o *j'>@s|ddlZddlmZddlmZddlmZmZmZm Z m Z m Z m Z m Z mZmZdZGdddejZGd d d ejZdS) N)nn) functional) MemoryEfficientSwishSwishcalculate_output_image_size drop_connectefficientnet_paramsget_model_paramsget_same_padding_conv2dload_pretrained_weights round_filters round_repeats) zefficientnet-b0zefficientnet-b1zefficientnet-b2zefficientnet-b3zefficientnet-b4zefficientnet-b5zefficientnet-b6zefficientnet-b7zefficientnet-b8zefficientnet-l2cs2eZdZd fdd Zd ddZd ddZZS) MBConvBlockNc st||_d|j|_|j|_|jjduo%d|jjko#dkn|_|j |_ |jj }|jj |jj }|jj dkrVt |d}|||ddd|_ tj||j|jd|_|jj}|jj}t |d}||||||dd|_tj||j|jd|_t||}|jrt dd}tdt|jj |jj} ||| dd |_|| |dd |_|jj} t |d}||| ddd|_tj| |j|jd|_t|_dS) Nrr image_sizeF) in_channels out_channels kernel_sizebiasZ num_featuresZmomentumeps)rrgroupsrstrider)rr)rrr)super__init__ _block_argsbatch_norm_momentumZ_bn_mombatch_norm_epsilonZ_bn_epsZse_ratiohas_seid_skip input_filters expand_ratior _expand_convr BatchNorm2d_bn0rr_depthwise_conv_bn1rmaxint _se_reduce _se_expandoutput_filters _project_conv_bn2r_swish) self block_args global_paramsrinpZoupConv2dksZnum_squeezed_channelsZ final_oup __class__r/var/www/html/Deteccion_Ine/venv/lib/python3.10/site-packages/modelscope/models/cv/face_emotion/efficient/model.pyrs|            zMBConvBlock.__init__cCs|}|jjdkr||}||}||}||}||}||}|jrEt |d}| |}||}| |}t ||}||}||}|jj|jj}}|jrs|jjdkrs||krs|rot|||jd}||}|S)zMBConvBlock's forward function. Args: inputs (tensor): Input tensor. drop_connect_rate (bool): Drop connect rate (float, between 0 and 1). Returns: Output of this block after processing. r)ptraining)rr"r#r%r/r&r'rFZadaptive_avg_pool2dr*r+torchZsigmoidr-r.r!r,r rrr<)r0inputsdrop_connect_ratexZ x_squeezedr!r,r9r9r:forwardLs0             zMBConvBlock.forwardTcCs|rt|_dSt|_dSzSets swish function as memory efficient (for training) or standard (for export). Args: memory_efficient (bool): Whether to use memory-efficient version of swish. N)rrr/)r0memory_efficientr9r9r: set_swishqszMBConvBlock.set_swish)NT)__name__ __module__ __qualname__rrBrE __classcell__r9r9r7r:rs 7%rcseZdZdZdfdd ZdddZdd Zd d Zd d Ze dddZ e    dddZ e ddZ e ddZ ddZZS) EfficientNetajEfficientNet model. Most easily loaded with the .from_name or .from_pretrained methods. Args: blocks_args (list[namedtuple]): A list of BlockArgs to construct blocks. global_params (namedtuple): A set of GlobalParams shared between blocks. References: [1] https://arxiv.org/abs/1905.11946 (EfficientNet) Example: >>> import torch >>> from efficientnet.model import EfficientNet >>> inputs = torch.rand(1, 3, 224, 224) >>> model = EfficientNet.from_pretrained('efficientnet-b0') >>> model.eval() >>> outputs = model(inputs) Nc stt|tsJdt|dksJd||_||_d|jj}|jj}|j }t |d}d}t d|j}|||ddd d |_ t j|||d |_t|d}t g|_|jD]N} | jt | j|jt | j|jt| j|jd } |jt| |j|dt|| j}| jdkr| j| jdd } t| jdD]} |jt| |j|dqqY| j}t d|j}t |d}|||dd d|_t j|||d |_t d|_|jj rt !|jj"|_#t $||jj%|_&t'|_(dS)Nzblocks_args should be a listrz!block args must be greater than 0rr Frrrr)r!r, num_repeat)r!ri)rr))rr isinstancelistlen_global_paramsZ _blocks_argsrrrr r _conv_stemrr$r%rZ ModuleList_blocks_replacer!r,rrPappendrrrange _conv_headr'ZAdaptiveAvgPool2d _avg_pooling include_topZDropoutZ dropout_rate_dropoutZLinear num_classes_fcrr/) r0 blocks_argsr2Zbn_momZbn_epsrr4rrr1_r7r9r:rs             zEfficientNet.__init__TcCs,|rtnt|_|jD]}||q dSrC)rrr/rVrE)r0rDblockr9r9r:rEs  zEfficientNet.set_swishcCst}||||}|}t|jD]G\}}|jj}|r*|t|t |j9}|||d}| d| dkrF||d t |d<n|t |jdkrZ||d t |d<|}q|| | |}||d t |d<|S)a?Use convolution layer to extract features from reduction levels i in [1, 2, 3, 4, 5]. Args: inputs (tensor): Input tensor. Returns: Dictionary of last intermediate features with reduction levels i in [1, 2, 3, 4, 5]. Example: >>> import torch >>> from efficientnet.model import EfficientNet >>> inputs = torch.rand(1, 3, 224, 224) >>> model = EfficientNet.from_pretrained('efficientnet-b0') >>> endpoints = model.extract_endpoints(inputs) >>> print(endpoints['reduction_1'].shape) # torch.Size([1, 16, 112, 112]) >>> print(endpoints['reduction_2'].shape) # torch.Size([1, 24, 56, 56]) >>> print(endpoints['reduction_3'].shape) # torch.Size([1, 40, 28, 28]) >>> print(endpoints['reduction_4'].shape) # torch.Size([1, 112, 14, 14]) >>> print(endpoints['reduction_5'].shape) # torch.Size([1, 320, 7, 7]) >>> print(endpoints['reduction_6'].shape) # torch.Size([1, 1280, 7, 7]) r@rNz reduction_{}r)dictr/r%rU enumeraterVrTr@floatrSsizeformatr'rZ)r0r?Z endpointsrAZprev_xidxrbr@r9r9r:extract_endpointss$  zEfficientNet.extract_endpointscCsr||||}t|jD]\}}|jj}|r%|t|t|j9}|||d}q|| | |}|S)zuse convolution layer to extract feature . Args: inputs (tensor): Input tensor. Returns: Output of the final convolution layer in the efficientnet model. rc) r/r%rUrerVrTr@rfrSr'rZ)r0r?rArirbr@r9r9r:extract_featuresszEfficientNet.extract_featurescCs@||}||}|jjr|jdd}||}||}|S)a EfficientNet's forward function. Calls extract_features to extract features, applies final linear layer, and returns logits. Args: inputs (tensor): Input tensor. Returns: Output of this model after processing. r)Z start_dim)rkr[rTr\flattenr]r_)r0r?rAr9r9r:rB s     zEfficientNet.forwardrLcKs0||t||\}}|||}|||S)aCreate an efficientnet model according to name. Args: model_name (str): Name for efficientnet. in_channels (int): Input data's channel number. override_params (other key word params): Params to override model's global_params. Optional key: 'width_coefficient', 'depth_coefficient', 'image_size', 'dropout_rate', 'num_classes', 'batch_norm_momentum', 'batch_norm_epsilon', 'drop_connect_rate', 'depth_divisor', 'min_depth' Returns: An efficientnet model. )_check_model_name_is_validr _change_in_channels)cls model_nameroverride_paramsr`r2modelr9r9r: from_names   zEfficientNet.from_nameFcKs$|j|fd|i|}|||S)a{Create an efficientnet model according to name. Args: model_name (str): Name for efficientnet. weights_path (None or str): str: path to pretrained weights file on the local disk. None: use pretrained weights downloaded from the Internet. advprop (bool): Whether to load pretrained weights trained with advprop (valid when weights_path is None). in_channels (int): Input data's channel number. num_classes (int): Number of categories for classification. It controls the output size for final linear layer. override_params (other key word params): Params to override model's global_params. Optional key: 'width_coefficient', 'depth_coefficient', 'image_size', 'dropout_rate', 'batch_norm_momentum', 'batch_norm_epsilon', 'drop_connect_rate', 'depth_divisor', 'min_depth' Returns: A pretrained efficientnet model. r^)rsrn)rorpZ weights_pathZadvproprr^rqrrr9r9r:from_pretrained5s  zEfficientNet.from_pretrainedcCs||t|\}}}}|S)zGet the input image size for a given efficientnet model. Args: model_name (str): Name for efficientnet. Returns: Input image size (resolution). )rmr )rorpraresr9r9r:get_image_sizeZs zEfficientNet.get_image_sizecCs|tvr tddtdS)zValidates model name. Args: model_name (str): Name for efficientnet. Returns: bool: Is a valid name or not. zmodel_name should be one of: z, N) VALID_MODELS ValueErrorjoin)rorpr9r9r:rmfs z'EfficientNet._check_model_name_is_validcCs>|dkrt|jjd}td|j}|||dddd|_dSdS)zAdjust model's first convolution layer to in_channels, if in_channels not equals 3. Args: in_channels (int): Input data's channel number. rLrrMrNFrON)r rTrr rU)r0rr4rr9r9r:rnrs   z EfficientNet._change_in_channels)NNrF)rL)NFrLrt)rGrHrI__doc__rrErjrkrB classmethodrsrurwrmrnrJr9r9r7r:rKys( < +  $   rK)r>rZtorch.nnrr=utilsrrrrr r r r r rrxModulerrKr9r9r9r:s  0f