[FIX] fh range should begin at 1; dependencies

[mabuimo]

What does this implement/fix?

• The range for the fh argument should begin at 1, as 0 would include the last point of the training dataset in the cross-validation.
• Added uv as an optional dependency manager.
• The dependencies of requirements.txt were incomplete.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[benHeid]

Thanks for the PR. I have a small suggestion to the code. The only major comment is could you try to rerun notebook 3. It doesn't run completely.

[benHeid]

Can you execute this notebook again and check that Moirai_small runs. The following cell seems to fail

results = benchmark.run("energy_benchmarking.json", force_rerun =True)

[benHeid]

I am unsure, do you think it is useful to push the result file?

[mabuimo]

Probably it is not very useful, we can remove it.

[mabuimo]

Regarding the lock file, I found this discussion astral-sh/uv#10730 but I agree with you that we don't need to include it.

[mabuimo]

Btw @benHeid testing TimeLLMForecast fails with the example of notebook 3.

Code to add to replicate the issue:

from sktime.forecasting.time_llm import TimeLLMForecaster
 benchmark.add_estimator(
     TimeLLMForecaster()
 )

[mabuimo]

@benHeid ready for final review. All comments addressed, thank you.

[fkiraly]

Nice, thanks!

(@benHeid is currently off sktime, so I am reviewing)

Could you kindly do a clean re-execution of the notebooks? I.e., reset and execute all of them.

[mabuimo]

Hi @fkiraly in theory the last PR includes a full re-run of all cells. Isn't that the case? Thanks.

[fkiraly]

I think you need to reset the notebook and then re-execute - if the first cell shows execution number 82 and not 1, it has not been reset properly