d.rstrip() + "\n")
with open("predict_gold_samples_test.txt", "w") as fw:
with open(test_datasets[0], "r") as f1:
with open(test_datasets[1], "r") as f2:
with open(test_datasets[-1], "r") as f3:
for a, b, c, d in zip(predict_labels_test, f3, f1, f2):
fw.write(
str(a) + "\t" + b.rstrip() + "\t" + c.rstrip() +
"\t" + d.rstrip() + "\n")
print("Done")