:j7 rddlZddlmZddlmZdededeeeffdZ d ded ed edefd Zdededefd Zy)N)Tensor)_check_same_shapepredstargetreturnct||||z jj}|jj}||fS)zUpdate 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 )rabssumrr sum_abs_error sum_scales }/home/mcse/projects/srt_converter/srt-converter-venv/lib/python3.12/site-packages/torchmetrics/functional/regression/wmape.py/_weighted_mean_absolute_percentage_error_updatersHeV$V^((*..0M   "I ) ##r r epsilonc6|tj||z S)zCompute 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 )min)torchclamp)r r rs r0_weighted_mean_absolute_percentage_error_computer*s 5;;yg> >>rc8t||\}}t||S)aCompute 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) )rrr s r'weighted_mean_absolute_percentage_errorr:s$2 OuV\]M9 ;M9 UUr)g-`>) rrtorchmetrics.utilities.checksrtuplerfloatrrrrrs ;$ $ $ 66>$0 ? ? ? ? ? V6V6VfVr