Random Seed Initialization for Reproducibility ...

1. Try

seed = 0 # any number you want
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.deterministic = True

2. More strictly,

seed = 0  # any number you want
os.environ['PYTHONHASHSEED'] = str(seed)
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.deterministic = True

#References

https://pytorch.org/docs/stable/notes/randomness.html
https://discuss.pytorch.org/t/random-seed-initialization/7854/17
https://discuss.pytorch.org/t/how-to-get-deterministic-behavior/18177/11