o @TiH@s<ddlZddlmZGdddejZGdddejZdS)N)nncsDeZdZdZdededeffdd Zdejdejfd d Z Z S) ConvNeXtBlockamConvNeXt Block adapted from https://github.com/facebookresearch/ConvNeXt to 1D audio signal. Args: dim (int): Number of input channels. intermediate_dim (int): Dimensionality of the intermediate layer. layer_scale_init_value (float, optional): Initial value for the layer scale. None means no scaling. Defaults to None. dimintermediate_dimlayer_scale_init_valuecsttj||dd|d|_tj|dd|_t|||_t |_ t|||_ |dkr>tj |t |dd|_dSd|_dS) N) kernel_sizepaddinggroupsư>epsrT) requires_grad)super__init__rConv1ddwconv LayerNormnormLinearpwconv1GELUactpwconv2 Parametertorchonesgamma)selfrrr __class__M/home/ubuntu/.local/lib/python3.10/site-packages/linacodec/module/convnext.pyrs  zConvNeXtBlock.__init__xreturncCsn|}||}|dd}||}||}||}||}|jdur+|j|}|dd}||}|S)N)r transposerrrrr)rr$residualr"r"r#forward$s         zConvNeXtBlock.forward) __name__ __module__ __qualname____doc__intfloatrrTensorr* __classcell__r"r"r r#rs rcseZdZdZ   ddedededededBd edBd effd d Zed efddZ ed efddZ ddZ de j d e j fddZZS)ConvNextBackbonea Backbone module built with ConvNeXt blocks. Args: input_channels (int): Number of input features channels. dim (int): Hidden dimension of the model. intermediate_dim (int): Intermediate dimension used in ConvNeXtBlock. num_layers (int): Number of ConvNeXtBlock layers. layer_scale_init_value (float, optional): Initial value for layer scaling. Defaults to `1 / num_layers`. NFinput_channelsrr num_layersoutput_channelsr skip_embedcst||_||_|_|stj|dddnt|_tj dd|_ p+d|t fddt |D|_ |rEt|nt|_tj dd|_||jdS) Nrr)r r r r r&csg|] }tdqS)rrr)r).0_r8r"r# Rsz-ConvNextBackbone.__init__..)rrr4r6rrrIdentityembedrr ModuleListrangeconvnextrproj_outfinal_layer_normapply _init_weights)rr4rrr5r6rr7r r8r#r@s    zConvNextBackbone.__init__r%cCs|jSN)r4rr"r"r# input_dim_szConvNextBackbone.input_dimcCs|jr|jS|jSrE)r6rrFr"r"r# output_dimcszConvNextBackbone.output_dimcCs<t|tjtjfrtjj|jddtj|jddSdS)Ng{Gz?)stdr) isinstancerrrinit trunc_normal_weight constant_bias)rmr"r"r#rDgszConvNextBackbone._init_weightsr$cKsh|dd}||}||dd}|dd}|jD]}||}q||dd}||}|S)ah Args: x (Tensor): Input tensor of shape (B, L, C), where B is the batch size, C denotes output features, and L is the sequence length. Returns: Tensor: Output of shape (B, L, H), where B is the batch size, L is the sequence length, and H denotes the model dimension. r&r')r(r=rr@rBrA)rr$kwargs conv_blockr"r"r#r*ls     zConvNextBackbone.forward)NNF)r+r,r-r.r/r0boolrpropertyrGrHrDrr1r*r2r"r"r r#r34s4r3)rrModulerr3r"r"r"r#s ,