Open michelole opened 4 years ago
Running LSTM architectures, even when using pre-trained models, gives different results each time.
Try to reach some determinism. This might require updating dl4j with a newer version (see #78) that contains, e.g., https://github.com/eclipse/deeplearning4j/pull/7087.
Running LSTM architectures, even when using pre-trained models, gives different results each time.
Try to reach some determinism. This might require updating dl4j with a newer version (see #78) that contains, e.g., https://github.com/eclipse/deeplearning4j/pull/7087.