tsai

tsai copied to clipboard

Great package!

I was wondering whether you would consider integrating with sktime or pytorch-forecasting?

To me, it seems like most of tsai is to sktime.classification what pytorch-forecasting is to sktime.forecasting.

You also have forecasting models, which is an intersection with pytorch-forecasting.

That is, we could talk about:

• interfacing the tsai classifiers as sktime classifiers. They would remain in tsai and you would be credited as authors, see for instance the estimator overview: https://www.sktime.net/en/latest/estimator_overview.html - this would give tsai more exposure, especially since you seem to have the largest array of deep learning classifiers.
• joining the current efforts to build a unified deep learning layer. There are some very interesting questions in design and architecture, see here the current state of discussions for pytorch-forecasting: https://github.com/sktime/pytorch-forecasting/issues/1736

sktime issue opened: https://github.com/sktime/sktime/issues/7885 pytorch-forecasting issue opened: https://github.com/sktime/pytorch-forecasting/issues/1779