[QUESTION] The whole history sequence predict, is there way that much faster that historical_forecast

[ruixu2]

I tried to history forecast and compute metrics including mae mape and r2 score.

But as mentiond #2780 , it was too too slow using backtest or historical_forecast functions and i can not find the reason.

So is there other ways to finish the history forecast.

• In pytorch's deep learning model, the torch_metrics can input and training with the metric validation.
• But in other models,

Deep learning train, 8 iter/s

backtest is about #1iter/s

[dennisbader]

Let's continue this discussion in #2780. I'll close this one to avoid duplicates.