search

lukasruff / Deep-SVDD-PyTorch

A PyTorch implementation of the Deep SVDD anomaly detection method
MIT License
698 stars 197 forks source link

Seed does not work for autoencoder (not reproducible results) #29

Closed omid-ghozatlou closed 2 years ago

omid-ghozatlou commented 2 years ago

As you when we set seed to any number except -1, the result must not change by rerunning. however, I set the seed to 10 and use preparing. So the results of both pretraining and training change widely. It is worth mentioning that if you set pretraining False, the network is trained without any pre-trained weights and the result of many times training is the same for the specific seed value. (of course except -1) Does anybody know why seed (use for reproducibility) does not work for autoencoder?

omid-ghozatlou commented 2 years ago

I solved the problem by adding folllowing lines for Cuda:

        torch.manual_seed(cfg.settings['seed'])
        torch.cuda.manual_seed(cfg.settings['seed'])
        torch.cuda.manual_seed_all(cfg.settings['seed'])
        torch.backends.cudnn.deterministic = True 
        torch.backends.cudnn.benchmark = False