:jtddlmZddlZddlmZddlmcmZdejdejdejfdZ GddejZ y) )OptionalN sequencesweightsreturnc|jd}|jddz}tj||zd|z }tj||jdz }tj|jd}tj||zd|||z z dzz }tj|}tj ||gdS)aHelper function to compute statistics pooling Assumes that weights are already interpolated to match the number of frames in sequences and that they encode the activation of only one speaker. Parameters ---------- sequences : (batch, features, frames) torch.Tensor Sequences of features. weights : (batch, frames) torch.Tensor (Already interpolated) weights. Returns ------- output : (batch, 2 * features) torch.Tensor Concatenation of mean and (unbiased) standard deviation. dimg:0yE>) unsqueezesumtorchsquaresqrtcat)rrv1meandx2v2varstds y/home/mcse/projects/srt_converter/srt-converter-venv/lib/python3.12/site-packages/pyannote/audio/models/blocks/pooling.py_poolrs&A&G  d "B 99Y(a 02 5D ,,y4>>!#44 5C g  " "q " )B ))C'Mq )R"r'\D-@ AC **S/C 99dC[a ((cjeZdZdZ ddej deej dej fdZy) StatsPoolzStatistics pooling Compute temporal mean and (unbiased) standard deviation and returns their concatenation. Reference --------- https://en.wikipedia.org/wiki/Weighted_arithmetic_mean Nrrrc |>|jd}|jdd}tj||gdS|j dk(rd}|j d}nd}|j \}}}|j \}}} || k7rtj||d }tjt|D cgc]} t||dd| ddfc} d} |s| jdS| Scc} w) aForward pass Parameters ---------- sequences : (batch, features, frames) torch.Tensor Sequences of features. weights : (batch, frames) or (batch, speakers, frames) torch.Tensor, optional Compute weighted mean and standard deviation, using provided `weights`. Note ---- `sequences` and `weights` might use a different number of frames, in which case `weights` are interpolated linearly to reach the number of frames in `sequences`. Returns ------- output : (batch, 2 * features) or (batch, speakers, 2 * features) torch.Tensor Concatenation of mean and (unbiased) standard deviation. When `weights` are provided with the `speakers` dimension, `output` is computed for each speaker separately and returned as (batch, speakers, 2 * channel)-shaped tensor. Nr r)r correctionr FTnearest)sizemode) rrrrr r r!F interpolatestackrangersqueeze) selfrrrrhas_speaker_dimension_ num_frames num_speakers num_weightsspeakeroutputs rforwardzStatsPool.forwardLs2 ?>>b>)D--B1-5C99dC[b1 1 ;;=A $) !''A'.G%) !%>>+1j'.||~$<  $mmG*9MG %\2 iGQ!78   %>>a>( (  s D)N)__name__ __module__ __qualname____doc__rTensorrr0rrrr@s< JN66080F6 6rr) typingrrtorch.nnnntorch.nn.functional functionalr#r5rModulerr6rrr=sO. )U\\)ELL)U\\)DB Br