seq2seq calc source data

calc.py

import math
import csv
import random

random.seed(10)

n_numbers = 3 # how many numbers do we want to operate on

largest = 10 #largest INT

# Character set
character_set = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '+', '*', ' ']
data_dim = len(character_set)  # size of our vocab; also one hot encoding size

input_seq_length = 8 # max_digit_len + n_operators = n_numbers*2 + 2
output_seq_length = n_numbers + 1 #MAX Output Length

def get_sum_pairs(n_examples):
    inputs, labels = list(), list()
    char_to_int = dict((c, i) for i, c in enumerate(character_set)) # lookup table

    for i in range(n_examples):
        lhs = [random.randint(1, largest) for _ in range(n_numbers)]
        op = random.choice(['+', '*'])
        if op == '+':
            rhs = sum(lhs)
        elif op == '*':
            rhs = 1
            for l in lhs:
                rhs *= l

        lhs = [str(l) for l in lhs]
        lhs_str = " * ".join(lhs) if op == "*" else " + ".join(lhs)  # + or * only
        inputs.append(lhs_str)
        labels.append(str(rhs))

    print len(inputs), len(labels)
    return inputs, labels

def write_to_file(lines, path):
    with open(path, 'ar') as f:
        for line in lines:
            f.write("".join(line) + '\n')

trainX, trainY = get_sum_pairs(10000)
devX, devY = get_sum_pairs(3000)

write_to_file(trainX, 'train.source')
write_to_file(trainY, 'train.target')
write_to_file(devX, 'dev.source')
write_to_file(devY, 'dev.target')