io# ddlZddlmZddlmZmZmZmZddlZddl m Z ddlm Z ddl m Z ddlmZmZddlmZd Zd ee ee fd e fd Zd e d e fd Zd e d e fdZd e d e fdZd e d e fdZd ed efdZd ed eeeffdZ d,de deed e fdZ d-d e deded e fdZ!d-de deded e fdZ"d.d e ded e fdZ#d e d e fdZ$ded efd Z%d,d e d!eed e fd"Z&d/d e deed#eejNd e fd$Z(d e d e fd%Z)d&e d'e d efd(Z*d e d)e d*e d e fd+Z+y)0N)Sequence)AnyListOptionalUnion)apply_to_collection)Tensor)TorchMetricsUserWarning)_TORCH_LESS_THAN_2_6_XLA_AVAILABLE)rank_zero_warngư>xreturnct|tjr|S|Dcgc]7}|jdk(r |jdk(r|j dn|9}}|s t dtj|dScc}w)z'Concatenation along the zero dimension.rzNo samples to concatenatedim) isinstancetorchr numelndim unsqueeze ValueErrorcat)rys p/Volumes/fast/ai/experiments/voice-extract-mac/.venv/lib/python3.12/site-packages/torchmetrics/utilities/data.py dim_zero_catrsr!U\\"JKL!Q1779>affkQq @!AL 455 99QA  Ms>> x = torch.tensor([1, 2, 3]) >>> to_onehot(x) tensor([[0, 1, 0, 0], [0, 0, 1, 0], [0, 0, 0, 1]]) Nrr)dtypedevice?) intr'detachr1rzerosshaper@rAlongr expand_asscatter_)r=r> tensor_onehotindexs r to_onehotrLQs*,**,335::zz#q!$$ 66A36  ' ''r" prob_tensorrUctj|tj}|dk(r4|j||j |dd|jS|j|t |||d|jS)awConvert a probability tensor to binary by selecting top-k the highest entries. Args: prob_tensor: dense tensor of shape ``[..., C, ...]``, where ``C`` is in the position defined by the ``dim`` argument topk: number of the highest entries to turn into 1s dim: dimension on which to compare entries Returns: A binary tensor of the same shape as the input tensor of type ``torch.int32`` Example: >>> x = torch.tensor([[1.1, 2.0, 3.0], [2.0, 1.0, 0.5]]) >>> select_topk(x, topk=2) tensor([[0, 1, 1], [1, 1, 0]], dtype=torch.int32) r@rT)rkeepdimrBrO)r zeros_likerCrIargmaxrX)rYrUr topk_tensors r select_topkr`|s&"";eii@K qyS+"4"4d"4"KSQ ??  S"D[TX^a"bdgh ?? r" argmax_dimc0tj||S)awConvert a tensor of probabilities to a dense label tensor. Args: x: probabilities to get the categorical label [N, d1, d2, ...] argmax_dim: dimension to apply Return: A tensor with categorical labels [N, d2, ...] Example: >>> x = torch.tensor([[0.2, 0.5], [0.9, 0.1]]) >>> to_categorical(x) tensor([1, 0]) r)rr^)rras rto_categoricalrcs <<z **r"cL|jdk(r|jS|S)Nr)rsqueezer s r_squeeze_scalar_element_tensorrfs'')q.199;/a/r"datac,t|ttSN)rr rf)rgs r_squeeze_if_scalarrjs tV-K LLr" minlengthc|ttj|}tjsts |j rptj ||jjt|d}tj|jdd|jdStj||S)aImplement custom bincount. PyTorch currently does not support ``torch.bincount`` when running in deterministic mode on GPU or when running MPS devices or when running on XLA device. This implementation therefore falls back to using a combination of `torch.arange` and `torch.eq` in these scenarios. A small performance hit can expected and higher memory consumption as `[batch_size, mincount]` tensor needs to be initialized compared to native ``torch.bincount``. Args: x: tensor to count minlength: minimum length to count Returns: Number of occurrences for each unique element in x Example: >>> x = torch.tensor([0,0,0,1,1,2,2,2,2]) >>> _bincount(x, minlength=3) tensor([3, 2, 4]) )rArrrrk) lenrunique$are_deterministic_algorithms_enabledr is_mpsarangerArepeateqreshaperbincount)rrkmeshs r _bincountrys* Q(  113~||Iahh7>>s1vqIxx "a($/333:: >>!y 11r"r@cdtjxr|jxr|j}tr_|r]t j dk7rJtdt|jj||j|jStj|||S)z\Implement custom cumulative summation for Torch versions which does not support it natively.win32zYou are trying to use a metric in deterministic mode on GPU that uses `torch.cumsum`, which is currently not supported. The tensor will be copied to the CPU memory to compute it and then copied back to GPU. Expect some slowdowns.)rr@) rrqrQis_floating_pointr sysplatformr r cpucumsumtorA)rrr@is_cuda_fp_deterministics r_cumsumrs$IIKsPQPYPYs^_^q^q^s 8S\\W=T & $  uuw~~#U~366qxx@@ <<s% 00r"cbtj|d\}}t|t|S)zSimilar to `_bincount`, but works also with tensor that do not contain continuous values. Args: x: tensor to count Returns: Number of occurrences for each unique element in x T)return_inversern)rrpryro)runique_xinverse_indicess r_flexible_bincountrs*!& Qt DHo _H >>r"tensor1tensor2c|j|jk7r|j|j}tj||S)z:Wrap torch.allclose to be robust towards dtype difference.r[)r@rrallclose)rrs rrrs7}} %**7==*1 >>'7 ++r"xpfpctj|}||}||}|dd|ddz |dd|ddz z }tj||dz }tj|dt |dz }|||||||z zzS)zInterpolation function comparable to numpy.interp. Args: x: x-coordinates where to evaluate the interpolated values xp: x-coordinates of the data points fp: y-coordinates of the data points rNrmr)rrR searchsortedclampro)rrrsorted_indicesslopesrVs rinterprs]]2&N N B N Bfr#2w2ab6BsG#3 4F  Q'!+Gkk'1c&kAo6G g;A7 O< <rs $-- 3EO8  E&$v,./FFv  V  &F&v& &F&v& 77T7  T eD$J/ &"& 1 1# 1  1F(&(S(3(v(V3V6+f+#+f+&0f00MSMSM22HSM2V2> 1v 1HSM 1hu{{6K 1W] 1 ?& ?V ?,f,v,$,=f=&=f==r"