Add Time-LLM

[marcopeix]

Add TimeLLM to neuralforecast.

User can load an LLM from HugginFace and use it for time series forecasting.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[CLAassistant]

All committers have signed the CLA.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ marcopeix
❌ Marco Peixeiro

---

Marco Peixeiro seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[AzulGarza]

awesome @marcopeix! as a first piece of feedback, could you include a tutorial regarding how to use TimeLLM? it can be based on experiments/test_timellm/test_timellm.py, also the tests should be included in models.timellm.ipynb, (throwing an error if transformers is not installed).

also we should include transformers in the dev requirements to prevent tests from breaking.

[marcopeix]

I added tests to the notebook
GPT-2 is the default LLM if the user does not specify another model
The usage example acts as a tutorial for using Time-LLM
Added transformers as a dependency

[AzulGarza]

thanks @marcopeix!

could you also include transformers in the dev_requirements part of the setttings.ini file?

also running nbdev_install_git_hooks and nbdev_clean should solve the nb_sync action.

[marcopeix]

Updated PR according to Azul's comments:

• Better documentation of default behaviour
• Remove transformers from core dependencies and added to dev dependencies
• Flag to check if transformers is installed and use GPT-2 as default LLM
• Remove test file from experiments folder since the script exists in the notebook

[AzulGarza]

awesome @marcopeix! thank you 🎉

let's wait for the review of @jmoralez or @cchallu to merge it to main❤️