o Si@sddlmZddlZddlmZddlmZmZddlmZm Z ddl m Z Gdddej Z Gd d d e ZGd d d e ZGd dde ZdS))OptionalN)nn) _hz_to_mel _mel_to_hz)IMDCTISTFT)symexpc@s&eZdZdZdejdejfddZdS) FourierHeadz'Base class for inverse fourier modules.xreturncCstd)aJ Args: x (Tensor): Input tensor of shape (B, L, H), where B is the batch size, L is the sequence length, and H denotes the model dimension. Returns: Tensor: Reconstructed time-domain audio signal of shape (B, T), where T is the length of the output signal. z-Subclasses must implement the forward method.)NotImplementedError)selfr r?/home/ubuntu/.local/lib/python3.10/site-packages/vocos/heads.pyforwards zFourierHead.forwardN)__name__ __module__ __qualname____doc__torchTensorrrrrrr sr c sJeZdZdZd dedededeffdd Zd ejd ejfd d Z Z S) ISTFTHeada ISTFT Head module for predicting STFT complex coefficients. Args: dim (int): Hidden dimension of the model. n_fft (int): Size of Fourier transform. hop_length (int): The distance between neighboring sliding window frames, which should align with the resolution of the input features. padding (str, optional): Type of padding. Options are "center" or "same". Defaults to "same". samedimn_fft hop_lengthpaddingcs8t|d}tj|||_t||||d|_dS)N)rr win_lengthr)super__init__rrLinearoutristft)r rrrrout_dim __class__rrr &s zISTFTHead.__init__r r cCsn||dd}|jddd\}}t|}tj|dd}t|}t|}||d|}||}|S)ay Forward pass of the ISTFTHead module. Args: x (Tensor): Input tensor of shape (B, L, H), where B is the batch size, L is the sequence length, and H denotes the model dimension. Returns: Tensor: Reconstructed time-domain audio signal of shape (B, T), where T is the length of the output signal. rrY@maxy?) r" transposechunkrexpclipcossinr#)r r magpySaudiorrrr,s     zISTFTHead.forward)r) rrrrintstrr rrr __classcell__rrr%rrs rc sXeZdZdZ   ddedededeed ef fd d Zd e j d e j fddZ Z S)IMDCTSymExpHeada| IMDCT Head module for predicting MDCT coefficients with symmetric exponential function Args: dim (int): Hidden dimension of the model. mdct_frame_len (int): Length of the MDCT frame. padding (str, optional): Type of padding. Options are "center" or "same". Defaults to "same". sample_rate (int, optional): The sample rate of the audio. If provided, the last layer will be initialized based on perceptual scaling. Defaults to None. clip_audio (bool, optional): Whether to clip the audio output within the range of [-1.0, 1.0]. Defaults to False. rNFrmdct_frame_lenr sample_rate clip_audioc st|d}t|||_t||d|_||_|durYt|d}t d||}t |} d| | } t |jj| ddWddS1sRwYdSdS)Nr frame_lenrrr')rr rr!r"rimdctr=rrlinspacerr+no_gradweightmul_view) r rr;rr<r=r$m_maxm_ptsf_ptsscaler%rrr Us   "zIMDCTSymExpHead.__init__r r cCsF||}t|}tj|ddd}||}|jr!tj|ddd}|S)a Forward pass of the IMDCTSymExpHead module. Args: x (Tensor): Input tensor of shape (B, L, H), where B is the batch size, L is the sequence length, and H denotes the model dimension. Returns: Tensor: Reconstructed time-domain audio signal of shape (B, T), where T is the length of the output signal. gYr)minr+?)r"rrr/rAr=)r r r6rrrrms  zIMDCTSymExpHead.forward)rNF) rrrrr7r8rboolr rrrr9rrr%rr:Hs"r:c sJeZdZdZddedededeffdd Zd ej d ej fd d Z Z S) IMDCTCosHeadu IMDCT Head module for predicting MDCT coefficients with parametrizing MDCT = exp(m) · cos(p) Args: dim (int): Hidden dimension of the model. mdct_frame_len (int): Length of the MDCT frame. padding (str, optional): Type of padding. Options are "center" or "same". Defaults to "same". clip_audio (bool, optional): Whether to clip the audio output within the range of [-1.0, 1.0]. Defaults to False. rFrr;rr=cs0t||_t|||_t||d|_dS)Nr>)rr r=rr!r"rrA)r rr;rr=r%rrr s zIMDCTCosHead.__init__r r cCs\||}|jddd\}}t|jdd}||t|}|jr,tj|ddd}|S)a| Forward pass of the IMDCTCosHead module. Args: x (Tensor): Input tensor of shape (B, L, H), where B is the batch size, L is the sequence length, and H denotes the model dimension. Returns: Tensor: Reconstructed time-domain audio signal of shape (B, T), where T is the length of the output signal. rr(r)r*rMrNrK)r"r-rr.r/rAr0r=)r r mr3r6rrrrs zIMDCTCosHead.forward)rF) rrrrr7r8rOr rrrr9rrr%rrPs rP)typingrrr torchaudio.functional.functionalrrvocos.spectral_opsrr vocos.modulesrModuler rr:rPrrrrs   .: