RЦi SSKrSSKrSSKrSSKJr SSKJr S\4SjrSrSSjr SSjr SS jr S r SS \RS \S \S\S\4 SjjrSS \RS \S\S\4Sjjrg)N)nn)_calculate_fan_in_and_fan_outreturncUb[U5S:H$[[S5(a/[R"5nUSL=(a URS:H$g)zACheck if targeting meta device (explicit arg or context manager).Nmetaget_default_deviceF)strhasattrtorchrtype)devicedefault_devices V/var/www/html/ai-image-ml/venv/lib/python3.13/site-packages/timm/layers/weight_init.pyis_meta_devicersQ 6{f$$u*++113T)Kn.A.AV.KK cSnXSU-- :d XSU--:a[R"SSS9 U"X1- U- 5nU"XA- U- 5nURSU-S- SU-S- 5 UR5 UR U[ R "S5-5 URU5 URX4S9 U$)NchS[R"U[R"S5- 5-S- $)N?@)matherfsqrt)xs rnorm_cdf _trunc_normal_..norm_cdfs(TXXa$))B-/00B66rzjmean is more than 2 std from [a, b] in nn.init.trunc_normal_. The distribution of values may be incorrect.) stacklevelr)minmax) warningswarnuniform_erfinv_mul_rradd_clamp_)tensormeanstdabrlus r_trunc_normal_r/s7 1s7{1s7{ 2 E!" $ !(c!"A!(c!"A OOAEAIq1uqy) NN KKdiim#$ KK MMaM Mrc|[R"5 [XX#U5sSSS5 $!,(df  g=f)ahFills the input Tensor with values drawn from a truncated normal distribution. The values are effectively drawn from the normal distribution :math:`\mathcal{N}(\text{mean}, \text{std}^2)` with values outside :math:`[a, b]` redrawn until they are within the bounds. The method used for generating the random values works best when :math:`a \leq \text{mean} \leq b`. NOTE: this impl is similar to the PyTorch trunc_normal_, the bounds [a, b] are applied while sampling the normal with mean/std applied, therefore a, b args should be adjusted to match the range of mean, std args. Args: tensor: an n-dimensional `torch.Tensor` mean: the mean of the normal distribution std: the standard deviation of the normal distribution a: the minimum cutoff value b: the maximum cutoff value Examples: >>> w = torch.empty(3, 5) >>> nn.init.trunc_normal_(w) N)r no_gradr/r(r)r*r+r,s r trunc_normal_r36s#. fCA6 s - ;c[R"5 [USSX45 URU5R U5 SSS5 U$!,(df  U$=f)a~Fills the input Tensor with values drawn from a truncated normal distribution. The values are effectively drawn from the normal distribution :math:`\mathcal{N}(\text{mean}, \text{std}^2)` with values outside :math:`[a, b]` redrawn until they are within the bounds. The method used for generating the random values works best when :math:`a \leq \text{mean} \leq b`. NOTE: this 'tf' variant behaves closer to Tensorflow / JAX impl where the bounds [a, b] are applied when sampling the normal distribution with mean=0, std=1.0 and the result is subsequently scaled and shifted by the mean and std args. Args: tensor: an n-dimensional `torch.Tensor` mean: the mean of the normal distribution std: the standard deviation of the normal distribution a: the minimum cutoff value b: the maximum cutoff value Examples: >>> w = torch.empty(3, 5) >>> nn.init.trunc_normal_(w) rrN)r r1r/r%r&r2s rtrunc_normal_tf_r5QsM. vq#q, Cd#  M  Ms /A AcF[U5upEUS:XaUnOUS:XaUnO US:XaXE-S- nUW- nUS:Xa"[U[R"U5S- S9 gUS:XaB[R "5 UR [R"U5S9 SSS5 gUS :XaK[R"S U-5n[R "5 URU*U5 SSS5 g[S U35e!,(df  g=f!,(df  g=f) Nfan_infan_outfan_avgrtruncated_normalg۶%?r*normaluniformzinvalid distribution ) rr5rrr r1normal_r# ValueError) r(scalemode distributionr7r8denomvariancebounds rvariance_scaling_rGns3F;OF x    !Q&u}H))TYYx%8;M%MN  ! ]]_ NNtyy2N 3_  " !h,' ]]_ OOUFE *_0?@@__s5$DD D D c[USSS9 g)Nr7r:)rBrC)rG)r(s r lecun_normal_rIsf8:LMrmodulename init_bias head_biasclassifier_namec[U[R[R[R[R 45(aUR U5(aT[RRUR5 [RRURU5 g[RRURSS9 [U[R5(a9URb+[RRURU5 ggg[US5(aUR5 gg)N{Gz?r; init_weights) isinstancerLinearConv1dConv2dConv3d startswithinitzeros_weight constant_biasr3r rQ)rJrKrLrMrNs rinit_weight_vitr]s&299biiBIIFGG ??? + + GGNN6== ) GG  fkk9 5 GG ! !&--T ! :&")),,1H!!&++y92I,  ( ( )rcP[U[R5(aURU5(aT[RR UR 5 [RRURU5 g[RRUR 5 URbXSU;a([RRURSS9O([RR UR5 gg[U[R5(aM[UR 5 URb*[RR UR5 gg[US5(aUR5 gg)Nmlpgư>r;rQ)rRrrSrWrXrYrZr[r\xavier_uniform_r?rUrIr rQ)rJrKrMrNs rinit_weight_jaxras &"))$$ ??? + + GGNN6== ) GG  fkk9 5 GG # #FMM 2{{&:?4- 6RWW^^\b\g\gMh' FBII & &fmm$ ;; " GGNN6;; ' #  ( ( )r)rgr)rr7r<)rPrbhead)rbrc)r rr!r torch.nn.initrboolrr/r3r5rGrIModuler floatr]rarrris 7d F76:A2N %     ,%   r