[DOC] Add DeepAR and NBeats usage examples

[tomatotomata]

Summary

This adds focused doctest-style usage examples for the v1 DeepAR and NBeats models and their package containers.

The model examples show the supported from_dataset entry point and link to the complete tutorials or example script. The package examples document the container-to-model resolution used by the estimator suite.

This is intentionally limited to two v1 models, as requested in issue #2377. The remaining models can follow in separate small pull requests.

Validation

• python -m ruff check pytorch_forecasting/models/deepar/_deepar.py pytorch_forecasting/models/deepar/_deepar_pkg.py pytorch_forecasting/models/nbeats/_nbeats.py pytorch_forecasting/models/nbeats/_nbeats_pkg.py passed.
• python -m compileall -q passed for the four changed files.
• Doctest syntax parsing passed for all four new examples.
• git diff --check passed.
• Full import and doctest execution was not available in this Windows checkout because the lightning dependency is not installed.

Closes #2377 for the DeepAR and NBeats slice.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (main@903e968). Learn more about missing BASE report.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2378   +/-   ##
=======================================
  Coverage        ?   88.82%           
=======================================
  Files           ?      218           
  Lines           ?    11671           
  Branches        ?        0           
=======================================
  Hits            ?    10367           
  Misses          ?     1304           
  Partials        ?        0           

Flag	Coverage Δ
cpu	88.82% <ø> (?)
pytest	88.82% <ø> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[phoeenniixx]

Thanks!
Can you please use dataset and show the complete flow of the model training and prediction in the example?

[tomatotomata]

I expanded both model docstrings into runnable end-to-end examples using the built-in autoregressive dataset. Each now creates a TimeSeriesDataSet, derives a validation set, builds the model with from_dataset, trains for one epoch, and calls predict on the validation loader.

The examples are in commit 7a1ab5f. python -m compileall passes for both changed modules. I could not execute the flow here because this checkout does not have the lightning dependency installed. Let me know what you think.

[phoeenniixx]

The code quality is still failing. Also, in place of adding a complete code block, can you add it command-wise? Using >>>? The doctests check >>> and not the code block

[tomatotomata]

I checked the failing code-quality run and found one concrete ruff E501 in the DeepAR example. I also reworked both model examples from literal code blocks into >>> doctest commands with continuation prompts, so the docstring checks can see the complete dataset, training, and prediction flow. The package-container examples already used >>> and stay unchanged. This is on 906c0e7; the changed lines are now formatted for the repository's 88-character limit. Let me know what you think.

[phoeenniixx]

The tests are still failing!

[tomatotomata]

I traced the failing doctests on the current head and fixed the actual failure in 5ef9bed.

Lightning was writing the progress bar and model summary to the doctest output, so both complete training examples failed even though the training itself finished. I disabled the progress bar and model summary in the example trainers while keeping the full dataset, fit, and prediction flow intact.

Checks on the updated head:

• pre-commit on all four changed model files: passed
• doctests for DeepAR, DeepAR_pkg, NBeats, and NBeats_pkg: 4 passed
• git diff --check: clean

Let me know what you think.