[XXXXXQ-0206]

predict() without series falls back to training_series (documented in the docstring). After fit(train, val_series=val) where val continues train, that fallback excludes the validation part, so the forecast restarts inside the validation period - silently.

Reproduction (0.48.1, master dcdde24, CPU):

train, val = series[:70], series[70:]
model.fit(train, val_series=val)
model.predict(n=3).start_time()                 # index 70  == train.end + freq (re-forecasts the val range)
model.predict(n=3, series=series).start_time()  # index 100 == val.end + freq

This PR keeps the fallback and the returned start point unchanged (documented behaviour others rely on) and only makes it visible: fit() records whether a val_series reached beyond training_series, and predict() warns when it falls back, naming the training series, the resulting start and how to avoid it. The flag is recorded before the model is saved during fit(), so it also survives save()/load() and load_from_checkpoint(). A val_series inside the training series, a fit() without val_series, explicit series=..., historical_forecasts() and backtest() are unaffected.

Test: test_predict_without_series_warns_when_fitted_with_val_series (no val_series / extending it / equal to it).

Verification (local, CPU torch 2.14.1): new test: 3 cases pass; with torch_forecasting_model.py restored to base they fail (1 failed, 2 passed); the full test file 363 passed; pre-commit passes.

AI disclosure: prepared with AI assistance under the account owner's direction; reproduction, tests and verification commands were run locally, outputs quoted verbatim.

Fixes #3278

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing. Review the Copilot code review billing settings.

[dennisbader]

Thanks for the PR @XXXXXQ-0206. I'll close this PR with the answer I gave here.