search

time-series-foundation-models / lag-llama

Lag-Llama: Towards Foundation Models for Probabilistic Time Series Forecasting
Apache License 2.0
1.09k stars 121 forks source link

True Future Predictions #19

Closed bradski7 closed 3 months ago

bradski7 commented 4 months ago

Unable to make future predictions for future timesteps beyond the validation window where there are no known target values. Is that something this model is capable of?

ashok-arjun commented 4 months ago

Yes of course - that's something any model should support :) So far the demo was just demonstrating the model's predictions compared to the ground truth.

We just need to write some code to support making predictions beyond the "test set" in the demo.

Until then, one way to do this would be: you extend your dataset to have the required timestamps at the end along with dummy values (no known target values), and then just make predictions there.

It's pretty easy to support this properly; I'll update here once we have a better way to do this.

kashif commented 3 months ago

@bradski7 yes since the estimator is a gluonts estimator class, you can call predict() method on the predictor

ashok-arjun commented 3 months ago

Hi @bradski7, is this resolved?

bradski7 commented 3 months ago

Yes, thanks for the help