venv/lib/python3.11/site-packages/torchmetrics/functional/regression/wmape.py

# Copyright The Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import torch
from torch import Tensor

from torchmetrics.utilities.checks import _check_same_shape


def _weighted_mean_absolute_percentage_error_update(
    preds: Tensor,
    target: Tensor,
) -> tuple[Tensor, Tensor]:
    """Update and returns variables required to compute Weighted Absolute Percentage Error.

    Check for same shape of input tensors.

    Args:
        preds: Predicted tensor
        target: Ground truth tensor

    """
    _check_same_shape(preds, target)

    sum_abs_error = (preds - target).abs().sum()
    sum_scale = target.abs().sum()

    return sum_abs_error, sum_scale


def _weighted_mean_absolute_percentage_error_compute(
    sum_abs_error: Tensor,
    sum_scale: Tensor,
    epsilon: float = 1.17e-06,
) -> Tensor:
    """Compute Weighted Absolute Percentage Error.

    Args:
        sum_abs_error: scalar with sum of absolute errors
        sum_scale: scalar with sum of target values
        epsilon: small float to prevent division by zero

    """
    return sum_abs_error / torch.clamp(sum_scale, min=epsilon)


def weighted_mean_absolute_percentage_error(preds: Tensor, target: Tensor) -> Tensor:
    r"""Compute weighted mean absolute percentage error (`WMAPE`_).

    The output of WMAPE metric is a non-negative floating point, where the optimal value is 0. It is computes as:

    .. math::
        \text{WMAPE} = \frac{\sum_{t=1}^n | y_t - \hat{y}_t | }{\sum_{t=1}^n |y_t| }

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

    Args:
        preds: estimated labels
        target: ground truth labels

    Return:
        Tensor with WMAPE.

    Example:
        >>> from torch import randn
        >>> preds = randn(20,)
        >>> target = randn(20,)
        >>> weighted_mean_absolute_percentage_error(preds, target)
        tensor(1.3967)

    """
    sum_abs_error, sum_scale = _weighted_mean_absolute_percentage_error_update(preds, target)
    return _weighted_mean_absolute_percentage_error_compute(sum_abs_error, sum_scale)