[Bug] Incorrect default assignment logic for n_plotting_samples in DeepAR

[lomesh2312]

Hi there,

I recently noticed a potential inconsistency in how n_plotting_samples is initialized in the DeepAR model. Based on the docstring, n_plotting_samples is expected to default to n_validation_samples if it's provided, or to 100 otherwise.

However, the current implementation in pytorch_forecasting/models/deepar/_deepar.py seems to have this logic inverted:

if n_plotting_samples is None:
    if n_validation_samples is None:
        n_plotting_samples = n_validation_samples  # Sets to None
    else:
        n_plotting_samples = 100  # Sets to 100, regardless of n_validation_samples

Expected Behavior

If the docstring is correct, the logic should probably look like this:

if n_plotting_samples is None:
    n_plotting_samples = n_validation_samples if n_validation_samples is not None else 100

Why this matters

1. Consistency: The current behavior deviates from the documented parameters, which can be confusing for users expecting their plotting sample size to match their validation sample size.
2. Robustness: If n_plotting_samples ends up as None (when n_validation_samples is None), it might cause issues in plotting utilities that expect an integer value.

Minimal Reproduction

A quick simulation of the current logic shows the mismatch:

def simulate_init(n_validation_samples=None, n_plotting_samples=None):
    if n_plotting_samples is None:
        if n_validation_samples is None:
            n_plotting_samples = n_validation_samples
        else:
            n_plotting_samples = 100
    return n_plotting_samples

print(f"Provided n_validation_samples=10: Resulting n_plotting_samples={simulate_init(10)}") # Returns 100
print(f"Provided n_validation_samples=None: Resulting n_plotting_samples={simulate_init(None)}") # Returns None

Suggested Resolution

Update the initialization in DeepAR to align with the docstring:

if n_plotting_samples is None:
    n_plotting_samples = n_validation_samples if n_validation_samples is not None else 100

Before proceeding with a PR, it would be helpful to confirm:

Whether the docstring reflects the intended behavior

Or if the current implementation is correct and the docstring needs updating instead

Happy to open a PR once the intended behavior is clarified.