You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thank you for your very interesting article and model. I would like to add covariates to my target data to improve the predictions. From what I understand from the articles and issues, pre-trained weights are only available for time series without covariates. However, in two issues you mention that it is possible to train the model from scratch with covariates. In these issues, you suggest guidelines on the methodology for adapting the code to add covariates.
I would be very interested in doing this, could you help me with this?
Thank you for your help.
The text was updated successfully, but these errors were encountered:
Hi, just to confirm, I can explain how to adapt the code to add covariates, but you should ensure your pretraining data all have the same covariates, so you can pretrain your model on it. Is this the case? (Just checking)
Hello,
Thank you for your very interesting article and model. I would like to add covariates to my target data to improve the predictions. From what I understand from the articles and issues, pre-trained weights are only available for time series without covariates. However, in two issues you mention that it is possible to train the model from scratch with covariates. In these issues, you suggest guidelines on the methodology for adapting the code to add covariates.
I would be very interested in doing this, could you help me with this?
Thank you for your help.
The text was updated successfully, but these errors were encountered: