sktime / sktime/pytorch-forecasting
[Feature Request] Improve predictions on new data for TFT
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 5k
- Forks
- 911
- Avg merge
- 3d 13h
- Merged PRs (30d)
- 12
Description
I think it would help to add a way to easily cast predictions on new data for TFT in a simpler way. One could pass a DF with nans on the decoder unknown dynamic covariates, and get the predictions as a result. To be more specific, this DF has both the future known covariates as well as the past known covariates.
I think this would constitute a great improvement over what is seen in: https://pytorch-forecasting.readthedocs.io/en/stable/tutorials/stallion.html?highlight=predict#Predict-on-new-data
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the linked Stallion tutorial, especially its “Predict on new data” section, and inspect the TFT prediction entry points it uses. Clarify how a dataframe containing future and past known covariates plus NaNs for decoder unknown covariates should be accepted, and define completion as returning predictions through that simpler workflow.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- pandas, python, pytorch
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100