def multi_test_epoch_end(self, outputs, dataloader_idx: int = 0):
test_loss_mean = torch.stack([x['test_loss'] for x in outputs]).mean()
correct_counts = torch.stack([x['test_correct_counts'].unsqueeze(0) for x in outputs])
total_counts = torch.stack([x['test_total_counts'].unsqueeze(0) for x in outputs])
topk_scores = compute_topk_accuracy(correct_counts, total_counts)
tensorboard_log = {'test_loss': test_loss_mean}
for top_k, score in zip(self._accuracy.top_k, topk_scores):
tensorboard_log['test_epoch_top@{}'.format(top_k)] = score
return {'log': tensorboard_log}
def multi_test_epoch_end(self, outputs, dataloader_idx: int = 0):
self.val_loss_mean = torch.stack([x['test_loss']
for x in outputs]).mean()
correct_counts = torch.stack(
[x['test_correct_counts'] for x in outputs])
total_counts = torch.stack([x['test_total_counts'] for x in outputs])
topk_scores = compute_topk_accuracy(correct_counts, total_counts)
logging.info("test_loss: {:.3f}".format(self.val_loss_mean))
tensorboard_log = {'test_loss': self.val_loss_mean}
for top_k, score in zip(self._accuracy.top_k, topk_scores):
tensorboard_log['test_epoch_top@{}'.format(top_k)] = score
self.accuracy = score * 100
return {'log': tensorboard_log}