def fetch_results(self):
men_data = []
women_data = []
men_append = men_data.append
women_append = women_data.append
for level in range(0, self.num_levels):
men_averager = Averager()
women_averager = Averager()
for result in self.results:
men_averager.add(result.men[level])
women_averager.add(result.women[level])
total_employees = men_averager.get_total(
) + women_averager.get_total()
men_avg = men_averager.get_average()
men_percentage = 100 * men_averager.get_total() / total_employees
women_avg = women_averager.get_average()
women_percentage = 100 * women_averager.get_total(
) / total_employees
men_append(men_percentage)
women_append(women_percentage)
return [men_data, women_data]
def __init__(self, bbox):
self._averager = Averager(AVG_SERIES, len(bbox))
self._seen = 0
self._unseen = 0
self.push(bbox)
input_content_text_vectors = helper.convert_lines_to_matrix(lines, word_to_index, content_text_length, 'c')
feed_dict = {input_header_text: input_header_text_vectors, input_content_text: input_content_text_vectors}
_predicted = session.run(predicted, feed_dict=feed_dict)
expected = helper.convert_lines_to_labels(lines)[:len(p_lines)]
(a_correct, _) = helper.prediction_assessment(expected, _predicted[:len(p_lines)])
correct += a_correct
total += len(p_lines)
print 'correct = ', correct, ' total = ', total, ' percentage = ', float(correct) / float(total)
print 'final results:'
print 'correct = ', correct, ' total = ', total, ' percentage = ', float(correct)/float(total)
elif train_or_test == 'train':
with open(data_filename) as data_file:
with tf.Session() as session:
session.run(init_op)
av = Averager(50)
for batch_index in range(number_of_batches):
lines = helper.read_file_in_loop(data_file, batch_size)
input_header_text_vectors = helper.convert_lines_to_matrix(lines, word_to_index, header_text_length, 'h')
input_content_text_vectors = helper.convert_lines_to_matrix(lines, word_to_index, content_text_length, 'c')
input_label_vector = helper.convert_lines_to_labels(lines)
feed_dict = {input_header_text: input_header_text_vectors, input_content_text: input_content_text_vectors, input_labels: input_label_vector}
(_loss, _, _predicted) = session.run([loss, train_op, predicted_as_vector], feed_dict=feed_dict)
assessment = helper.prediction_assessment(input_label_vector, _predicted)
(_, _percent) = assessment
av.add(_percent)
if (batch_index % 50) == 0 or batch_index == (number_of_batches-1):
print 'batch: ', batch_index, ' loss: ', _loss
print 'assessment: ', assessment
print 'Last 50 iterations average: ', av.average()
magn_file.write(hdr_line_fmt.format(M=M_hdr, E=E_hdr,
deltaE2=deltaE2_hdr))
print_options(magn_file, options)
domain_file.write('T domain_sizes\n')
print_options(domain_file, options)
for T in (float(T_str) for T_str in options.T.split(',')):
if options.verbose > 0:
sys.stderr.write('# computing T = {0:.4f}\n'.format(T))
ising = ising_module.IsingSystem(options.N, options.J, options.H, T)
ising.init_random(seed)
runner = DomainSizeRunner(ising=None, steps=options.steps,
is_verbose=options.verbose - 2,
burn_in=options.burn_in,
sample_period=options.sample_period,
window=options.window)
averager = Averager(runner, ising, is_verbose=options.verbose - 1)
averager.average(options.runs)
M_values = averager.get('M mean')
M_str = val_fmt.format(**M_values)
E_values = averager.get('E mean')
E_str = val_fmt.format(**E_values)
deltaE2_values = averager.get('deltaE^2')
deltaE2_str = val_fmt.format(**deltaE2_values)
magn_file.write(val_line_fmt.format(T=T, M=M_str, E=E_str,
deltaE2=deltaE2_str))
magn_file.flush()
domains = averager.get('domains')
distrubtion = ','.join(['{0:d}:{1:.8e}'.format(k, v)
for k, v in domains.items()])
domain_file.write('{0:.4f} {1:s}\n'.format(T, distrubtion))
domain_file.flush()
