More time steps incurs longer training time, which above certain limit becomes impractical. What is the recommended maximum limit of the number of time steps for RNN or LSTM? I'm using a powerful workstation with GPU, but not a server. Around a few hundreds time steps? A few thousands time steps?

  • 1
    $\begingroup$ This has zero context. What is your dataset? $\endgroup$ – Alex R. Jan 24 at 6:24
  • $\begingroup$ as a very rough rule of thumb, RNNs struggle to learn dependencies more than 50 timesteps long. LSTMs struggle with more than maybe 500. $\endgroup$ – shimao Jul 11 at 5:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.