o
    fôjC  ã                   @   sz   d dl Z d dl mZ d dlZd dlmZ ddlmZ ddlmZ G dd„ dej	ƒZ
G d	d
„ d
ej	ƒZG dd„ dej	ƒZdS )é    N)ÚList)Únné   )ÚAudioSignal©Ú
STFTParamsc                       sz   e Zd ZdZddge ¡ dddddddf	d	ee d
ej	de
de
de
de
de
dedef‡ fdd„Zdedefdd„Z‡  ZS )ÚMultiScaleSTFTLossa%  Computes the multi-scale STFT loss from [1].

    Parameters
    ----------
    window_lengths : List[int], optional
        Length of each window of each STFT, by default [2048, 512]
    loss_fn : typing.Callable, optional
        How to compare each loss, by default nn.L1Loss()
    clamp_eps : float, optional
        Clamp on the log magnitude, below, by default 1e-5
    mag_weight : float, optional
        Weight of raw magnitude portion of loss, by default 1.0
    log_weight : float, optional
        Weight of log magnitude portion of loss, by default 1.0
    pow : float, optional
        Power to raise magnitude to before taking log, by default 2.0
    weight : float, optional
        Weight of this loss, by default 1.0
    match_stride : bool, optional
        Whether to match the stride of convolutional layers, by default False

    References
    ----------

    1.  Engel, Jesse, Chenjie Gu, and Adam Roberts.
        "DDSP: Differentiable Digital Signal Processing."
        International Conference on Learning Representations. 2019.
    é   é   çñhãˆµøä>ç      ð?ç       @FNÚwindow_lengthsÚloss_fnÚ	clamp_epsÚ
mag_weightÚ
log_weightÚpowÚweightÚmatch_strideÚwindow_typec
           
         sH   t ƒ  ¡  ‡ ‡fdd„|D ƒ| _|| _|| _|| _|| _|| _|| _d S )Nc                    ó    g | ]}t ||d  ˆ ˆd�‘qS ©é   )Úwindow_lengthÚ
hop_lengthr   r   r   ©Ú.0Úw©r   r   © úl/var/www/html/academy/tools/voices/clone/sv-venv/lib/python3.10/site-packages/audiotools/metrics/spectral.pyÚ
<listcomp>6   ó    úüÿz/MultiScaleSTFTLoss.__init__.<locals>.<listcomp>)	ÚsuperÚ__init__Ústft_paramsr   r   r   r   r   r   )
Úselfr   r   r   r   r   r   r   r   r   ©Ú	__class__r   r!   r%   )   s   
ù	
zMultiScaleSTFTLoss.__init__ÚxÚyc              	   C   s˜   d}| j D ]D}| |j|j|j¡ | |j|j|j¡ || j|  |j | j	¡ 
| j
¡ ¡ |j | j	¡ 
| j
¡ ¡ ¡ 7 }|| j|  |j|j¡ 7 }q|S )a?  Computes multi-scale STFT between an estimate and a reference
        signal.

        Parameters
        ----------
        x : AudioSignal
            Estimate signal
        y : AudioSignal
            Reference signal

        Returns
        -------
        torch.Tensor
            Multi-scale STFT loss.
        ç        )r&   Ústftr   r   r   r   r   Ú	magnitudeÚclampr   r   Úlog10r   )r'   r*   r+   ÚlossÚsr    r    r!   ÚforwardF   s   

þzMultiScaleSTFTLoss.forward©Ú__name__Ú
__module__Ú__qualname__Ú__doc__r   ÚL1Lossr   ÚintÚtypingÚCallableÚfloatÚboolÚstrr%   r   r3   Ú__classcell__r    r    r(   r!   r      s>    öþýüûúùø	÷
ör   c                       s¤   e Zd ZdZddgddge ¡ dddddd	d
d
gddgdfdee dee dej	de
de
de
de
de
dedee
 dee
 def‡ fdd„Zdedefdd„Z‡  ZS )ÚMelSpectrogramLossaÅ  Compute distance between mel spectrograms. Can be used
    in a multi-scale way.

    Parameters
    ----------
    n_mels : List[int]
        Number of mels per STFT, by default [150, 80],
    window_lengths : List[int], optional
        Length of each window of each STFT, by default [2048, 512]
    loss_fn : typing.Callable, optional
        How to compare each loss, by default nn.L1Loss()
    clamp_eps : float, optional
        Clamp on the log magnitude, below, by default 1e-5
    mag_weight : float, optional
        Weight of raw magnitude portion of loss, by default 1.0
    log_weight : float, optional
        Weight of log magnitude portion of loss, by default 1.0
    pow : float, optional
        Power to raise magnitude to before taking log, by default 2.0
    weight : float, optional
        Weight of this loss, by default 1.0
    match_stride : bool, optional
        Whether to match the stride of convolutional layers, by default False
    é–   éP   r	   r
   r   r   r   Fr,   NÚn_melsr   r   r   r   r   r   r   r   Úmel_fminÚmel_fmaxr   c                    sZ   t ƒ  ¡  ‡ ‡fdd„|D ƒ| _|| _|| _|| _|| _|| _|| _|
| _	|| _
|| _d S )Nc                    r   r   r   r   r   r    r!   r"   Œ   r#   z/MelSpectrogramLoss.__init__.<locals>.<listcomp>)r$   r%   r&   rD   r   r   r   r   r   rE   rF   r   )r'   rD   r   r   r   r   r   r   r   r   rE   rF   r   r(   r   r!   r%   |   s   
ù	
zMelSpectrogramLoss.__init__r*   r+   c              	   C   sÂ   d}t | j| j| j| jƒD ]Q\}}}}|j|j|jdœ}|j|f||dœ|¤Ž}	|j|f||dœ|¤Ž}
|| j	|  
|	 | j¡ | j¡ ¡ |
 | j¡ | j¡ ¡ ¡ 7 }|| j|  
|	|
¡ 7 }q|S )a*  Computes mel loss between an estimate and a reference
        signal.

        Parameters
        ----------
        x : AudioSignal
            Estimate signal
        y : AudioSignal
            Reference signal

        Returns
        -------
        torch.Tensor
            Mel loss.
        r,   )r   r   r   )rE   rF   )ÚziprD   rE   rF   r&   r   r   r   Úmel_spectrogramr   r   r/   r   r   r0   r   )r'   r*   r+   r1   rD   ÚfminÚfmaxr2   ÚkwargsZx_melsZy_melsr    r    r!   r3   Ÿ   s    ÿý
þzMelSpectrogramLoss.forwardr4   r    r    r(   r!   rA   b   sP    óþýüûúùø	÷
öõôó#rA   c                       sD   e Zd ZdZ	ddededef‡ fdd	„Zd
edefdd„Z‡  Z	S )Ú	PhaseLossa0  Difference between phase spectrograms.

    Parameters
    ----------
    window_length : int, optional
        Length of STFT window, by default 2048
    hop_length : int, optional
        Hop length of STFT window, by default 512
    weight : float, optional
        Weight of loss, by default 1.0
    r	   r
   r   r   r   r   c                    s    t ƒ  ¡  || _t||ƒ| _d S )N)r$   r%   r   r   r&   )r'   r   r   r   r(   r    r!   r%   Ð   s   
zPhaseLoss.__init__r*   r+   c           	      C   s°   | j }| |j|j|j¡ | |j|j|j¡ |j|j }||tj k   dtj 7  < ||tjk  dtj 8  < |j 	¡ |j 
¡ }}|j| ||  }|| d  ¡ }|S )a.  Computes phase loss between an estimate and a reference
        signal.

        Parameters
        ----------
        x : AudioSignal
            Estimate signal
        y : AudioSignal
            Reference signal

        Returns
        -------
        torch.Tensor
            Phase loss.
        r   éþÿÿÿ)r&   r-   r   r   r   ÚphaseÚnpÚpir.   ÚminÚmaxÚmean)	r'   r*   r+   r2   ÚdiffÚx_minÚx_maxÚweightsr1   r    r    r!   r3   Ø   s   zPhaseLoss.forward)r	   r
   r   )
r5   r6   r7   r8   r:   r=   r%   r   r3   r@   r    r    r(   r!   rL   Ã   s    ÿÿÿÿrL   )r;   r   ÚnumpyrO   Útorchr   Ú r   r   ÚModuler   rA   rL   r    r    r    r!   Ú<module>   s    Wa