Another foundation model we could interface is LagLLama: https://github.com/time-series-foundation-models/lag-llama

While the weights are hosted on HuggingFace, I suppose that they are not usable within the HFTransformersForecaster, since lagLlama is not implemented using the transformers library. But I suppose, It should not be that hard to implement it. A potential difficulty might be that this package is currently not released on pipit