在进行网络训练的时候为了之后可以成功复现当前结果,需要设置随机种子
废话少说,直接上函数,在train.py最初调用此函数即可
def init_seeds(seed=0): random.seed(seed) # seed for module random np.random.seed(seed) # seed for numpy torch.manual_seed(seed) # seed for PyTorch CPU torch.cuda.manual_seed(seed) # seed for current PyTorch GPU torch.cuda.manual_seed_all(seed) # seed for all PyTorch GPUs if seed == 0: # if True, causes cuDNN to only use deterministic convolution algorithms. torch.backends.cudnn.deterministic = True # if True, causes cuDNN to benchmark multiple convolution algorithms and select the fastest. torch.backends.cudnn.benchmark = False
欢迎分享,转载请注明来源:内存溢出
评论列表(0条)