͑igcSSKJr SSKrSSKrSSKJr SSKrSSKr\(aSSKJ r SSK J r SSK J r SSKJrJrJrJr "S S 5r\"5r\R+5S+S j5r\R+5S,S-S jj5r\R+5S,S.S jj5r\R+5S,S-Sjj5r\R+5S/Sj5r\R+5S0Sj5r\R+5S1Sj5r\R+5S2Sj5r\R+5S,S-Sjj5r\R+5S,S3Sjj5r\R+5S,S4Sjj5r \R+5S5S-Sjj5r!\R+5S,S-Sjj5r"\R+5S,S-Sjj5r#\R+5S6S-Sjj5r$\R+5S,S7Sjj5r%\R+5S8S-Sjj5r&\R+5S,S-Sjj5r'\R+5S,S-Sjj5r(\R+5S,S-Sjj5r)\R+5S,S-Sjj5r*S,S9S jjr+SSS!S!S".S:S#jjr,S;S$SSS!S!S%.SS'jjjr.S?S$SSS!S!S%.S@S(jjjr/S=S$SSS!S!S%.S>S)jjjr0S=S$SSS!S!S%.S>S*jjjr1g)A) annotationsN) TYPE_CHECKING)Tensor) PlaceLike)_WindowLiteral)_current_expected_place_get_paddle_placecorein_dynamic_or_pir_modec*\rSrSrSrSSjrSrSrg)WindowFunctionRegister%c0UlgN_functions_dict)selfs ^/var/www/html/banglarbhumi/venv/lib/python3.13/site-packages/paddle/audio/functional/window.py__init__WindowFunctionRegister.__init__&s !Nc^U4SjnU$)Nc>>URnUTRU'U$r)__name__r)funcnamers radd_subfunction8WindowFunctionRegister.register..add_subfunction*s!==D)-D  &Kr)rrrs` rregisterWindowFunctionRegister.register)s  rc URU$rr)rrs rgetWindowFunctionRegister.get1s##D))rrr)r __module__ __qualname____firstlineno__rr!r$__static_attributes__r rrrr%s"*rrc@/nUHn[R"U5(a>[U[5(d)UR [ R "U/U55 M\UR [ R "X155 M [ R"U5$r)npisscalar isinstancestrappendpaddle to_tensorconcat)x data_typelts r_catr78sm A  ;;q>>*Q"4"4 HHV%%qc95 6 HHV%%a3 4  == rc^[U5(a[R"U4US9$[X5up[R"SXS9n[R "XS9n[R "[R"X@S- S- 5SU-US- - SSU-US- - - 5n[XS5$)za Computes the Bartlett window. This function is consistent with scipy.signal.windows.bartlett(). dtyper@) _len_guardsr0ones_extendaranger1where less_equal _truncateMsymr: needs_truncnws r _bartlettrJCs 1~~{{A4u--Q_NA a(A(A !!es]+ a1q5 cAgQ A Q $$rc[U5(a[R"U4US9$[X5up[R"XS9n[R "SXS9n[R"XS9nUS- S- n[R "U[R"SXV- U- S-- 5-5[R "U5- n[Xt5$)z[Compute the Kaiser window. This function is consistent with scipy.signal.windows.kaiser(). r9rr;r<) r=r0r>r?r1r@i0sqrtrC)rEbetarFr:rGrHalpharIs r_kaiserrPXs1~~{{A4u--Q_NA   D .D a(A(A UcME  v{{1e 3;;<<  $ A Q $$rcD[R"/SQUS9n[XXS9$)zQNuttall window. This function is consistent with scipy.signal.windows.nuttall(). )gzD?g;%N?g1|?gC˅?r9)arFr:)r0r1_general_cosine)rErFr:rRs r_nuttallrTos( 4E A 1s 88rc[U[5(a3[R"U[R"US-S- 5-5$[ R"U[ R"[ R "U5S- 5-5$)Nrr;)r-floatmathlogrMr0square)r3s r_acoshrZzs_!UxxDIIadQh//00 ::a&++fmmA&6&:;; <r?r@expabsrC)rEpsigrFr:rGrHrIs r_general_gaussianrhsz1~~{{A4u--Q_NA a(AGs?:A 4&**QW-!a%889A Q $$rc[U5(a[R"U4US9$[X5up[R"[ R *[ R XS9n[R"U4US9n[[U55H#nXaU[R"Xu-5-- nM% [Xd5$)zCompute a generic weighted sum of cosine terms window. This function is consistent with scipy.signal.windows.general_cosine(). r9) r=r0r>r?linspacerWpizerosrangelencosrC)rErRrFr:rGfacrIks rrSrSs1~~{{A4u--Q_NA //477(DGGQ SS[R"TRS5[R"S[R -TRS5-UTS- - S--T- 55--$)Nr;rrr<rx)r0matmul unsqueezerorWrk)rHFmrEmas rW_taylor..Wse1v}} LLO JJq477{R\\!_4AG c8IJQN O    rrc)r=r0r>r?rZrWrkr@empty empty_likermrnprodsqueezerC)rEnbarsllnormrFr:rGBAs2signsm2minumberdenomr~rIscaler|r}s` @@r_taylorrs81~~{{A4u--Q_NA{ sRxAq DGGA qAqDD3J1,, -B q$ ,B taxk /B   b !EE#A#JE!$Q$K bBCGnV[[  q!trCxAo56 6   7 AFH(=$=>>E 3r7Q;  ACR(8$899E++a"&2cr7"2234++a"&21fh<"7789  2" &--1E *+A aQ! n$ U  A Q $$rc[USXS9$)zCompute a Hamming window. The Hamming window is a taper formed by using a raised cosine with non-zero endpoints, optimized to minimize the nearest side lobe. HzG?r9rtrErFr:s r_hammingrs AtS 66rc[USXS9$)z}Compute a Hann window. The Hann window is a taper formed by using a raised cosine or sine-squared with ends that touch zero. rxr9rrs r_hannr s AsC 55rc [U5(a[R"U4US9$US::a[R"U4US9$US:a [XS9$[ X5up[R "SXS9n[ XS- -S- 5nUSUS-nXVS-X- S- nXPU- S- Sn SS[R"[RS SU-U- US- - --5--n [R"URUS9n SS[R"[RS U- S-SU -U- US- - --5--n [R"XU /5n [X5$) zSCompute a Tukey window. The Tukey window is also known as a tapered cosine window. r9rrc)rFr;r<Nrxr`g) r=r0r>rr?r@r]rorWrkshaper2rC)rErOrFr:rGrHwidthn1n2n3w1w2w3rIs r_tukeyrse1~~{{A4u-- z{{A4u-- #Q  Q_NA a(A Q#% &E 1uqy B 19qy1} %B u9q=? B FJJtww"sRx%/?1q5/I*IJKK LB RXXU +B  **TWWu q 038e3Cq1u3M MN O P B  rrl#A Q $$rc[U5(a[R"U4US9$[X5up[R"SXS9US- S- - nSU-U-n[R "US-*U- 5n[ Xt5$)zlCompute a Gaussian window. The Gaussian widows has a Gaussian shape defined by the standard deviation(std). r9rrcr<r)r=r0r>r?r@rdrC)rEstdrFr:rGrHsig2rIs r _gaussianr5sx1~~{{A4u--Q_NA a(AGs?:A s7S=D QT7T>"A Q $$rcHU(aUb [S5e[U5(a[R"U4US9$[ X5upUcUS- S- n[R "SXS9n[R "[R"Xa- 5*U- 5n[Xu5$)z+Compute an exponential (or Poisson) window.z"If sym==True, center must be None.r9r;rr) r^r=r0r>r?r@rdrerC)rEcentertaurFr:rGrHrIs r _exponentialrGs  v!=>>1~~{{A4u--Q_NA ~a%1 a(A FJJqz**S01A Q $$rcp[U5(a[R"U4US9$[X5up[R"SUS-S-S-US9nUS-S:Xa)SU-S- U- n[R "XUSSS2/5nO(SU-US-- n[R "XUSSS2/5n[ XS5$) zCompute a triangular window.r9r;rrrcNr`)r=r0r>r?r@r2rCrDs r_triangr[s1~~{{A4u--Q_NA a!a%A)7A1uz US[A  MM1"g, ' EQW  MM1Bi. ) Q $$rc [U5(a[R"U4US9$[X5up[R"[R "SSXS9SS5nSU- [R "[RU-5-S[R- [R"[RU-5--n[SUS/U5n[XS5$)zVCompute a Bohman window. The Bohman window is the autocorrelation of a cosine window. r9r`r;rcr) r=r0r>r?rerjrorWrksinr7rC)rErFr:rGrprIs r_bohmanrms 1~~{{A4u--Q_NA **V__RA;AbA BC SFJJtww}--dgg   # A1 A aAYA Q $$rc[U/SQXS9$)zCompute a Blackman window. The Blackman window is a taper formed by using the first three terms of a summation of cosines. It was designed to have close to the minimal leakage possible. It is close to optimal, only slightly worse than a Kaiser window. )gzG?rxg{Gz?r9rsrs r _blackmanrs 10# CCrc [U5(a[R"U4US9$[X5up[R"[ R U- [R"SXS9S--5n[XC5$)z,Compute a window with a simple cosine shape.r9rrx) r=r0r>r?rrWrkr@rC)rErFr:rGrIs r_cosinersa1~~{{A4u--Q_NA 477Q;&--1"BS"HIJA Q $$rcU(+nSn[U[5(aUSn[U5S:aUSSnOF[U[5(aUS;a[ SU-S-5eUnO[ [ U5S35e[ RS U-5nU/UQ7n S U0n U"U S U0U D6$![an[ S 5UeSnAff=f) aReturn a window of a given length and type. Args: window (Union[str, Tuple[str, float]]): The window function applied to the signal before the Fourier transform. Supported window functions: 'hamming', 'hann', 'gaussian', 'general_gaussian', 'exponential', 'triang', 'bohman', 'blackman', 'cosine', 'tukey', 'taylor', 'bartlett', 'kaiser', 'nuttall'. win_length (int): Number of samples. fftbins (bool, optional): If True, create a "periodic" window. Otherwise, create a "symmetric" window, for use in filter design. Defaults to True. dtype (str, optional): The data type of the return window. Defaults to 'float64'. Returns: Tensor: The window represented as a tensor. Examples: .. code-block:: python >>> import paddle >>> n_fft = 512 >>> cosine_window = paddle.audio.functional.get_window('cosine', n_fft) >>> std = 7 >>> gaussian_window = paddle.audio.functional.get_window(('gaussian', std), n_fft) r rr;N)gaussian exponentialkaiserzThe 'z6' window needs one or more parameters -- pass a tuple.z! as window type is not supported._zUnknown window type.rFr:) r-tuplernr.r^typewindow_function_registerr$KeyError) window win_lengthfftbinsr:rFargswinstrwinfunceparamskwargss r get_windowrs8+C D&%   v;?!":D FC : :& $33  FDL>)JKLL8*..sV|< 4 FS\F F 2% 26 22 8/0a78sB,, C6 CCFlayoutdevice pin_memory requires_gradcUb[R"S5 [5(GaNUb [U5O [ 5nU(Ga.[ R "5(GaUGb[U[R[R45(d[U[R5(d4[U[R5(a+UR5(a[R"5nOw[U[R5(d4[U[R5(a+UR5(a[R"5nO[!SU35eUR#US9nU(a*[ R "5(aUR%5nUSLaSUlU$)Nz1layout only supports 'strided' in Paddle; ignoredz$Pinning memory is not supported for )rTF)warningswarnr r r r0in_dynamic_moder-r CUDAPinnedPlaceXPUPinnedPlace CUDAPlacePlace is_gpu_placeXPUPlace is_xpu_place RuntimeErrortor stop_gradient)rIrrrrs r_apply_window_postprocessrsQ IJ! f %(*  &&(("--t/B/BC&$..1164::..63F3F3H3H--/FDMM2264::..63F3F3H3H,,.":6(C FAf,,.. LLN Hrfloat32)r:rrrrc^[SXUS9n SupX;- n X,U -- n XU --n[UUUUUS9$)u Compute a generalized Hamming window. Args: window_length (int): The size of the returned window. Must be positive. periodic (bool, optional): If True, returns a window for use as a periodic function; if False, returns a symmetric window. Defaults to True. alpha (float, optional): The coefficient α in the equation above. Defaults to 0.54. beta (float, optional): The coefficient β in the equation above. Defaults to 0.46. dtype (str, optional): The data type of the returned tensor. Defaults to 'float32'. layout (str, optional): Only included for API consistency with PyTorch; ignored in Paddle. Defaults to None. device(PlaceLike|None, optional): The desired device of returned tensor. if None, uses the current device for the default tensor type (see paddle.device.set_device()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types. Default: None. pin_memory(bool, optional): If set, return tensor would be allocated in the pinned memory. Works only for CPU tensors. Default: False requires_grad(bool, optional): If autograd should record operations on the returned tensor. Default: False. Returns: Tensor: A 1-D tensor of shape `(window_length,)` containing the Hamming window. Examples: .. code-block:: python >>> import paddle >>> win = paddle.hamming_window(400, requires_grad=True) >>> win = paddle.hamming_window(256, alpha=0.5, beta=0.5) hammingrr:)rq= ףp?rrr) window_lengthperiodicrOrNr:rrrrw0alpha0beta0rrrIs rhamming_windowrsTN I}e LBMF A F A F A $ #  rc2[SXUS9n[UUUUUS9$)a Compute a Hann window. Args: window_length (int): The size of the returned window. Must be positive. periodic (bool, optional): If True, returns a window for use as a periodic function; if False, returns a symmetric window. Defaults to True. dtype (str, optional): The data type of the returned tensor. Defaults to 'float32'. layout (str, optional): Only included for API consistency with PyTorch; ignored in Paddle. Defaults to None. device(PlaceLike|None, optional): The desired device of returned tensor. if None, uses the current device for the default tensor type (see paddle.device.set_device()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types. Default: None. pin_memory(bool, optional): If set, return tensor would be allocated in the pinned memory. Works only for CPU tensors. Default: False requires_grad(bool, optional): If autograd should record operations on the returned tensor. Default: False. Returns: Tensor: A 1-D tensor of shape `(window_length,)` containing the Hann window. Examples: .. code-block:: python >>> import paddle >>> win = paddle.hann_window(512) >>> win = paddle.hann_window(512, requires_grad=True) hannrrrrrr:rrrrrIs r hann_windowr-s/F 6=%HA $ #  rc6[SU4XUS9n[UUUUUS9$)aO Compute a Kaiser window. Args: window_length (int): The size of the returned window. Must be positive. periodic (bool, optional): If True, returns a window for use as a periodic function; if False, returns a symmetric window. Defaults to True. beta (float, optional): Shape parameter for the window. Defaults to 12.0. dtype (str, optional): The data type of the returned tensor. Defaults to 'float32'. layout (str, optional): Only included for API consistency with PyTorch; ignored in Paddle. Defaults to None. device(PlaceLike|None, optional): The desired device of returned tensor. if None, uses the current device for the default tensor type (see paddle.device.set_device()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types. Default: None. pin_memory(bool, optional): If set, return tensor would be allocated in the pinned memory. Works only for CPU tensors. Default: False requires_grad(bool, optional): If autograd should record operations on the returned tensor. Default: False. Returns: Tensor: A 1-D tensor of shape `(window_length,)` containing the Kaiser window. Examples: .. code-block:: python >>> import paddle >>> win = paddle.kaiser_window(400, beta=8.6) >>> win = paddle.kaiser_window(400, requires_grad=True) rrrr) rrrNr:rrrrrIs r kaiser_windowrZs9J  4- A % #  rc2[SXUS9n[UUUUUS9$)a Compute a Blackman window. Args: window_length (int): The size of the returned window. Must be positive. periodic (bool, optional): If True, returns a window for use as a periodic function; if False, returns a symmetric window. Defaults to True. dtype (str, optional): The data type of the returned tensor. Defaults to 'float32'. layout (str, optional): Only included for API consistency with PyTorch; ignored in Paddle. Defaults to None. device(PlaceLike|None, optional): The desired device of returned tensor. if None, uses the current device for the default tensor type (see paddle.device.set_device()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types. Default: None. pin_memory(bool, optional): If set, return tensor would be allocated in the pinned memory. Works only for CPU tensors. Default: False requires_grad(bool, optional): If autograd should record operations on the returned tensor. Default: False. Returns: Tensor: A 1-D tensor of shape `(window_length,)` containing the Blackman window. Examples: .. code-block:: python >>> import paddle >>> win = paddle.blackman_window(256) >>> win = paddle.blackman_window(256, requires_grad=True) blackmanrrrrs rblackman_windowrs/F :}eLA $ #  rc2[SXUS9n[UUUUUS9$)a Compute a Bartlett window. Args: window_length (int): The size of the returned window. Must be positive. periodic (bool, optional): If True, returns a window for use as a periodic function; if False, returns a symmetric window. Defaults to True. dtype (str, optional): The data type of the returned tensor. Defaults to 'float32'. layout (str, optional): Only included for API consistency with PyTorch; ignored in Paddle. Defaults to None. device(PlaceLike|None, optional): The desired device of returned tensor. if None, uses the current device for the default tensor type (see paddle.device.set_device()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types. Default: None. pin_memory(bool, optional): If set, return tensor would be allocated in the pinned memory. Works only for CPU tensors. Default: False requires_grad(bool, optional): If autograd should record operations on the returned tensor. Default: False. Returns: Tensor: A 1-D tensor of shape `(window_length,)` containing the Bartlett window. Examples: .. code-block:: python >>> import paddle >>> n_fft = 512 >>> win = paddle.bartlett_window(n_fft) >>> win = paddle.bartlett_window(n_fft, requires_grad=True) bartlettrrrrs rbartlett_windowrs/J :}eLA $ #  r)r3z list[Tensor]r4r.returnr)Tfloat64)rEr]rFboolr:r.rr) rEr]rNrVrFrr:r.rr)r3zTensor | floatrr)rEr]rFrrr)rEr]rr)rIrrarrr) rEr]rRz list[float]rFrr:r.rr) rEr]rOrVrFrr:r.rr)TTr)rxTr) rEr]rrVrFrr:r.rr)NrcTr) rz-_WindowLiteral | tuple[_WindowLiteral, float]rr]rrr:r.rr) rIrr str | NonerPlaceLike | Nonerrrrrr)Trr)rr]rrrOrVrNrVr:r.rrrrrrrr)T)rr]rrr:r.rrrrrrrr)Tg(@)rr]rrrNrVr:r.rrrrrrrr)2 __future__rrWrtypingrnumpyr+r0rpaddle._typingrfeatures.layersrpaddle.base.frameworkr r r r rrr!r7rJrPrTrZr?r=rCrhrSrtrrrrrrrrrrrrrrrrrr rrrs#   (0 * * 23""$%""$%%%(""$8A% %%"%25% %%%,""$9%9""$=%= ""$%""$%""$%""$3< %  % %-0 %  %% % ""$;D %  % %!% %58 %  %% % ""$9BF FF#F36F F%F""$FO5% 5%,05%@C5% 5%%5%p""$7%7""$6%6""$6?% % %03% %%%B""$7@% %%!%14% %%%"""$AJ% %'+%;>% %%%&""$%%%"""$%%%"""$D%D""$%%% 33 9333333  33  33r# *  *  *   *  *  *  * ^ 2 #222 2  2  2 2 222n*#***  *  *  ***^. #... .  .  . ...f*#***  *  *  ***^,#,,,  ,  ,  ,,,r