`idZddlZddlmZmZddlZddlmZm Z m Z m Z ddl m Z ddlmZejeZGdd e Zd gZdS) zXcodec model configurationN)OptionalUnion) AutoConfig DacConfig HubertConfig WavLMConfig)PretrainedConfig)loggingceZdZdZdZeedZdddddgddgddgdddd ddf d ee e d e d e d e e de e de e de de dee de de e efde e efffd Zede fdZede fdZede fdZede fdZede fdZede fdZxZS) XcodecConfiga This is the configuration class to store the configuration of an [`XcodecModel`]. It is used to instantiate a Xcodec model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the [Manel/X-Codec](https://huggingface.co/Manel/X-Codec) architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: target_bandwidths (`List[float]`, *optional*, defaults to `[0.5, 1, 1.5, 2, 4]`): The range of different bandwidths (in kbps) the model can encode audio with. sample_rate (`int`, *optional*, defaults to 16000): The sampling rate at which the audio waveform should be digitalized, in hertz (Hz). kernel_size (`int`, *optional*, defaults to 3): Kernel size for the initial semantic convolution. channel_ratios (`List[float]`, *optional*, defaults to `[1, 1]`): Expansion factors for the number of output channels in each semantic block. strides (`List[int]`, *optional*, defaults to `[1, 1]`): Strides for each semantic encoder block. block_dilations (`List[int]`, *optional*, defaults to `[1, 1]`): Dilation factors for the residual units in semantic blocks. unit_kernel_size (`int`, *optional*, defaults to 3): Kernel size inside each ResidualUnit in semantic blocks. codebook_size (`int`, *optional*, defaults to 1024): Number of entries in each residual quantizer's codebook. codebook_dim (`int`, *optional*): Dimensionality of each codebook vector. Defaults to sum of hidden size of acoustic and semantic models. initializer_range (`float`, *optional*, defaults to 0.02): Standard deviation of the truncated normal initializer for all weight matrices. acoustic_model_config (`Union[Dict, DacConfig]`, *optional*): An instance of the configuration for the acoustic (DAC) model. semantic_model_config (`Union[Dict, HubertConfig, WavLMConfig]`, *optional*): An instance of the configuration object for the semantic (HuBERT) model. Example: ```python >>> from transformers import XcodecModel, XcodecConfig >>> # Initializing configuration >>> configuration = XcodecConfig() >>> # Initializing a model (with random weights) from the configuration >>> model = XcodecModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```xcodec)acoustic_model_configsemantic_model_configNi>r g{Gz?target_bandwidths sample_rate kernel_sizechannel_ratiosstridesblock_dilationsunit_kernel_size codebook_size codebook_diminitializer_rangerrc  tjd i| | tdgddgdd|_nct | t rtd i| |_nD & &tW[\qWrWrtt  ! ()5D & & -t 4 4 "777-7-GH]^mHn-o-o**u.:-R-RD & &FhlnCiDiDFF   $ 3 3 3 !2&&, . 0*!2  5ADD^DjjL(returncDtj|j|jz SN)mathceilr hop_lengthr4s r7 frame_ratezXcodecConfig.frame_ratesy)DO;<<zXcodecConfig.hop_lengths 2745IJJKKKr8cXtjtj|jSr;)r<r=log2rr?s r7codebook_nbitszXcodecConfig.codebook_nbitssy4#566777r8c4|jj|jjzSr;)rr(rr?s r7r(zXcodecConfig.hidden_sizes)58R8^^^r8c\td|jdz|j|jzzS)Ni)rDrr@rIr?s r7num_quantizerszXcodecConfig.num_quantizerss,4$0444K^9^_```r8)__name__ __module__ __qualname____doc__ model_typerr sub_configsrlistfloatrDrr.rr,propertyr@rBr>rIr(rM __classcell__)r6s@r7r r sM0 0 dJ"+!+K48 '(!fV&'V !!&*#'8<;?F)F)#DK0F)F) F) U F) c F)cF)F)F)smF)!F) %T9_5F) %T<%78F)F)F)F)F)F)P=C===X=6c666X6LCLLLXL8888X8_S___X_aaaaXaaaaar8r )rQr<typingrrnumpyrE transformersrrrrconfiguration_utilsr utilsr get_loggerrNr2r __all__r*r8r7r_s! """"""""IIIIIIIIIIII333333  H % %XaXaXaXaXa#XaXaXav  r8