set_determ.py

def set_determ(SEED: int, cudnn_determ: bool = True):
    """Set deterministic modes of `random`, `torch` and `numpy` to ensure reproducibility.

    :param SEED: Random seed.
    :param cudnn_determ: Make cuDNN deterministic.
    """
    import random
    import torch
    import numpy as np

    random.seed(SEED)                               # Set python seed for custom operators.

    torch.manual_seed(SEED)                         # Seed the RNG for all devices (both CPU and CUDA).
    torch.cuda.manual_seed_all(SEED)                # If you are using multi-GPU. In case of one GPU, you can use # torch.cuda.manual_seed(SEED).

    if cudnn_determ:
        torch.backends.cudnn.benchmark = False      # Causes cuDNN to deterministically select an algorithm, 
                                                    # possibly at the cost of reduced performance 
                                                    # (the algorithm itself may be nondeterministic).
        torch.backends.cudnn.deterministic = True   # Causes cuDNN to use a deterministic convolution algorithm, 
                                                    # but may slow down performance.
                                                    # It will not guarantee that your training process is deterministic 
                                                    # if you are using other libraries that may use nondeterministic algorithms 
    else:
        torch.backends.cudnn.enabled = False        # Controls whether cuDNN is enabled or not. 
                                                    # If you want to enable cuDNN, set it to True.
    np.random.RandomState(
        np.random.MT19937(
        np.random.SeedSequence(SEED)))              # If any of the libraries or code rely on NumPy seed the global NumPy RNG.
    np.random.seed(SEED)