validate_ensemble(args, epoch):

[shasso2s]

hallo , i don't understand this method(validate_ensemble(args, epoch):) can you plz explain it to me

[YuanGongND]

Hi there,

I assume you mean

ast/src/traintest.py

Lines 311 to 327 in 3f53567

	def validate_ensemble(args, epoch):
	exp_dir = args.exp_dir
	target = np.loadtxt(exp_dir+'/predictions/target.csv', delimiter=',')
	if epoch == 1:
	cum_predictions = np.loadtxt(exp_dir + '/predictions/predictions_1.csv', delimiter=',')
	else:
	cum_predictions = np.loadtxt(exp_dir + '/predictions/cum_predictions.csv', delimiter=',') * (epoch - 1)
	predictions = np.loadtxt(exp_dir+'/predictions/predictions_' + str(epoch) + '.csv', delimiter=',')
	cum_predictions = cum_predictions + predictions
	# remove the prediction file to save storage space
	os.remove(exp_dir+'/predictions/predictions_' + str(epoch-1) + '.csv')
	cum_predictions = cum_predictions / epoch
	np.savetxt(exp_dir+'/predictions/cum_predictions.csv', cum_predictions, delimiter=',')
	stats = calculate_stats(cum_predictions, target)
	return stats

Please refer to this paper, Section VI.B.1 Checkpoint Averaging. This function implement the above algorithm that 1) save checkpoint model of every epoch (or start from a specific epoch), 2) use each model checkpoint to do an inference on the data, and 3) average the prediction of each check point model. For this, you don't need to train multiple times, but just train once and ensemble the checkpoint models. Note: though it saves training cost compared with normal ensemble, it performs worse. See the above paper for details.

-Yuan