o yi @sddlmZddlZddlmZddlmZddlmZdedefdd Zdedefd d Z d ed ede deeeffddZ dd ed ede defddZ d ed edefddZdS))TupleN)Tensor) _check_data_shape_to_num_outputs)_check_same_shapedatareturncCs|}|d}ttjdg|jd|dd|ddkg}||}tt|tj|gg|jdg }|dd|dd}|dk}||S)z_find and return values which have repeats i.e. the same value are more than once in the tensor.rT)deviceN) detachclonesorttorchcattensorrnonzeronumelflatten)rtempchangeunique change_idxfreqatleast2r_/home/ubuntu/.local/lib/python3.10/site-packages/torchmetrics/functional/regression/spearman.py _find_repeatss  0,rcCsn|}t|}|}tjd|d|j|jd||d|<t|}|D]}||k}||||<q&|S)aCalculate the rank for each element of a tensor. The rank refers to the indices of an element in the corresponding sorted tensor (starting from 1). Duplicates of the same value will be assigned the mean of their rank. Adopted from `Rank of element tensor`_ r )dtyperN) rr empty_likeargsortarangerrrmean)rnrankidxrepeatsr conditionrrr _rank_data$s &r(predstarget num_outputscCs6|r|s tdt||t|||||fS)zUpdates and returns variables required to compute Spearman Correlation Coefficient. Checks for same shape and type of input tensors. Args: preds: Predicted tensor target: Ground truth tensor zhExpected `preds` and `target` both to be floating point tensors, but got {pred.dtype} and {target.dtype})is_floating_point TypeErrorrr)r)r*r+rrr_spearman_corrcoef_update8s   r.ư>epsc Cs|jdkrt|}t|}ntdd|jDj}tdd|jDj}||d}||d}||d}t||d}t||d}||||}t|ddS)aComputes Spearman Correlation Coefficient. Args: preds: Predicted tensor target: Ground truth tensor eps: Avoids ``ZeroDivisionError``. Example: >>> target = torch.tensor([3, -0.5, 2, 7]) >>> preds = torch.tensor([2.5, 0.0, 2, 8]) >>> preds, target = _spearman_corrcoef_update(preds, target, num_outputs=1) >>> _spearman_corrcoef_compute(preds, target) tensor(1.0000) r cSg|]}t|qSrr().0prrr _z._spearman_corrcoef_compute..cSr1rr2)r3trrrr5`r6rgg?)ndimr(rstackTr!sqrtclamp) r)r*r0 preds_diff target_diffcov preds_std target_stdcorrcoefrrr_spearman_corrcoef_computeLs  rCcCs0t|||jdkr dn|jdd\}}t||S)aComputes `spearmans rank correlation coefficient`_: .. math: r_s = = \frac{cov(rg_x, rg_y)}{\sigma_{rg_x} * \sigma_{rg_y}} where :math:`rg_x` and :math:`rg_y` are the rank associated to the variables x and y. Spearmans correlations coefficient corresponds to the standard pearsons correlation coefficient calculated on the rank variables. Args: preds: estimated scores target: ground truth scores Example (single output regression): >>> from torchmetrics.functional import spearman_corrcoef >>> target = torch.tensor([3, -0.5, 2, 7]) >>> preds = torch.tensor([2.5, 0.0, 2, 8]) >>> spearman_corrcoef(preds, target) tensor(1.0000) Example (multi output regression): >>> from torchmetrics.functional import spearman_corrcoef >>> target = torch.tensor([[3, -0.5], [2, 7]]) >>> preds = torch.tensor([[2.5, 0.0], [2, 8]]) >>> spearman_corrcoef(preds, target) tensor([1.0000, 1.0000]) r r )r+)r.r8shaperC)r)r*rrrspearman_corrcoefms& rE)r/)typingrrr(torchmetrics.functional.regression.utilsrtorchmetrics.utilities.checksrrr(intr.floatrCrErrrrs     "!