9i3SSKrSSKJr SS\S\4Sjjr"SS\R 5rSSjrSSjr "S S \R 5r "S S \R 5r "S S\R 5r g)N drop_probtrainingcUS:XdU(dU$SU- nURS4SURS- --nU[R"X@RUR S9-nUR 5 URU5U-nU$)aDrop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). This is the same as the DropConnect impl I created for EfficientNet, etc networks, however, the original name is misleading as 'Drop Connect' is a.sh different form of dropout in a.sh separate paper... See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... I've opted for changing the layer and argument names to 'drop path' rather than mix DropConnect as a.sh layer name and use 'survival rate' as the argument. rr)dtypedevice)shapendimtorchrandr r floor_div)xrr keep_probr random_tensoroutputs h/var/www/html/land-doc-ocr/venv/lib/python3.13/site-packages/modelscope/models/multi_modal/ofa/resnet.py drop_pathrsBhI I WWQZNU   E  WWQXX!//M UU9  -F Mc6^\rSrSrSrSU4SjjrSrSrU=r$)DropPath'z^Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). c6>[[U] 5 XlgN)superr__init__r)selfr __class__s rrDropPath.__init__+s h&("rcB[XRUR5$r)rrrrrs rforwardDropPath.forward/sNNDMM::r)rr) __name__ __module__ __qualname____firstlineno____doc__rr$__static_attributes__ __classcell__r s@rrr's#;;rrc 8[R"UUSUUUSUS9$)z3x3 convolution with paddingF) kernel_sizestridepaddinggroupsbiasdilationnnConv2d) in_planes out_planesr1r3r5s rconv3x3r;3s+ 99  rc0[R"XSUSS9$)z1x1 convolutionrF)r0r1r4r6)r9r:r1s rconv1x1r=@s 991V% IIrcB^\rSrSrSrSU4SjjrSrSrU=r$) BasicBlockFrc `>[[U] 5 Uc[RnUS:wdUS:wa [ S5eUS:a [ S5e[XU5UlU"U5Ul [R"SS9Ul [X"5Ul U"U5Ul X@lX0lg)Nr@z3BasicBlock only supports groups=1 and base_width=64z(Dilation > 1 not supported in BasicBlockTinplace)rr?rr7 BatchNorm2d ValueErrorNotImplementedErrorr;conv1bn1ReLUreluconv2bn2 downsampler1) rinplanesplanesr1rNr3 base_widthr5 norm_layerr s rrBasicBlock.__init__Is j$(*  J Q;**EG G a<%:< <Xv6 f%GGD) V, f%$ rcer)rHrIrKrLrMrNrridentityouts rr$BasicBlock.forwarddsur)rIrMrHrLrNrKr1)rNrrBrN r&r'r(r) expansionrr$r+r,r-s@rr?r?Fs+I   6rr?cD^\rSrSrSrSU4SjjrSrSrU=r$) Bottleneckxc >[[U] 5 Uc[Rn[ X&S- -5U-n [ X5UlU"U 5Ul[XX5U5Ul U"U 5Ul [ XUR-5Ul U"X R-5Ul[R"SS9UlX@lX0lU S:a['U 5Ulg[R("5Ulg)NgP@TrCr)rr\rr7rEintr=rHrIr;rLrMrZconv3bn3rJrKrNr1rIdentityr) rrOrPr1rNr3rQr5rRdrop_path_ratewidthr s rrBottleneck.__init__s j$(*  JF3./069X- e$U68D e$UT^^$;< f~~56GGD) $ -3" 9; rcUnURU5nURU5nURU5nURU5nUR U5nURU5nUR U5nUR U5nURbURU5nX RU5-nURU5nU$r) rHrIrKrLrMrarbrNrrUs rr$Bottleneck.forwardsjjmhhsmiinjjohhsmiinjjohhsm ?? &q)H,,iin r) rIrMrbrHrLrarNrrKr1)rNrrBrNrrYr-s@rr\r\xs1I   #G6rr\cX^\rSrSrSrSU4SjjrS SjrSrSrSr U=r $) ResNetak Deep residual network, copy from https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py. You can see more details from https://arxiv.org/abs/1512.03385 step 1. Get image embedding with `7` as the patch image size, `2` as stride. step 2. Do layer normalization, relu activation and max pooling. step 3. Go through three times residual branch. c >[[U] 5 Uc[RnX`lSUlSUlUc/SQn[U5S:wa[SRU55eX0l X@l [R"SUR SSSS S 9UlU"UR 5Ul[R "S S 9Ul[R$"SSSS 9UlUR)[*SUSUS9UlUR)[*SUSSUSUS9UlUR)[*SUSSUSUS9UlUR35Hn[5U[R5(a+[R6R9UR:SSS9 MM[5U[R<[R[R>45(dM[R6RAUR:S5 [R6RAURBS5 M U(aUR35Hn[5U[*5(a6[R6RAURDR:S5 MN[5U[F5(dMe[R6RAURHR:S5 M gg)aX Args: layers (`Tuple[int]`): There are three layers in resnet, so the length of layers should greater then three. And each element in `layers` is the number of `Bottleneck` in relative residual branch. zero_init_residual (`bool`, **optional**, default to `False`): Whether or not to zero-initialize the last BN in each residual branch. groups (`int`, **optional**, default to `1`): The number of groups. So far, only the value of `1` is supported. width_per_group (`int`, **optional**, default to `64`): The width in each group. So far, only the value of `64` is supported. replace_stride_with_dilation (`Tuple[bool]`, **optional**, default to `None`): Whether or not to replace stride with dilation in each residual branch. norm_layer (`torch.nn.Module`, **optional**, default to `None`): The normalization module. If `None`, will use `torch.nn.BatchNorm2d`. drop_path_rate (`float`, **optional**, default to 0.0): Drop path rate. See more details about drop path from https://arxiv.org/pdf/1605.07648v4.pdf. NrBr)FFFr/zHreplace_stride_with_dilation should be None or a 3-element tuple, got {}F)r0r1r2r4TrC)r0r1r2r)rd)r1dilaterdfan_outrK)mode nonlinearity)%rrjrr7rE _norm_layerrOr5lenrFformatr3rQr8rHrIrJrK MaxPool2dmaxpool _make_layerr\layer1layer2layer3modules isinstanceinitkaiming_normal_weight SyncBatchNorm GroupNorm constant_r4rbr?rM) rlayerszero_init_residualr3width_per_groupreplace_stride_with_dilationrRrdmr s rrResNet.__init__sk6 fd$&  J%  ' /,A ( + , 1<!JK!JAvvx!JKq&!A MMMM;;#!]])#&6+ ,"}}f%%Ls FcURU5nURU5nURU5nURU5nUR U5nUR U5nUR U5nU$r)rHrIrKryr{r|r}r#s r _forward_implResNet._forward_implQsb JJqM HHQK IIaL LLO KKN KKN KKNrc$URU5$r)rr#s rr$ResNet.forward[s!!!$$r) rurQrIrHr5r3rOr{r|r}ryrK)FrrBNNr)rFr) r&r'r(r)r*rrzrr$r+r,r-s@rrjrjsC%*!#.2 #S7r #& <&|%%rrj)rF)rrrr) r torch.nnr7floatboolrModulerr;r=r?r\rjrrrso E$( ;ryy ; I //d88vi%RYYi%r