Setting seed in a ML simulation

A code snippet and tips for setting the seed in a ML python script

#machinelearning

Use the following code snippet to set the seed in a python machine learning script.

seed = 0       
torch.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
torch.cuda.manual_seed(seed)
np.random.seed(seed)
random.seed(seed)
torch.backends.cudnn.deterministic=True
torch.backends.cudnn.benchmarks=False
os.environ['PYTHONHASHSEED'] = str(seed)

Tips to make a code deterministic

  • If try-except blocks behave differently for the same seed in different simulation runs, then the results may tend to differ (e.g. instances where different simulation runs with the same seed raise different errors in try).