3j `SSKrSSKrSSKrSSKJr S\R S\S\S\S\4 SjrSS\R S\S\S\S\S \R 4 S jjrSS\R S\S\S\S\S \R 4 S jjr SS\R S \S \ S\ 4Sjjr S\R 4Sjr g)N)_calculate_fan_in_and_fan_outtensormeanstdabcS[4SjnXSU-- :d XSU--:a[R"SSS9 [R"5 U"X1- U- 5nU"XA- U- 5nUR SU-S- SU-S- 5 UR 5 URU[R"S5-5 URU5 URX4S9 UsSSS5 $!,(df  g=f) NxchS[R"U[R"S5- 5-S- $)N?@)matherfsqrt)r s Z/home/wildlama/miniconda3/lib/python3.13/site-packages/spandrel/util/timm/__weight_init.pynorm_cdf(_no_grad_trunc_normal_..norm_cdf s(dhhq499S>122c99zjmean is more than 2 std from [a, b] in nn.init.trunc_normal_. The distribution of values may be incorrect.) stacklevelr )minmax) floatwarningswarntorchno_graduniform_erfinv_mul_rradd_clamp_)rrrrrrlus r_no_grad_trunc_normal_r&s :E: 1s7{1s7{ 2  ;  ah#% & ah#% & A 1q519-   C$))C.() D  ! #+ s BC(( C6returnc[XX#U5$)ak Fills the input Tensor with values drawn from a truncated normal distribution. The values are effectively drawn from the normal distribution :math:`\mathcal{N}(\text{mean}, \text{std}^2)` with values outside :math:`[a, b]` redrawn until they are within the bounds. The method used for generating the random values works best when :math:`a \leq \text{mean} \leq b`. NOTE: this impl is similar to the PyTorch `trunc_normal_`, the bounds [a, b] are applied while sampling the normal with mean/std applied, therefore a, b args should be adjusted to match the range of mean, std args. Args: tensor: an n-dimensional `torch.Tensor` mean: the mean of the normal distribution std: the standard deviation of the normal distribution a: the minimum cutoff value b: the maximum cutoff value Examples: >>> w = torch.empty(3, 5) >>> nn.init.trunc_normal_(w) )r&rrrrrs r trunc_normal_r*0s: "& ::rc[USSX45 [R"5 URU5R U5 SSS5 U$!,(df  U$=f)a}Fills the input Tensor with values drawn from a truncated normal distribution. The values are effectively drawn from the normal distribution :math:`\mathcal{N}(\text{mean}, \text{std}^2)` with values outside :math:`[a, b]` redrawn until they are within the bounds. The method used for generating the random values works best when :math:`a \leq \text{mean} \leq b`. NOTE: this 'tf' variant behaves closer to Tensorflow / JAX impl where the bounds [a, b] are applied when sampling the normal distribution with mean=0, std=1.0 and the result is subsquently scaled and shifted by the mean and std args. Args: tensor: an n-dimensional `torch.Tensor` mean: the mean of the normal distribution std: the standard deviation of the normal distribution a: the minimum cutoff value b: the maximum cutoff value Examples: >>> w = torch.empty(3, 5) >>> nn.init.trunc_normal_(w) rr N)r&rrr!r"r)s rtrunc_normal_tf_r,PsI861c10  Cd#  M  Ms !A Ascalemode distributionc[U5upEUS:XaUnOUS:XaUnO US:XaXE-S- nUW- nUS:Xa"[U[R"U5S- S9 gUS:Xa$UR [R"U5S9 gUS :Xa-[R"S U-5nUR U*U5 g[ S U35e) Nfan_infan_outfan_avgrtruncated_normalg۶%?)rnormaluniformzinvalid distribution )rr,rrnormal_r ValueError) rr-r.r/r1r2denomvariancebounds rvariance_scaling_r=rs 4F;OF x    !Q&u}H))TYYx%8;N%NO  !499X./  " !h,'&0?@@rc[USSS9 g)Nr1r4)r.r/)r=)rs r lecun_normal_r?sf8:LMr)gr gr )r r1r5) rrr torch.nn.initrTensorrr&r*r,strr=r?rrrDs:  7% LL% %%,1%6;%@E%T ; LL; ; ; ;  ;  \\ ;D  LL       \\ H A LLA A A A:N%,,Nr