Meansquaredlogerror

meansquaredlogerror

Metric MeanSquaredLogError from torchmetrics (torchmetrics.MeanSquaredLogError)

When to invoke this skill

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

Reference signature

from torchmetrics import MeanSquaredLogError

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

Library docstring

Compute `mean squared logarithmic error`_ (MSLE).

.. math:: \text{MSLE} = \frac{1}{N}\sum_i^N (\log_e(1 + y_i) - \log_e(1 + \hat{y_i}))^2

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:

- ``mean_squared_log_error`` (:class:`~torch.Tensor`): A tensor with the mean squared log error

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

Example:
    >>> from torch import tensor
    >>> from torchmetrics.regression import MeanSquaredLogError
    >>> target = tensor([2.5, 5, 4, 8])
    >>> preds = tensor([3, 5, 2.5, 7])
    >>> mean_squared_log_error = MeanSquaredLogError()
    >>> mean_squared_log_error(preds, target)
    tensor(0.0397)

.. attention::
    Half precision is only support on GPU for this metric.

Quick recipe

import torchmetrics as _m
score = _m.MeanSquaredLogError(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).