def main():
end_epoch = 0
for i in range(1, config.epoch + 1):
end_epoch = i
try:
if not opt.notrain:
flag = train(i)
if flag:
break
else:
eval(i, 'Valid')
except KeyboardInterrupt:
logging('Interupt\n')
break
#Select the best micro_f1 score
idx = np.argmax(scores['overall_micro_f1'])
best_epoch = idx + 1
logging("Summary (validation):\n")
d = {}
for metric in all_metrics:
logging("{}:{:.3f}\n".format(metric, scores[metric][idx]))
d[metric] = scores[metric][idx]
test_d = test(best_epoch)
val_losses = utils.combine_dict(valid_loss_printer.losses[best_epoch], d)
test_losses = utils.combine_dict(test_loss_printer.losses[best_epoch],
test_d)
results = utils.combine_results(start_local_time, time.localtime(),
time.time() - start_time, best_epoch,
'wgan', end_epoch, threshold, config,
train_loss_printer.losses[best_epoch],
val_losses, test_losses)
# Recorder
recorder = Recorder(opt.result_csv)
recorder.add_result(results)
recorder.write_csv()
print(log_path)
def main():
end_epoch = 0
for i in range(1, config.epoch + 1):
end_epoch = i
try:
if not opt.notrain:
train(i)
else:
eval(i)
except KeyboardInterrupt:
logging('Interupt\n')
break
idx = np.argmax(scores['overall_auc'])
best_epoch = idx + 1
logging("Summary (validation):\n")
for metric in all_metrics:
logging("{}:{:.3f}\n".format(metric, scores[metric][idx]))
logging("\nPerformance on test set:\n")
test_d = test(best_epoch)
d = {}
for metric in all_metrics:
logging("{}:{:.3f}\n".format(metric, scores[metric][idx]))
d[metric] = scores[metric][idx]
train_loss_d = {'logit_loss': train_loss[idx]}
results = utils.combine_results(start_local_time, time.localtime(),
time.time() - start_time, best_epoch,
'baseline', end_epoch, threshold, config,
train_loss_d, d, test_d)
# Recorder
recorder = Recorder(opt.result_csv)
recorder.add_result(results)
recorder.write_csv()
print(log_path)
