def main():
parser = argparse.ArgumentParser()
parser.add_argument('--traindata', help='training data file', required=True)
args = parser.parse_args()
with open(args.traindata) as f:
data = corpus.read(f)
class_counts = defaultdict(lambda: Counter())
for doc in data:
for class_name, method_name in doc:
class_counts[class_name][method_name] += 1
print '{} classes'.format(len(class_counts))
for class_name, counts in class_counts.iteritems():
print class_name
print counts.most_common(n=10)
print
def main():
logging.basicConfig(level=logging.INFO, format='%(message)s')
parser = argparse.ArgumentParser()
parser.add_argument('--traindata', help='path to training data', required=True)
parser.add_argument('--discount', help='discount parameter for PYP', required=True, type=float)
parser.add_argument('--strength', help='strength parameter for PYP', required=True, type=float)
parser.add_argument('--niter', help='number of iterations of sampling', type=int, default=10)
parser.add_argument('--char_lm_order', help='order of character language model', type=int, default=10)
args = parser.parse_args()
with open(args.traindata) as f:
_, identifiers = corpus.read(f)
char_lm = SRILMWrapper()
char_lm.train(set(identifiers), args.char_lm_order, 'wbdiscount')
base = CharLM(char_lm.ngram_file)
assert args.strength > - args.discount
model = PYP(args.discount, args.strength, base)
run_sampler(model, identifiers, args.niter)
