o Siw@sldZddlmZmZddlZddlZddlZddlmZddl m Z eGdddZ Gd d d ej Z dS) z)Residual vector quantizer implementation.) dataclassfieldN)nn)ResidualVectorQuantizationc@sReZdZUejed<ejed<ejed<dZejejed<e e dZ e ed<dS)QuantizedResult quantizedcodes bandwidthNpenalty)default_factorymetrics) __name__ __module__ __qualname__torchTensor__annotations__r tpOptionalrdictr rrK/home/ubuntu/.local/lib/python3.10/site-packages/encodec/quantization/vq.pyrs   rcseZdZdZ       d"d ed ed ed ed ededeffdd Zd#dej dede j ede fddZ d#dede j edefddZdefddZd#dej dede j edej fddZdej dej fd d!ZZS)$ResidualVectorQuantizeraResidual Vector Quantizer. Args: dimension (int): Dimension of the codebooks. n_q (int): Number of residual vector quantizers used. bins (int): Codebook size. decay (float): Decay for exponential moving average over the codebooks. kmeans_init (bool): Whether to use kmeans to initialize the codebooks. kmeans_iters (int): Number of iterations used for kmeans initialization. threshold_ema_dead_code (int): Threshold for dead code expiration. Replace any codes that have an exponential moving average cluster size less than the specified threshold with randomly selected vector from the current batch. Gz?T2 dimensionn_qbinsdecay kmeans_init kmeans_itersthreshold_ema_dead_codec s^t||_||_||_||_||_||_||_t |j|j|j|j|j|j|jd|_ dS)N)dim codebook_sizenum_quantizersr#r$r%r&) super__init__r!r r"r#r$r%r&rvq)selfr r!r"r#r$r%r& __class__rrr+)s"  z ResidualVectorQuantizer.__init__Nx sample_rater returnc CsT||}|||}|j||d\}}}t|||} t||| t|dS)aResidual vector quantization on the given input tensor. Args: x (torch.Tensor): Input tensor. sample_rate (int): Sample rate of the input tensor. bandwidth (float): Target bandwidth. Returns: QuantizedResult: The quantized (or approximately quantized) representation with the associated bandwidth and any penalty term for the loss. r!)r )get_bandwidth_per_quantizer get_num_quantizers_for_bandwidthr,rtensortormean) r-r0r1r bw_per_qr!rr commit_lossbwrrrforwardEs zResidualVectorQuantizer.forwardcCs8||}|j}|r|dkrttdt||}|S)z8Return n_q based on specified target bandwidth. gr)r4r!intmaxmathfloor)r-r1r r9r!rrrr5Vs  z8ResidualVectorQuantizer.get_num_quantizers_for_bandwidthcCst|j|dS)zFReturn bandwidth per quantizer for a given input sample rate. i)r?log2r")r-r1rrrr4_sz3ResidualVectorQuantizer.get_bandwidth_per_quantizercCs |||}|jj||d}|S)zEncode a given input tensor with the specified sample rate at the given bandwidth. The RVQ encode method sets the appropriate number of quantizer to use and returns indices for each quantizer. r3)r5r,encode)r-r0r1r r!r rrrrBds zResidualVectorQuantizer.encoder cCs|j|}|S)z@Decode the given codes to the quantized representation. )r,decode)r-r rrrrrCms zResidualVectorQuantizer.decode)rrrrTrr)N)rrr__doc__r=floatboolr+rrrrrr<r5r4rBrC __classcell__rrr.rrs:$ & r)rD dataclassesrrr?typingrrrcore_vqrrModulerrrrrs