Pranav Prajapati

> I know, I think the conversion needs to be fixed. It should result in `None` or `[None, None]` index names rather than `[None, "timepoints"]` Okay I will take a...

Yes this one is resolved now, Also it would be nice if we update the `CONTRIBUTING_MODELS.md` with the latest steps to contribute a pretrained model in keras_nlp. Let me know...

I think the tests are passing right now, cause lag llama isn't installed yet. I'll start with vendoring the package into sktime.

> Is there no pypi release? No its not available on pypi

> Can you explain what is going on with the conflicts and changes to `gluonts` adapters and checkers? > > I think we need to sort these out before merging,...

> code quality checks seem to be failing, see below yes, will fix it once the PR is ready and the forecaster is in working state

I also encountered the same error while implementing `ELECTRA` in keras. The previous version of `TransformerEncoder` was able to save successfully which I found from this [notebook](https://colab.research.google.com/drive/1SeZWJorKWmwWJax8ORSdxKrxE25BfhHa?usp=sharing) written by @abheesht17...

The respective old weights of MOIRAI with all variations are fetched under my username, along with commits. * [small](https://huggingface.co/Pranavv/moirai-small/tree/temp) * [base](https://huggingface.co/Pranavv/moirai-base/tree/main) * [large](https://huggingface.co/Pranavv/moirai-large/tree/main) I plan to push them to sktime...

I have uploaded the weights on hugging face and wanted a feedback on the README of [model](https://hackmd.io/@TxN8vTB0T6aexuvfRxlqhQ/SkMAFaxs0),

> left some comments, and I see you already addressed them is it good to go now ??