Symmetricmeanabsolutepercentageerror

symmetricmeanabsolutepercentageerror

Metric SymmetricMeanAbsolutePercentageError from torchmetrics (torchmetrics.SymmetricMeanAbsolutePercentageError)

When to invoke this skill

The user has predictions + ground truth and asks to evaluate with SymmetricMeanAbsolutePercentageError, or mentions torchmetrics.SymmetricMeanAbsolutePercentageError directly, or wants the standard torchmetrics implementation.

Reference signature

from torchmetrics import SymmetricMeanAbsolutePercentageError

# SymmetricMeanAbsolutePercentageError(**kwargs: Any) -> None

Library docstring

Compute symmetric mean absolute percentage error (`SMAPE`_).

.. math:: \text{SMAPE} = \frac{2}{n}\sum_1^n\frac{|   y_i - \hat{y_i} |}{\max(| y_i | + | \hat{y_i} |, \epsilon)}

Where :math:`y` is a tensor of target values, and :math:`\hat{y}` is a tensor of predictions.

As input to ``forward`` and ``update`` the metric accepts the following input:

- ``preds`` (:class:`~torch.Tensor`): Predictions from model
- ``target`` (:class:`~torch.Tensor`): Ground truth values

As output of ``forward`` and ``compute`` the metric returns the following output:

- ``smape`` (:class:`~torch.Tensor`): A tensor with non-negative floating point smape value between 0 and 2

Args:
    kwargs: Additional keyword arguments, see :ref:`Metric kwargs` for more info.

Example:
    >>> from torchmetrics.regression import SymmetricMeanAbsolutePercentageError
    >>> target = tensor([1, 10, 1e6])
    >>> preds = tensor([0.9, 15, 1.2e6])
    >>> smape = SymmetricMeanAbsolutePercentageError()
    >>> smape(preds, target)
    tensor(0.2290)

Quick recipe

import torchmetrics as _m
score = _m.SymmetricMeanAbsolutePercentageError(y_true, y_pred)

Don'ts

• Don't reimplement when the library version handles edge cases (NaN, ties, empty inputs) better than a hand-rolled formula.
• Always check the library version's argument order — sklearn is (y_true, y_pred) while torchmetrics is (preds, target).