nateemma
commented
1 year ago The training is done in backtesting only if the model file does not already exist. If it does, then I assume you have already trained the model and I just use it. The reason I do it this way is that training a neural network requires a lot of samples for training, and also takes a very long time (hours), so I cannot really run it in real time. I could incrementally train, but that still likely takes too long, and I didn't want to play with running in a separate thread - it would be better to use freqai (which I will look at soon).
So, what I do is run backtesting over a very long time period to create the model, then just run backtesting/hyperopt/dryrun using the model that this created.The model files should be in github - some of them are so big that I cannot upload them though.
Hope that helps
Cheers
Phil
On Tue, May 2, 2023 at 1:50 PM Sergey Sokolov @.***> wrote:
Tell me how to start training a model I decided to use NNPredict I understand that the training of the model here is not as in the classic freqtrade example, because when you run the backtest the training stage of the model does not occur Logs...
Adding technical indicators... training model... fitting model...
— Reply to this email directly, view it on GitHub https://github.com/nateemma/strategies/issues/23, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABD4X53SE4XLT3ZZ3B6BNJDXEFXP5ANCNFSM6AAAAAAXTRV7RQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Tell me how to start training a model I decided to use NNPredict I understand that the training of the model here is not as in the classic freqtrade example, because when you run the backtest the training stage of the model does not occur Logs...