get_more_info(): different seed but same 'train-accuracy', 'train-all-time', 'train-all-time'

[MU-PING]

Why do different trials have same "train-accuracy", "train-per-time" and "train-all-time", it's not intuitive.

Code

print("Trial-1--------------------------------------")
print(self.api.get_more_info(index=arch, dataset='cifar10', hp='200', is_random=777), "\n")
print("Trial-2--------------------------------------")
print(self.api.get_more_info(index=arch, dataset='cifar10', hp='200', is_random=888), "\n")
print("Trial-3--------------------------------------")
print(self.api.get_more_info(index=arch, dataset='cifar10', hp='200', is_random=999), "\n")

Output

Trial-1--------------------------------------
{'train-loss': 0.0019287928102910518, 'train-accuracy': 99.986, 'train-per-time': 28.7974933385849, 'train-all-time': 5759.49866771698, 'comment': 'In this dict, train-loss/accuracy/time is the metric on the train+valid sets of CIFAR-10. The test-loss/accuracy/time is the performance of the CIFAR-10 test set after training on the train+valid sets by 200 epochs. The per-time and total-time indicate the per epoch and total time costs, respectively.', 'test-loss': 0.273262268447876, 'test-accuracy': 94.26, 'test-per-time': 1.8184241397040222, 'test-all-time': 363.68482794080444}

Trial-2--------------------------------------
{'train-loss': 0.0019176186656206847, 'train-accuracy': 99.986, 'train-per-time': 28.7974933385849, 'train-all-time': 5759.49866771698, 'comment': 'In this dict, train-loss/accuracy/time is the metric on the train+valid sets of CIFAR-10. The test-loss/accuracy/time is the performance of the CIFAR-10 test set after training on the train+valid sets by 200 epochs. The per-time and total-time indicate the per epoch and total time costs, respectively.', 'test-loss': 0.25168872060775754, 'test-accuracy': 94.68, 'test-per-time': 1.8184241397040222, 'test-all-time': 363.68482794080444}

Trial-3--------------------------------------
{'train-loss': 0.0018634130948781966, 'train-accuracy': 99.986, 'train-per-time': 28.7974933385849, 'train-all-time': 5759.49866771698, 'comment': 'In this dict, train-loss/accuracy/time is the metric on the train+valid sets of CIFAR-10. The test-loss/accuracy/time is the performance of the CIFAR-10 test set after training on the train+valid sets by 200 epochs. The per-time and total-time indicate the per epoch and total time costs, respectively.', 'test-loss': 0.26081861381530763, 'test-accuracy': 94.18, 'test-per-time': 1.8184241397040222, 'test-all-time': 363.68482794080444}