Fixed seed for reproducing the results

[JunwenBai]

Has anyone found a way (or a set of fixed random seed) to reproduce the results? I tried to fix every random seed, but it seems that there still exists randomness. Based on my experience, more than half of my runs fail to escape 0.5 accuracy, which is quite annoying. Any suggestions would be helpful! Thx!

[943fansi]

Has anyone found a way (or a set of fixed random seed) to reproduce the results? I tried to fix every random seed, but it seems that there still exists randomness. Based on my experience, more than half of my runs fail to escape 0.5 accuracy, which is quite annoying. Any suggestions would be helpful! Thx!

Add np.random.seed(123) in data_utils.py