Thanks to share your code.
I'm trying to fix the seed for reproducibility like these.
os.environ["CUBLAS_WORKSPACE_CONFIG"] = ":4096:8"
torch.use_deterministic_algorithms(True)
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.deterministic = True
torch.backends.cudnn.benchmark = False
But I can't fix my seed in your source code.
Can you advise to me how to fix seed?
Thanks to share your code.
I'm trying to fix the seed for reproducibility like these.
os.environ["CUBLAS_WORKSPACE_CONFIG"] = ":4096:8"
torch.use_deterministic_algorithms(True)
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.deterministic = True
torch.backends.cudnn.benchmark = False
But I can't fix my seed in your source code.
Can you advise to me how to fix seed?