i nddlmZmZddlZddlmZddlmZddlmZdededefd Z dededefd Z d%d ed ed ee edfdefdZ d&dedee dedededededefdZdededeeeffdZd'dedede dedef dZd(dedededefdZd(dedededefdZdedededefd Zd!ed"eed#defd$Zy)))OptionalUnionN)Tensor)Literal)rank_zero_warnxyreturnc|jtjk(s|jtjk(r9|j|jjzj S||jzS)zSafe calculation of matrix multiplication. If input is float16, will cast to float32 for computation and back again. )dtypetorchfloat16floatThalfrr s s/Volumes/fast/ai/experiments/voice-extract-mac/.venv/lib/python3.12/site-packages/torchmetrics/utilities/compute.py _safe_matmulrsT  ww%--177emm#; ACCIIK'--// qss7NcF|tj|z}d||dk(<|S)zCompute x * log(y). Returns 0 if x=0. Example: >>> import torch >>> x = torch.zeros(1) >>> _safe_xlogy(x, 1/x) tensor([0.]) r)r log)rr ress r _safe_xlogyr"s( eiil CCQK Jrnumdenom zero_division)warnnanc|jr|n|j}|jr|n|j}t|ttfs|dk(r{|dk(r#t j |dk(r t d|dk(rdn|}t jd||j|j}t j|dk7||z |St j||S)a Safe division, by preventing division by zero. Function will cast to float if input is not already to secure backwards compatibility. Args: num: numerator tensor denom: denominator tensor, which may contain zeros zero_division: value to replace elements divided by zero Example: >>> import torch >>> num = torch.tensor([1.0, 2.0, 3.0]) >>> denom = torch.tensor([0.0, 1.0, 2.0]) >>> _safe_divide(num, denom) tensor([0.0000, 2.0000, 1.5000]) rrz:Detected zero division in _safe_divide. Setting 0/0 to 0.0r)r device) is_floating_pointr isinstanceintr anyrfullr r"where true_divide)rrrzero_division_tensors r _safe_divider+1s,&&(#ciikC,,.EEKKME-%.-62I F "uyy!'< W X,6M $zz"m399UXU_U_`{{5A:sU{4HII   S% ((rscoreaverage multilabeltpfpfntop_kc||dk(r|S|dk(r||z}n2tj|}|sd||dk(r ||z|zdk(n||zdk(<t||z|jddjdS) NnoneweightedrrT)keepdim)r ones_liker+sum)r,r-r.r/r0r1r2weightss r_adjust_weights_safe_divider<Rs'V+ *r'//%(ILG!BGbLA%bA F %R)F G K KB OOrc|jdkDr|jn|}|jdkDr|jn|}|jdkDs|jdkDr%td|jd|j|j|jk7r-td|jd|j||fS)z Check that auc input is correct.r6zJExpected both `x` and `y` tensor to be 1d, but got tensors with dimension z and zHExpected the same number of elements in `x` and `y` tensor but received )ndimsqueeze ValueErrornumelrs r_auc_format_inputsrB`svvz qAvvz qAvvzQVVaZXYZY_Y_X``efgflflem n   wwyAGGIVWXW^W^W`Vaafghgngngpfq r   a4Kr directionaxisctj5tj||||z}ddd|S#1swYSxYw)zrCompute area under the curve using the trapezoidal rule. Assumes increasing or decreasing order of `x`. dimN)r no_gradtrapz)rr rCrD auc_scores r_auc_compute_without_checkrKps; !KK1$7)C     s ;Areorderc4tj5|rtj|d\}}||}|dd|ddz }|dkjr!|dkj rd}n t dd }t |||cdddS#1swYyxYw) zCompute area under the curve using the trapezoidal rule. Example: >>> import torch >>> x = torch.tensor([1, 2, 3, 4]) >>> y = torch.tensor([1, 2, 3, 4]) >>> _auc_compute(x, y) tensor(7.5000) T)stabler6Nr7rgz_The `x` tensor is neither increasing or decreasing. Try setting the reorder argument to `True`.g?)r rHsortr&allr@rK)rr rLx_idxdxrCs r _auc_computerS{s  zz!D1HAu%A qrUQsV^ F<<>a}}  uI)!Q : s A/BBc<t||\}}t|||S)a8Compute Area Under the Curve (AUC) using the trapezoidal rule. Args: x: x-coordinates, must be either increasing or decreasing y: y-coordinates reorder: if True, will reorder the arrays to make it either increasing or decreasing Return: Tensor containing AUC score )rL)rBrS)rr rLs raucrUs# a #DAq 1g ..rxpc 2t|dd|ddz |dd|ddz }|dd||ddzz }tjtj|dddf|dddfddz }tj|dt |dz }|||z||zS)aiOne-dimensional linear interpolation for monotonically increasing sample points. Returns the one-dimensional piecewise linear interpolation to a function with given discrete data points :math:`(xp, fp)`, evaluated at :math:`x`. Adjusted version of this https://github.com/pytorch/pytorch/issues/50334#issuecomment-1000917964 Args: x: the :math:`x`-coordinates at which to evaluate the interpolated values. xp: the :math:`x`-coordinates of the data points, must be increasing. fp: the :math:`y`-coordinates of the data points, same length as `xp`. Returns: the interpolated values, same size as `x`. Example: >>> x = torch.tensor([0.5, 1.5, 2.5]) >>> xp = torch.tensor([1, 2, 3]) >>> fp = torch.tensor([1, 2, 3]) >>> interp(x, xp, fp) tensor([0.5000, 1.5000, 2.5000]) r6Nr7r)r+r r:geclamplen)rrVr0mbindicess rinterpr^s0 RVb"g%r!"v3B'78A 3B1r#2w;Aii1d7Ra[91=AGkk'1c!fqj1G W:>AgJ &&rtensor normalization)sigmoidsoftmaxc|s|S|jtjdk(rLtj|dk\|dkzs,|dk(r|jntj|d}|S|dk|dkDzj }tj ||dk(rtj||Stj|d|S)aNormalize logits if needed. If input tensor is outside the [0,1] we assume that logits are provided and apply the normalization. Use torch.where to prevent device-host sync. Args: tensor: input tensor that may be logits or probabilities normalization: normalization method, either 'sigmoid' or 'softmax' Returns: normalized tensor if needed Example: >>> import torch >>> tensor = torch.tensor([-1.0, 0.0, 1.0]) >>> normalize_logits_if_needed(tensor, normalization="sigmoid") tensor([0.2689, 0.5000, 0.7311]) >>> tensor = torch.tensor([[-1.0, 0.0, 1.0], [1.0, 0.0, -1.0]]) >>> normalize_logits_if_needed(tensor, normalization="softmax") tensor([[0.0900, 0.2447, 0.6652], [0.6652, 0.2447, 0.0900]]) >>> tensor = torch.tensor([0.0, 0.5, 1.0]) >>> normalize_logits_if_needed(tensor, normalization="sigmoid") tensor([0.0000, 0.5000, 1.0000]) cpurr6rarF)r"r rPrarbr&r()r_r` conditions rnormalize_logits_if_neededrfs8   }} U++yy&A+&A+67)6))CV^^%W]cdIeF 1*!,113I ;;!.)!; f AFv[\A] r)r)r6)r7)F)typingrrr rtyping_extensionsrtorchmetrics.utilitiesrrrrr+strboolr%r<tuplerBrKrSrUr^rfr!rrrms# %1Fv& 6 f  $;>) ) ) 667) )Dop P  P$SM P7; PAG PMS PY_ Phk P  P & V ffn0E  &VSZ`;F;v;;;: /6 /f /t / / 'f'&'f''B*v*hwOcGd>e*jp*r