o ߥi@sddlZddlZddlZddlmZddlZddlZddlmZddl m Z ddl m Z ddl mZeddd Zd d ZGd d d ejZGdddejZ   dddZ  dddZddZdS)N)repeat)nn) functional)init) _BatchNormcKst|ts|g}|D]k}|D]d}t|tjr7tj|jfi||jj|9_|j dur6|j j |qt|tj r\tj|jfi||jj|9_|j dur[|j j |qt|t rtt |jd|j durt|j j |qq dS)aeInitialize network weights. Args: module_list (list[nn.Module] | nn.Module): Modules to be initialized. scale (float): Scale initialized weights, especially for residual blocks. Default: 1. bias_fill (float): The value to fill bias. Default: 0 kwargs (dict): Other arguments for initialization function. Nr) isinstancelistmodulesrConv2drkaiming_normal_weightdatabiasfill_Linearr constant_) module_listscale bias_fillkwargsmodulemrc/home/ubuntu/.local/lib/python3.10/site-packages/modelscope/models/cv/super_resolution/arch_util.pydefault_init_weightss.        rcKs0g}t|D] }||di|qtj|S)zMake layers by stacking the same blocks. Args: basic_block (nn.module): nn.module class for basic block. num_basic_block (int): number of blocks. Returns: nn.Sequential: Stacked blocks in nn.Sequential. Nr)rangeappendr Sequential) basic_blocknum_basic_blockkwarglayers_rrr make_layer/s  r$cs*eZdZdZd fdd ZddZZS) ResidualBlockNoBNaResidual block without BN. It has a style of: ---Conv-ReLU-Conv-+- |________________| Args: num_feat (int): Channel number of intermediate features. Default: 64. res_scale (float): Residual scale. Default: 1. pytorch_init (bool): If set to True, use pytorch default init, otherwise, use default_init_weights. Default: False. @rFcsptt|||_tj||ddddd|_tj||ddddd|_tjdd|_ |s6t |j|jgddSdS)NrT)r)inplaceg?) superr%__init__ res_scalerr conv1conv2ReLUrelur)selfnum_featr+ pytorch_init __class__rrr*NszResidualBlockNoBN.__init__cCs(|}||||}|||jS)N)r-r/r,r+)r0xidentityoutrrrforwardXszResidualBlockNoBN.forward)r&rF)__name__ __module__ __qualname____doc__r*r8 __classcell__rrr3rr%?s r%cs eZdZdZfddZZS)UpsamplezUpsample module. Args: scale (int): Scale factor. Supported scales: 2^n and 3. num_feat (int): Channel number of intermediate features. c sg}||d@dkr.ttt|dD]}|t|d|ddd|tdqn#|dkrI|t|d|ddd|tdntd|dt t |j |dS) Nrrr' zscale z/ is not supported. Supported scales: 2^n and 3.) rintmathlogrrr PixelShuffle ValueErrorr)r>r*)r0rr1rr#r3rrr*fs zUpsample.__init__)r9r:r;r<r*r=rrr3rr>^sr>bilinearzerosTcCs|dd|ddksJ|\}}}}ttd||td||\}} t| |fd} d| _| |} d| dddddddft|ddd } d| dddddddft|ddd } tj| | fdd }t j |||||d }|S) akWarp an image or feature map with optical flow. Args: x (Tensor): Tensor with size (n, c, h, w). flow (Tensor): Tensor with size (n, h, w, 2), normal value. interp_mode (str): 'nearest' or 'bilinear'. Default: 'bilinear'. padding_mode (str): 'zeros' or 'border' or 'reflection'. Default: 'zeros'. align_corners (bool): Before pytorch 1.3, the default value is align_corners=True. After pytorch 1.3, the default value is align_corners=False. Here, we use the True as default. Returns: Tensor: Warped image or feature map. Nrr'rr?Fg@g?)dim)mode padding_mode align_corners) sizetorchmeshgridarangetype_asstackfloat requires_gradmaxF grid_sample)r5flow interp_moderLrMr#hwgrid_ygrid_xgridvgridvgrid_xvgrid_y vgrid_scaledoutputrrr flow_warpvs($22reFcCs|\}}}}|dkrt||dt||d}} n|dkr,|d|d}} ntd|d|} ||} | |} | dddddddf| 9<| dddddddf| 9<tj| || f||d} | S) aResize a flow according to ratio or shape. Args: flow (Tensor): Precomputed flow. shape [N, 2, H, W]. size_type (str): 'ratio' or 'shape'. sizes (list[int | float]): the ratio for resizing or the final output shape. 1) The order of ratio should be [ratio_h, ratio_w]. For downsampling, the ratio should be smaller than 1.0 (i.e., ratio < 1.0). For upsampling, the ratio should be larger than 1.0 (i.e., ratio > 1.0). 2) The order of output_size should be [out_h, out_w]. interp_mode (str): The mode of interpolation for resizing. Default: 'bilinear'. align_corners (bool): Whether align corners. Default: False. Returns: Tensor: Resized flow. ratiorrshapez1Size type should be ratio or shape, but got type .N)inputrNrKrM)rNrBrFclonerW interpolate)rY size_typesizesrZrMr#flow_hflow_woutput_houtput_w input_flowratio_hratio_w resized_flowrrr resize_flows($ $$rvc Cs||\}}}}||d}||dkr||dksJ||}||}|||||||} | dddddd||||S)z Pixel unshuffle. Args: x (Tensor): Input feature with shape (b, c, hh, hw). scale (int): Downsample ratio. Returns: Tensor: the pixel unshuffled feature. r?rrr'r@)rNviewpermutereshape) r5rbchhhw out_channelr[r\x_viewrrrpixel_unshuffles  r)rr)rGrHT)rGF)collections.abc collectionsrCwarnings itertoolsrrO torchvisionrtorch.nnrrWrtorch.nn.modules.batchnormrno_gradrr$Moduler%rr>rervrrrrrs.       0 /