def test_epoch(model, data_loader, log_interval):
model.eval()
test_loss = 0
correct = 0
histogram_data = {key: [] for key in range(NUMBER_OF_CLASSES)}
histogram_data['all'] = []
for batch_index, (x_data_raw, target) in enumerate(data_loader):
# x_data_ts = Variable(x_data_ts, volatile=True)
x_data_raw = Variable(x_data_raw, volatile=True)
target = Variable(target)
output = model(x_data_raw)
if type(output.data) == torch.cuda.DoubleTensor:
output = output.cpu()
test_loss += functional.binary_cross_entropy(output, target).data[0]
pred = output.data.max(1)[1]
target_column = target.data.max(1)[1]
correct += pred.eq(target_column).sum()
build_histogram(output, target_column, histogram_data)
if batch_index % log_interval == 0 and batch_index > 1:
LOGGER.info('Test iteration: {}, Correct count: {}'.format(
batch_index, correct))
test_loss /= len(data_loader.dataset)
LOGGER.info(
'Test set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
test_loss, correct, len(data_loader.dataset),
100. * correct / len(data_loader.dataset)))
for key, value in histogram_data.items():
histogram = Histogram(
value,
title='Percentage of Correctly Predicted: {}'.format(key),
bins=10,
number_range=(0.0, 1.0),
histogram_type='numbers')
print(histogram.horizontal())
