[ENH] Add v2 interface support for N-BEATS model

[harshsomankar123-tech]

Reference Issues/PRs

Fixes #2034.

What does this implement/fix? Explain your changes.

This PR adds v2 interface support for the N-BEATS model, following the v2 patterns recently established by DLinear, SAMformer, and TFT in the codebase.

Specifically, this implements:

• Model Core (pytorch_forecasting/models/nbeats/_nbeats_v2.py): * Implemented NBeats inheriting from the experimental BaseModel v2.
  • Added dynamic parameter metadata extraction for lookback (context_length) and horizon (prediction_length).
  • Implemented the forward pass to return prediction, backcast, trend, seasonality, and generic forecasts.
  • Squeezed target/prediction shapes where necessary to maintain strict compatibility with 2D loss metrics (e.g., MAE, QuantileLoss, MASE) and v2 training loops.
  • Integrated an optional backcast loss scaling hook when backcast_loss_ratio > 0.
  • Preserved model interpretation functionality (plot_interpretation) and scaling transformation (transform_output).
• Package Wrapper & Registration (pytorch_forecasting/models/nbeats/_nbeats_pkg_v2.py & __init__.py): * Created package container NBeats_pkg_v2 inheriting from Base_pkg to represent the high-level workflow wrapper.
  • Registered model parameters, metadata, and testing configurations (get_test_train_params).
  • Updated imports and exports for the new v2 components.

What should a reviewer concentrate their feedback on?

• The implementation of the target/prediction shape squeezing to ensure robust compatibility with 2D loss metrics.
• The integration strategy with the new BaseModel (v2) and Base_pkg classes.
• The optional backcast loss scaling hook logic.
• Verification that the new get_test_train_params adequately covers required testing configurations for the package wrapper.

Did you add any tests for the change?

Yes. Robust test coverage was added in tests/test_models/test_nbeats_v2.py.

The new tests cover initialization, forward pass configurations, quantile loss output shapes, step functions (training, validation, test), packaging/fitting pipelines, and interpretation plotting (including matplotlib soft-dependency checking).

Verification: All new v2 tests (7 passes, 1 skipped due to matplotlib missing) and legacy v1 tests (2 passes, 1 skipped) execute successfully without regressions.

Any other comments?

N/A

PR checklist

• The PR title starts with either [ENH], [MNT], [DOC], or [BUG].
• Added/modified tests
• Used pre-commit hooks when committing to ensure that code is compliant with hooks. Install hooks with pre-commit install.
  To run hooks independent of commit, execute pre-commit run --all-files

[codecov]

Codecov Report

❌ Patch coverage is 92.52336% with 16 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@4d8d97c). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
pytorch_forecasting/models/nbeats/_nbeats_v2.py	91.62%	16 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2322   +/-   ##
=======================================
  Coverage        ?   87.41%           
=======================================
  Files           ?      169           
  Lines           ?     9965           
  Branches        ?        0           
=======================================
  Hits            ?     8711           
  Misses          ?     1254           
  Partials        ?        0           

Flag	Coverage Δ
cpu	87.41% <92.52%> (?)
pytest	87.41% <92.52%> (?)

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.