§
    9²¸jºÆ  ã                   ó  — d dl Z d dlm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 ddlmZ ddlmZ dd	lmZmZ dd
lmZ ddlmZ ddlmZmZ ddlmZmZ ddlmZ ddl m!Z!m"Z"m#Z#m$Z$m%Z% ddl&m'Z' ddl(m)Z) ddl*m+Z+ ddl,m-Z-  e%¦   «         rd dl.Z.d dl/mc m0Z1 e#e G d„ de!¦  «        ¦   «         ¦   «         Z2e#e G d„ de!¦  «        ¦   «         ¦   «         Z3e#e G d„ de!¦  «        ¦   «         ¦   «         Z4 G d„ dej5        ¦  «        Z6d„ Z7 ed¦  «        dYd „¦   «         Z8d!e.j9        d"e:d#e.j9        fd$„Z;	 dZd&ej5        d'e.j9        d(e.j9        d)e.j9        d*e.j9        dz  d+e<d,e<d-ee"         fd.„Z= ee8¦  «         G d/„ d0ej5        ¦  «        ¦   «         Z> ed1¦  «         G d2„ d3ej5        ¦  «        ¦   «         Z? G d4„ d5e¦  «        Z@ G d6„ d7ej5        ¦  «        ZAd8„ ZB G d9„ d:ej5        ¦  «        ZC G d;„ d<ej5        ¦  «        ZD G d=„ d>ej5        ¦  «        ZE G d?„ d@ej5        ¦  «        ZFe# G dA„ dBe¦  «        ¦   «         ZG G dC„ dDej5        ¦  «        ZH G dE„ dFej5        ¦  «        ZI G dG„ dHej5        ¦  «        ZJ G dI„ dJej5        ¦  «        ZK G dK„ dLej5        ¦  «        ZL G dM„ dNej5        ¦  «        ZM G dO„ dPej5        ¦  «        ZN G dQ„ dRej5        ¦  «        ZO G dS„ dTej5        ¦  «        ZP e#dU¬V¦  «         G dW„ dXeG¦  «        ¦   «         ZQdXdBgZRdS )[é    N)ÚCallable)Ú	dataclass)Ú	Parameteré   )Úinitialization)ÚACT2FN)ÚCache)Úuse_kernel_forward_from_hubÚuse_kernelized_func)Úcreate_bidirectional_mask)ÚGradientCheckpointingLayer)ÚROPE_INIT_FUNCTIONSÚdynamic_rope_update)ÚALL_ATTENTION_FUNCTIONSÚPreTrainedModel)ÚUnpack)ÚModelOutputÚTransformersKwargsÚauto_docstringÚcan_return_tupleÚis_torch_available)Údeprecate_kwarg)Úmaybe_autocasté   )Ú	AutoModelé   )ÚNeuCodecConfigc                   óŒ   — e Zd ZU dZdZej        dz  ed<   dZej	        dz  ed<   dZ
ej        dz  ed<   dZej        dz  ed<   dS )ÚNeuCodecOutputaM  
    audio_values (`torch.FloatTensor` of shape `(batch_size, 1, sequence_length)`, *optional*):
        Decoded audio waveform values in the time domain, obtained using the decoder
        part of NeuCodec. These represent the reconstructed audio signal.
    audio_codes (`torch.LongTensor` of shape `(batch_size, 1, codes_length)`, *optional*):
        Discrete code embeddings computed using `model.encode`. These are the quantized
        representations of the input audio used for further processing or generation.
    latents (`torch.Tensor` of shape `(batch_size, dimension, time_steps)`):
        Quantized continuous representation of input's embedding.
    audio_codes_mask (`torch.int32` of shape `(batch_size, 1, codes_length)`, *optional*):
        Downsampled `padding_mask` for indicating valid audio codes in `audio_codes`.
    NÚaudio_valuesÚaudio_codesÚlatentsÚaudio_codes_mask)Ú__name__Ú
__module__Ú__qualname__Ú__doc__r    ÚtorchÚFloatTensorÚ__annotations__r!   Ú
LongTensorr"   ÚTensorr#   © ó    ú~/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/transformers/models/neucodec/modeling_neucodec.pyr   r   2   s}   € € € € € € ðð ð .2€L�%Ô# dÑ*Ð1Ð1Ñ1Ø+/€K�Ô! DÑ(Ð/Ð/Ñ/Ø#'€GˆUŒ\˜DÑ Ð'Ð'Ñ'Ø,0Ð�e”l TÑ)Ð0Ð0Ñ0Ð0Ð0r.   r   c                   ón   — e Zd ZU dZdZej        dz  ed<   dZej	        dz  ed<   dZ
ej	        dz  ed<   dS )ÚNeuCodecEncoderOutputat  
    audio_codes (`torch.LongTensor` of shape `(batch_size, 1, codes_length)`, *optional*):
        Discrete code embeddings computed using `model.encode`. These represent
        the compressed, quantized form of the input audio signal that can be
        used for storage, transmission, or generation.
    latents (`torch.Tensor` of shape `(batch_size, dimension, time_steps)`):
        Quantized continuous representation of input's embedding.
    audio_codes_mask (`torch.int32` of shape `(batch_size, 1, codes_length)`, *optional*):
        Downsampled `padding_mask` for indicating valid audio codes in `audio_codes`.
    Nr!   r"   r#   )r$   r%   r&   r'   r!   r(   r+   r*   r"   r,   r#   r-   r.   r/   r1   r1   H   se   € € € € € € ð	ð 	ð ,0€K�Ô! DÑ(Ð/Ð/Ñ/Ø#'€GˆUŒ\˜DÑ Ð'Ð'Ñ'Ø,0Ð�e”l TÑ)Ð0Ð0Ñ0Ð0Ð0r.   r1   c                   ó2   — e Zd ZU dZdZej        dz  ed<   dS )ÚNeuCodecDecoderOutputa=  
    audio_values (`torch.FloatTensor` of shape `(batch_size, 1, segment_length)`, *optional*):
        Decoded audio waveform values in the time domain, obtained by converting
        the discrete codes back into continuous audio signals. This represents
        the reconstructed audio that can be played back.
    Nr    )r$   r%   r&   r'   r    r(   r)   r*   r-   r.   r/   r3   r3   [   s6   € € € € € € ðð ð .2€L�%Ô# dÑ*Ð1Ð1Ñ1Ð1Ð1r.   r3   c                   óH   ‡ — e Zd Zdefˆ fd„Zdej        dej        fd„Zˆ xZS )ÚNeuCodecMLPÚconfigc                 ó  •— t          ¦   «                              ¦   «          || _        t          |j                 | _        t          j        |j        |j	        d¬¦  «        | _
        t          j        |j	        |j        d¬¦  «        | _        d S )NF©Úbias)ÚsuperÚ__init__r6   r   Ú
hidden_actÚactivation_fnÚnnÚLinearÚhidden_sizeÚintermediate_sizeÚfc1Úfc2©Úselfr6   Ú	__class__s     €r/   r;   zNeuCodecMLP.__init__i   sr   ø€ Ý‰Œ×ÒÑÔÐØˆŒÝ# FÔ$5Ô6ˆÔÝ”9˜VÔ/°Ô1IÐPUÐVÑVÔVˆŒÝ”9˜VÔ5°vÔ7IÐPUÐVÑVÔVˆŒˆˆr.   Úhidden_statesÚreturnc                 ó„   — |                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|S ©N)rB   r=   rC   ©rE   rG   s     r/   ÚforwardzNeuCodecMLP.forwardp   s=   € ØŸš Ñ/Ô/ˆØ×*Ò*¨=Ñ9Ô9ˆØŸš Ñ/Ô/ˆØÐr.   ©	r$   r%   r&   r   r;   r(   r,   rL   Ú__classcell__©rF   s   @r/   r5   r5   h   sq   ø€ € € € € ðW˜~ð Wð Wð Wð Wð Wð Wð U¤\ð °e´lð ð ð ð ð ð ð ð r.   r5   c                 óœ   — | dd| j         d         dz  …f         }| d| j         d         dz  d…f         }t          j        | |fd¬¦  «        S )z*Rotates half the hidden dims of the input..Néÿÿÿÿr   ©Údim)Úshaper(   Úcat)ÚxÚx1Úx2s      r/   Úrotate_halfrY   w   s]   € à	
ˆ3Ð"�!”'˜"”+ Ñ"Ð"Ð"Ô	#€BØ	
ˆ3�”˜”˜qÑ Ð"Ð"Ð"Ô	#€BÝŒ9�r�c˜2�Y BÐ'Ñ'Ô'Ð'r.   Úrotary_pos_embc                 ó¾   — |                      |¦  «        }|                      |¦  «        }| |z  t          | ¦  «        |z  z   }||z  t          |¦  «        |z  z   }||fS )a…  Applies Rotary Position Embedding to the query and key tensors.

    Args:
        q (`torch.Tensor`): The query tensor.
        k (`torch.Tensor`): The key tensor.
        cos (`torch.Tensor`): The cosine part of the rotary embedding.
        sin (`torch.Tensor`): The sine part of the rotary embedding.
        unsqueeze_dim (`int`, *optional*, defaults to 1):
            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
    Returns:
        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
    )Ú	unsqueezerY   )ÚqÚkÚcosÚsinÚunsqueeze_dimÚq_embedÚk_embeds          r/   Úapply_rotary_pos_embrd   ~   sc   € ð& �-Š-˜Ñ
&Ô
&€CØ
�-Š-˜Ñ
&Ô
&€CØ�3‰w�; q™>œ>¨CÑ/Ñ0€GØ�3‰w�; q™>œ>¨CÑ/Ñ0€GØ�GÐÐr.   rG   Ún_reprH   c                 ó¸   — | j         \  }}}}|dk    r| S | dd…dd…ddd…dd…f                              |||||¦  «        } |                      |||z  ||¦  «        S )zÔ
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    r   N)rT   ÚexpandÚreshape)rG   re   ÚbatchÚnum_key_value_headsÚslenÚhead_dims         r/   Ú	repeat_kvrm   ˜   s„   € ð
 2?Ô1DÑ.€EÐ  hØ�‚z€zØÐØ! ! ! ! Q Q Q¨¨a¨a¨a°°°Ð"2Ô3×:Ò:¸5ÐBUÐW\Ð^bÐdlÑmÔm€MØ× Ò  Ð(;¸eÑ(CÀTÈ8ÑTÔTÐTr.   ç        ÚmoduleÚqueryÚkeyÚvalueÚattention_maskÚscalingÚdropoutÚkwargsc                 ó  — t          || j        ¦  «        }t          || j        ¦  «        }	t          j        ||                     dd¦  «        ¦  «        |z  }
|�|
|z   }
t
          j                             |
dt          j        ¬¦  «         	                    |j
        ¦  «        }
t
          j                             |
|| j        ¬¦  «        }
t          j        |
|	¦  «        }|                     dd¦  «                             ¦   «         }||
fS )Nr   r   rQ   ©rS   Údtype©ÚpÚtrainingr   )rm   Únum_key_value_groupsr(   ÚmatmulÚ	transposer>   Ú
functionalÚsoftmaxÚfloat32Útory   ru   r|   Ú
contiguous)ro   rp   rq   rr   rs   rt   ru   rv   Ú
key_statesÚvalue_statesÚattn_weightsÚattn_outputs               r/   Úeager_attention_forwardr‰   ¤   sé   € õ ˜3 Ô ;Ñ<Ô<€JÝ˜U FÔ$?Ñ@Ô@€Lå”<  z×';Ò';¸A¸qÑ'AÔ'AÑBÔBÀWÑL€LØÐ!Ø# nÑ4ˆå”=×(Ò(¨¸2ÅUÄ]Ð(ÑSÔS×VÒVÐW\ÔWbÑcÔc€LÝ”=×(Ò(¨¸È6Ì?Ð(Ñ[Ô[€LÝ”,˜|¨\Ñ:Ô:€KØ×'Ò'¨¨1Ñ-Ô-×8Ò8Ñ:Ô:€Kà˜Ð$Ð$r.   c                   óÎ   ‡ — e Zd ZdZdedefˆ fd„Z	 	 	 ddej        de	ej        ej        f         dz  dej        dz  d	e
dz  d
ee         de	ej        ej        f         fd„Zˆ xZS )ÚNeuCodecAttentionz=Multi-headed attention from 'Attention Is All You Need' paperr6   Ú	layer_idxc                 ó®  •— t          ¦   «                              ¦   «          || _        || _        t	          |d|j        |j        z  ¦  «        | _        |j        |j        z  | _	        | j        dz  | _
        |j        | _        d| _        t          j        |j        |j        | j        z  |j        ¬¦  «        | _        t          j        |j        |j        | j        z  |j        ¬¦  «        | _        t          j        |j        |j        | j        z  |j        ¬¦  «        | _        t          j        |j        | j        z  |j        |j        ¬¦  «        | _        d S )Nrl   g      à¿Fr8   )r:   r;   r6   rŒ   Úgetattrr@   Únum_attention_headsrl   rj   r}   rt   Úattention_dropoutÚ	is_causalr>   r?   Úattention_biasÚq_projÚk_projÚv_projÚo_proj©rE   r6   rŒ   rF   s      €r/   r;   zNeuCodecAttention.__init__Á   sB  ø€ Ý‰Œ×ÒÑÔÐØˆŒØ"ˆŒÝ ¨
°FÔ4FÈ&ÔJdÑ4dÑeÔeˆŒØ$*Ô$>À&ÔB\Ñ$\ˆÔ!Ø”} dÑ*ˆŒØ!'Ô!9ˆÔØˆŒå”iØÔ Ô :¸T¼]Ñ JÐQWÔQfð
ñ 
ô 
ˆŒõ ”iØÔ Ô :¸T¼]Ñ JÐQWÔQfð
ñ 
ô 
ˆŒõ ”iØÔ Ô :¸T¼]Ñ JÐQWÔQfð
ñ 
ô 
ˆŒõ ”iØÔ&¨¬Ñ6¸Ô8JÐQWÔQfð
ñ 
ô 
ˆŒˆˆr.   NrG   Úposition_embeddingsrs   Úpast_key_valuesrv   rH   c                 ó&  — |j         d d…         }g |¢d‘| j        ‘R }|                      |¦  «                             |¦  «                             dd¦  «        }|                      |¦  «                             |¦  «                             dd¦  «        }	|                      |¦  «                             |¦  «                             dd¦  «        }
|\  }}t          ||	||d¬¦  «        \  }}	|�|                     |	|
| j	        ¦  «        \  }	}
t          j        | j        j        t          ¦  «        } || ||	|
|f| j        sdn| j        | j        dœ|¤Ž\  }} |j        g |¢d‘R Ž                      ¦   «         }|                      |¦  «        }||fS )NrQ   r   r   )ra   rn   )ru   rt   )rT   rl   r“   Úviewr   r”   r•   rd   ÚupdaterŒ   r   Úget_interfacer6   Ú_attn_implementationr‰   r|   r�   rt   rh   r„   r–   )rE   rG   r˜   rs   r™   rv   Úinput_shapeÚhidden_shapeÚquery_statesr…   r†   r_   r`   Úattention_interfacerˆ   r‡   s                   r/   rL   zNeuCodecAttention.forwardØ   sÈ  € ð $Ô)¨#¨2¨#Ô.ˆØ8˜Ð8 bÐ8¨$¬-Ð8Ð8ˆà—{’{ =Ñ1Ô1×6Ò6°|ÑDÔD×NÒNÈqÐRSÑTÔTˆØ—[’[ Ñ/Ô/×4Ò4°\ÑBÔB×LÒLÈQÐPQÑRÔRˆ
Ø—{’{ =Ñ1Ô1×6Ò6°|ÑDÔD×NÒNÈqÐRSÑTÔTˆà&‰ˆˆSõ $8¸ÀjÐRUÐWZÐjkÐ#lÑ#lÔ#lÑ ˆ�jàÐ&Ø'6×'=Ò'=¸jÈ,ÐX\ÔXfÑ'gÔ'gÑ$ˆJ˜å(?Ô(MØŒKÔ,Õ.Eñ)
ô )
Ðð %8Ð$7ØØØØØð	%
ð  $œ}ÐH�C�C°$Ô2HØ”Lð	%
ð 	%
ð ð	%
ð 	%
Ñ!ˆ�\ð *�kÔ)Ð;¨;Ð;¸Ð;Ð;Ð;×FÒFÑHÔHˆØ—k’k +Ñ.Ô.ˆØ˜LÐ(Ð(r.   ©NNN)r$   r%   r&   r'   r   Úintr;   r(   r,   Útupler	   r   r   rL   rN   rO   s   @r/   r‹   r‹   ½   så   ø€ € € € € àGÐGð
˜~ð 
¸#ð 
ð 
ð 
ð 
ð 
ð 
ð4 IMØ.2Ø(,ð))ð ))à”|ð))ð # 5¤<°´Ð#=Ô>ÀÑEð))ð œ tÑ+ð	))ð
  ™ð))ð Ð+Ô,ð))ð 
ˆuŒ|˜Uœ\Ð)Ô	*ð))ð ))ð ))ð ))ð ))ð ))ð ))ð ))r.   r‹   ÚRMSNormc                   óT   ‡ — e Zd Zd	deddfˆ fd„Zdej        dej        fd„Zd„ Zˆ xZ	S )
ÚNeuCodecRMSNormç�íµ ÷Æ°>ÚepsrH   Nc                 ó¬   •— t          ¦   «                              ¦   «          t          j        t	          j        |¦  «        ¦  «        | _        || _        dS )z>
        NeuCodecRMSNorm is equivalent to T5LayerNorm
        N)r:   r;   r>   r   r(   ÚonesÚweightÚvariance_epsilon)rE   r@   rª   rF   s      €r/   r;   zNeuCodecRMSNorm.__init__  sD   ø€ õ 	‰Œ×ÒÑÔÐÝ”l¥5¤:¨kÑ#:Ô#:Ñ;Ô;ˆŒØ #ˆÔÐÐr.   rG   c                 ó  — |j         }|                     t          j        ¦  «        }|                     d¦  «                             dd¬¦  «        }|t          j        || j        z   ¦  «        z  }| j        |                     |¦  «        z  S )Nr   rQ   T)Úkeepdim)	ry   rƒ   r(   r‚   ÚpowÚmeanÚrsqrtr®   r­   )rE   rG   Úinput_dtypeÚvariances       r/   rL   zNeuCodecRMSNorm.forward  s|   € Ø#Ô)ˆØ%×(Ò(­¬Ñ7Ô7ˆØ ×$Ò$ QÑ'Ô'×,Ò,¨R¸Ð,Ñ>Ô>ˆØ%­¬°H¸tÔ?TÑ4TÑ(UÔ(UÑUˆØŒ{˜]×-Ò-¨kÑ:Ô:Ñ:Ð:r.   c                 óH   — t          | j        j        ¦  «        › d| j        › �S )Nz, eps=)r¥   r­   rT   r®   )rE   s    r/   Ú
extra_reprzNeuCodecRMSNorm.extra_repr  s&   € Ý˜œÔ)Ñ*Ô*ÐIÐI°$Ô2GÐIÐIÐIr.   )r©   )
r$   r%   r&   Úfloatr;   r(   r,   rL   r·   rN   rO   s   @r/   r¨   r¨     sŒ   ø€ € € € € ð$ð $¨ð $¸$ð $ð $ð $ð $ð $ð $ð; U¤\ð ;°e´lð ;ð ;ð ;ð ;ðJð Jð Jð Jð Jð Jð Jr.   r¨   c                   óÒ   ‡ — e Zd Zdedefˆ fd„Z	 	 	 	 	 ddej        dej        dz  dej        dz  d	e	dz  d
e
dz  deej        ej        f         dz  dee         dej        fd„Zˆ xZS )ÚNeuCodecDecoderLayerr6   rŒ   c                 ó4  •— t          ¦   «                              ¦   «          |j        | _        t          ||¬¦  «        | _        t          |¦  «        | _        t          |j        |j        ¬¦  «        | _	        t          |j        |j        ¬¦  «        | _
        d S )N)r6   rŒ   ©rª   )r:   r;   r@   r‹   Ú	self_attnr5   Úmlpr¨   Úrms_norm_epsÚinput_layernormÚpost_attention_layernormr—   s      €r/   r;   zNeuCodecDecoderLayer.__init__  s„   ø€ Ý‰Œ×ÒÑÔÐØ!Ô-ˆÔå*°&ÀIÐNÑNÔNˆŒå˜vÑ&Ô&ˆŒÝ.¨vÔ/AÀvÔGZÐ[Ñ[Ô[ˆÔÝ(7¸Ô8JÐPVÔPcÐ(dÑ(dÔ(dˆÔ%Ð%Ð%r.   NFrG   rs   Úposition_idsr™   Ú	use_cacher˜   rv   rH   c           
      óÎ   — |}|                       |¦  «        } | j        d||||||dœ|¤Ž\  }}	||z   }|}|                      |¦  «        }|                      |¦  «        }||z   }|S )N)rG   rs   rÂ   r™   rÃ   r˜   r-   )rÀ   r½   rÁ   r¾   )
rE   rG   rs   rÂ   r™   rÃ   r˜   rv   ÚresidualÚ_s
             r/   rL   zNeuCodecDecoderLayer.forward$  s¡   € ð !ˆØ×,Ò,¨]Ñ;Ô;ˆà)˜4œ>ð 
Ø'Ø)Ø%Ø+ØØ 3ð
ð 
ð ð
ð 
Ñˆ�qð ! =Ñ0ˆð !ˆØ×5Ò5°mÑDÔDˆØŸš Ñ/Ô/ˆØ  =Ñ0ˆØÐr.   )NNNFN)r$   r%   r&   r   r¤   r;   r(   r,   r+   r	   Úboolr¥   r   r   rL   rN   rO   s   @r/   rº   rº     sÿ   ø€ € € € € ðe˜~ð e¸#ð eð eð eð eð eð eð /3Ø04Ø(,Ø!&ØHLðð à”|ðð œ tÑ+ðð Ô&¨Ñ-ð	ð
  ™ðð ˜$‘;ðð # 5¤<°´Ð#=Ô>ÀÑEðð Ð+Ô,ðð 
Œðð ð ð ð ð ð ð r.   rº   c                   ó*   ‡ — e Zd ZdZdˆ fd„	Zd„ Zˆ xZS )ÚNeuCodecSnakeBetaa  
    A modified Snake function which uses separate parameters for the magnitude of the periodic components
    Shape:
        - Input: (B, C, T)
        - Output: (B, C, T), same shape as the input
    Parameters:
        - alpha - trainable parameter that controls frequency
        - beta - trainable parameter that controls magnitude
    References:
        - This activation function is a modified version based on this paper by Liu Ziyin, Tilman Hartwig, Masahito Ueda:
        https://huggingface.co/papers/2006.08195
    ç      ð?c                 ó  •— t          ¦   «                              ¦   «          || _        t          t	          j        |¦  «        |z  ¦  «        | _        t          t	          j        |¦  «        |z  ¦  «        | _        d| _        d S )Ng•Ö&è.>)	r:   r;   Úin_featuresr   r(   ÚzerosÚalphaÚbetaÚno_div_by_zero)rE   rÌ   rÎ   rF   s      €r/   r;   zNeuCodecSnakeBeta.__init__R  sm   ø€ Ý‰Œ×ÒÑÔÐØ&ˆÔõ �uœ{¨;Ñ7Ô7¸%Ñ?Ñ@Ô@ˆŒ
Ý�eœk¨+Ñ6Ô6¸Ñ>Ñ?Ô?ˆŒ	à)ˆÔÐÐr.   c                 ó€  — | j                              d¦  «                             d¦  «        }| j                             d¦  «                             d¦  «        }t          j        |¦  «        }t          j        |¦  «        }|d|| j        z   z  t          j        t          j        ||z  ¦  «        d¦  «        z  z   }|S )u’   
        Forward pass of the function.
        Applies the function to the input elementwise.
        SnakeBeta âˆ¶= x + 1/b * sin^2 (xa)
        r   rQ   rÊ   r   )rÎ   r\   rÏ   r(   ÚexprÐ   r±   r`   )rE   rG   rÎ   rÏ   s       r/   rL   zNeuCodecSnakeBeta.forward\  s°   € ð ”
×$Ò$ QÑ'Ô'×1Ò1°"Ñ5Ô5ˆØŒy×"Ò" 1Ñ%Ô%×/Ò/°Ñ3Ô3ˆÝ”	˜%Ñ Ô ˆÝŒy˜‰ŒˆØ%¨°°tÔ7JÑ0JÑ)KÍuÌyÝŒI�m eÑ+Ñ,Ô,¨añP
ô P
ñ )
ñ 
ˆð Ðr.   )rÊ   ©r$   r%   r&   r'   r;   rL   rN   rO   s   @r/   rÉ   rÉ   D  sV   ø€ € € € € ðð ð*ð *ð *ð *ð *ð *ðð ð ð ð ð ð r.   rÉ   c                 óX  — |dz  dk    }|dz  }d|z  }d|dz
  z  t           j        z  |z  dz   }|dk    r	d|d	z
  z  }n|d
k    rd|dz
  dz  z  d|d
z
  z  z   }nd}t          j        ||dt          j        ¬¦  «        }|rt          j        | |¦  «        dz   }	nt          j        |¦  «        |z
  }	| dk    r#t          j        dd|ft          j        ¬¦  «        S t          j        d| z  |	z  ¦  «        }
d| z  |z  |
z  }||                     ¦   «         z  }| 	                    dd|¦  «        S )aB  Generates a 1D Kaiser-windowed sinc filter.

    Args:
        cutoff (float): Normalized cutoff frequency (0 to 0.5).
        half_width (float): Transition bandwidth.
        kernel_size (int): Number of filter taps.

    Returns:
        torch.Tensor: A tensor of shape (1, 1, kernel_size) representing the filter.
    r   r   é   gHáz®G@r   gÍÌÌÌÌÌ@g      I@gKê46¼?gffffff!@g      5@g¨WÊ2Ä±â?é   gš™™™™™Ù?gUjö@+0´?rn   F)rÏ   Úperiodicry   ç      à?©ry   )
ÚmathÚpir(   Úkaiser_windowr‚   ÚarangerÍ   ÚsincÚsumr›   )ÚcutoffÚ
half_widthÚkernel_sizeÚis_evenÚ	half_sizeÚdelta_fÚattenuationrÏ   rÜ   Útime_indicesÚsinc_filterÚnormalized_filters               r/   Úkaiser_sinc_filter1drê   m  ss  € ð ˜A‰o Ò"€GØ˜qÑ €Ið �*‰n€GØ˜9 q™=Ñ)­D¬GÑ3°gÑ=ÀÑD€Kà�TÒÐØ˜ sÑ*Ñ+ˆˆØ	˜Ò	Ð	Ø˜ rÑ)¨cÑ1Ñ1°G¸{ÈTÑ?QÑ4RÑRˆˆàˆåÔ'¨¸$ÈÕV[ÔVcÐdÑdÔd€Mð ð =Ý”| Y J°	Ñ:Ô:¸SÑ@ˆˆå”| KÑ0Ô0°9Ñ<ˆð �‚{€{ÝŒ{˜A˜q +Ð.µe´mÐDÑDÔDÐDå”*˜Q ™Z¨,Ñ6Ñ7Ô7€KØ˜F™
 ]Ñ2°[Ñ@Ðð Ð*×.Ò.Ñ0Ô0Ñ0Ðà×!Ò! ! Q¨Ñ4Ô4Ð4r.   c                   ó&   ‡ — e Zd Zdˆ fd„	Zd„ Zˆ xZS )ÚNeuCodecDownSample1dr   Nc                 ó®  •— t          ¦   «                              ¦   «          d|z  }d|z  }|| _        || _        || _        |dk     rt          d¦  «        ‚|dk    rt          d¦  «        ‚|dz  dk    | _        |dz  t          | j        ¦  «        z
  | _        |dz  | _	        || _
        t          |||¦  «        }t          j        |d¬	¦  «        | _        d S )
NrØ   ç333333ã?rn   z(Minimum cutoff must be larger than zero.z'A cutoff above 0.5 does not make sense.r   r   F©Ú
persistent)r:   r;   rà   rá   râ   Ú
ValueErrorÚevenr¤   Úpad_leftÚ	pad_rightÚstriderê   r>   ÚBufferÚfilter)rE   Úratiorâ   rà   rá   r÷   rF   s         €r/   r;   zNeuCodecDownSample1d.__init__œ  sØ   ø€ Ý‰Œ×ÒÑÔÐØ�u‘ˆØ˜5‘[ˆ
ØˆŒØ$ˆŒØ&ˆÔà�CŠ<ˆ<ÝÐGÑHÔHÐHØ�CŠ<ˆ<ÝÐFÑGÔGÐGà !‘O qÒ(ˆŒ	Ø# qÑ(­3¨t¬y©>¬>Ñ9ˆŒØ$¨Ñ)ˆŒØˆŒÝ% f¨j¸+ÑFÔFˆÝ”i °5Ð9Ñ9Ô9ˆŒˆˆr.   c                 ó  — |j         d         }t          j        || j        | j        fd¬¦  «        }t          j        || j                             |j        ¦  «         	                    |dd¦  «        | j
        |¬¦  «        }|S )Nr   Ú	replicate©ÚmoderQ   ©rõ   Úgroups)rT   ÚFÚpadró   rô   Úconv1dr÷   rƒ   ry   rg   rõ   )rE   rG   ÚchannelsÚouts       r/   rL   zNeuCodecDownSample1d.forward°  s}   € Ø Ô& qÔ)ˆÝœ˜m¨d¬m¸T¼^Ð-LÐS^Ð_Ñ_Ô_ˆÝŒhØàŒK�NŠN˜=Ô.Ñ/Ô/×6Ò6°xÀÀRÑHÔHØ”;Øð
ñ 
ô 
ˆð ˆ
r.   ©r   N©r$   r%   r&   r;   rL   rN   rO   s   @r/   rì   rì   ›  sL   ø€ € € € € ð:ð :ð :ð :ð :ð :ð(
ð 
ð 
ð 
ð 
ð 
ð 
r.   rì   c                   ó&   ‡ — e Zd Zdˆ fd„	Zd„ Zˆ xZS )ÚNeuCodecUpSample1dr   Nc                 óÜ  •— t          ¦   «                              ¦   «          || _        |€t          d|z  dz  ¦  «        dz  n|| _        || _        | j        |z  dz
  | _        | j        | j        z  | j        | j        z
  dz  z   | _        | j        | j        z  | j        | j        z
  dz   dz  z   | _        t          d|z  d|z  | j        ¬¦  «        }t          j        |d¬¦  «        | _        d S )	Né   r   r   rØ   rî   )rà   rá   râ   Frï   )r:   r;   rø   r¤   râ   rõ   r   ró   rô   rê   r>   rö   r÷   )rE   rø   râ   r÷   rF   s       €r/   r;   zNeuCodecUpSample1d.__init__¾  sì   ø€ Ý‰Œ×ÒÑÔÐØˆŒ
Ø6AÐ6I�3˜q 5™y¨A™~Ñ.Ô.°Ñ2Ð2È{ˆÔØˆŒØÔ# uÑ,¨qÑ0ˆŒØœ 4¤;Ñ.°$Ô2BÀTÄ[Ñ2PÐUVÑ1VÑVˆŒØœ D¤KÑ/°4Ô3CÀdÄkÑ3QÐTUÑ3UÐZ[Ñ2[Ñ[ˆŒå%¨S°5©[ÀSÈ5Á[Ð^bÔ^nÐoÑoÔoˆÝ”i °5Ð9Ñ9Ô9ˆŒˆˆr.   c           	      óB  — |j         d         }t          j        || j        | j        fd¬¦  «        }| j        t          j        || j                             |j        ¦  «                             |dd¦  «        | j	        |¬¦  «        z  }|d| j
        | j         …f         }|S )Nr   rú   rû   rQ   rý   .)rT   rÿ   r   rø   Úconv_transpose1dr÷   rƒ   ry   rg   rõ   ró   rô   )rE   rG   r  s      r/   rL   zNeuCodecUpSample1d.forwardÊ  s¡   € Ø Ô& qÔ)ˆÝœ˜m¨d¬h¸¼Ð-AÈÐTÑTÔTˆØœ
¥QÔ%7ØàŒK�NŠN˜=Ô.Ñ/Ô/×6Ò6°xÀÀRÑHÔHØ”;Øð&
ñ &
ô &
ñ 
ˆð & c¨4¬=¸D¼N¸?Ð+JÐ&JÔKˆØÐr.   r  r  rO   s   @r/   r  r  ½  sL   ø€ € € € € ð
:ð 
:ð 
:ð 
:ð 
:ð 
:ðð ð ð ð ð ð r.   r  c                   ó¼   ‡ — e Zd ZdZdefˆ fd„Zdd„Zdej        dej        fd„Z	dd
ej        de
dej        fd„Zd
ej        deej        ej        f         fd„Zˆ xZS )Ú NeuCodecFiniteScalarQuantizationa#  
    Finite Scalar Quantization (FSQ) module that quantizes continuous latent representations into discrete codes.
    Original code: https://github.com/lucidrains/vector-quantize-pytorch/blob/353d46027888dfb140c3c65a67a7356f1492d71d/vector_quantize_pytorch/finite_scalar_quantization.py#L64

    Original modeling uses `ResidualFSQ` with a single quantizer: https://huggingface.co/HKUSTAudio/neucodec/blob/main/vq/codec_decoder_vocos.py#L389
    But we can directly use FSQ since a main feature of NeuCodec is that it uses a single codebook.
    r6   c                 óL  •— t          ¦   «                              ¦   «          t          |j        ¦  «        | _        |                      ¦   «         \  }}}t          j        |d¬¦  «        | _        t          j        |d¬¦  «        | _        t          j        |d¬¦  «        | _	        d S )NFrï   )
r:   r;   ÚlistÚquantization_levelsÚ_compute_buffersr>   rö   ÚlevelsÚbasisÚcodebook)rE   r6   r  r  r  rF   s        €r/   r;   z)NeuCodecFiniteScalarQuantization.__init__á  sŠ   ø€ Ý‰Œ×ÒÑÔÐÝ#'¨Ô(BÑ#CÔ#CˆÔ Ø"&×"7Ò"7Ñ"9Ô"9Ñˆ��xÝ”i °5Ð9Ñ9Ô9ˆŒÝ”Y˜u°Ð7Ñ7Ô7ˆŒ
Ýœ	 (°uÐ=Ñ=Ô=ˆŒˆˆr.   Nc                 ó¨  — t          j        | j        t           j        |¬¦  «        }t          j        t          j        dg| j        dd…         z   |¬¦  «        dt           j        ¬¦  «        }t          j        t          t          j        | j        ¦  «        ¦  «        |¬¦  «         	                    d¦  «        }||z  |z  }|dz  }||z
  |z  }|||fS )	zFCompute the levels, basis, and codebook buffers for the FSQ quantizer.©ry   Údevicer   NrQ   ©r  r   rx   r   )
r(   Útensorr  Úint32ÚcumprodrÝ   r¤   ÚnpÚprodr\   )rE   r  r  r  ÚindicesÚlevel_indicesrá   r  s           r/   r  z1NeuCodecFiniteScalarQuantization._compute_buffersé  sÐ   € å”˜dÔ6½e¼kÐRXÐYÑYÔYˆÝ”ÝŒL˜!˜˜tÔ7¸¸¸Ô<Ñ<ÀVÐLÑLÔLÐRSÕ[`Ô[fð
ñ 
ô 
ˆõ ”,�s¥2¤7¨4Ô+CÑ#DÔ#DÑEÔEÈfÐUÑUÔU×_Ò_Ð`bÑcÔcˆØ  EÑ)¨VÑ3ˆØ˜q‘[ˆ
Ø! JÑ.°*Ñ<ˆØ�u˜hÐ&Ð&r.   r  rH   c                 óx   — |                      d¦  «        }|| j        z  | j        z  }| j        dz  }||z
  |z  }|S )z`
        Convert integer codebook indices to normalized per-dimension codes in [-1, 1].
        rQ   r   )r\   r  r  )rE   r  r  rá   Úcodess        r/   Ú_indices_to_codesz2NeuCodecFiniteScalarQuantization._indices_to_codesõ  sJ   € ð ×#Ò# BÑ'Ô'ˆØ  D¤JÑ.°$´+Ñ=ˆØ”[ AÑ%ˆ
Ø Ñ+¨zÑ9ˆØˆr.   çü©ñÒMbP?rG   rª   c                 óÔ   — | j         dz
  d|z   z  dz  }t          j        | j         dz  dk    dd¦  «        }||z                       ¦   «         }||z                        ¦   «         |z  |z
  S )að  
        Constrain `hidden_states` to the valid quantization range for each dimension.

        Uses a scaled tanh to soft-clip values into the interval
        $[-(L-1)/2, (L-1)/2]$ (offset by 0.5 for even-level dimensions), where $L$ is
        the number of quantization levels. The small `eps` margin prevents values from
        saturating exactly at the boundary, which would zero out gradients.

        Args:
            hidden_states (`torch.Tensor`): Continuous input to be bounded.
            eps (`float`, *optional*, defaults to `1e-3`):
                Small margin added to the level range to avoid gradient saturation at boundaries.

        Returns:
            `torch.Tensor`: Bounded values in the valid quantization range.
        r   r   r   rØ   rn   )r  r(   ÚwhereÚatanhÚtanh)rE   rG   rª   Ú
half_rangeÚoffsetÚshifts         r/   Úboundz&NeuCodecFiniteScalarQuantization.boundÿ  sr   € ð" ”k A‘o¨!¨c©'Ñ2°QÑ6ˆ
Ý”˜Tœ[¨1™_°Ò1°3¸Ñ<Ô<ˆØ˜*Ñ$×+Ò+Ñ-Ô-ˆØ Ñ%×+Ò+Ñ-Ô-°
Ñ:¸VÑCÐCr.   c                 ó\  — |j         }t          |j        j        t          ¦  «        r|j        j        dk    r|j        j        nd}t          |d¬¦  «        5  |                     ¦   «         }| j        dz  }|                      |¦  «        }| 	                    ¦   «         }|||z
   
                    ¦   «         z   }||z  }||z  |z   }|| j        z                       d¬¦  «                             t          j        ¦  «        }d d d ¦  «         n# 1 swxY w Y   |                     |¦  «        |fS )NÚmpsÚcpuF©Údevice_typeÚenabledr   rQ   rR   )ry   Ú
isinstancer  ÚtypeÚstrr   r¸   r  r+  ÚroundÚdetachr  rß   rƒ   r(   r  )	rE   rG   Úoriginal_dtyper0  rá   Úroundedr!  Úcode_scaledr  s	            r/   rL   z(NeuCodecFiniteScalarQuantization.forward  s€  € à&Ô,ˆõ ˜-Ô.Ô3µSÑ9Ô9ðØ>KÔ>RÔ>WÐ[`Ò>`Ð>`ð Ô Ô%Ð%àð 	õ
 ¨¸UÐCÑCÔCð 
	Mð 
	MØ)×/Ò/Ñ1Ô1ˆMØœ¨Ñ)ˆJà ŸJšJ }Ñ5Ô5ˆMØ#×)Ò)Ñ+Ô+ˆGØ! W¨}Ñ%<×$DÒ$DÑ$FÔ$FÑFˆEØ˜JÑ&ˆEà  :Ñ-°Ñ;ˆKØ" T¤ZÑ/×4Ò4¸Ð4Ñ<Ô<×?Ò?ÅÄÑLÔLˆGð
	Mð 
	Mð 
	Mñ 
	Mô 
	Mð 
	Mð 
	Mð 
	Mð 
	Mð 
	Mð 
	Møøøð 
	Mð 
	Mð 
	Mð 
	Mð �xŠx˜Ñ'Ô'¨Ð0Ð0s   ÁB*DÄDÄDrJ   )r#  )r$   r%   r&   r'   r   r;   r  r(   r,   r"  r¸   r+  r¥   rL   rN   rO   s   @r/   r  r  Ø  sô   ø€ € € € € ðð ð>˜~ð >ð >ð >ð >ð >ð >ð
'ð 
'ð 
'ð 
'ð¨¬ð ¸%¼,ð ð ð ð ðDð D 5¤<ð D°eð DÀuÄ|ð Dð Dð Dð Dð,1 U¤\ð 1°e¸E¼LÈ%Ì,Ð<VÔ6Wð 1ð 1ð 1ð 1ð 1ð 1ð 1ð 1r.   r  c                   óL   ‡ — e Zd ZdZdefˆ fd„Zdej        dej        fd„Zˆ xZ	S )ÚNeuCodecISTFTHeadzù
    Head for converting decoder outputs to waveform via STFT projection and ISTFT.

    Uses custom "same" padding ISTFT from Vocos:
    https://github.com/gemelo-ai/vocos/blob/c859e3b7b534f3776a357983029d34170ddd6fc3/vocos/spectral_ops.py#L47
    r6   c                 ó\  •— t          ¦   «                              ¦   «          t          j        |j        |j        dz   ¦  «        | _        |j        | _        |j        | _        | j        | j        z
  dz  | _        t          j
        |j        ¦  «        }t          j        |d¬¦  «        | _        d S )Nr   Frï   )r:   r;   r>   r?   r@   Ún_fftÚlinearÚ
hop_lengthÚpaddingr(   Úhann_windowrö   Úwindow)rE   r6   rB  rF   s      €r/   r;   zNeuCodecISTFTHead.__init__3  sŠ   ø€ Ý‰Œ×ÒÑÔÐÝ”i Ô 2°F´LÀ1Ñ4DÑEÔEˆŒØ”\ˆŒ
Ø Ô+ˆŒØœ
 T¤_Ñ4¸Ñ:ˆŒÝÔ" 6¤<Ñ0Ô0ˆÝ”i °5Ð9Ñ9Ô9ˆŒˆˆr.   rG   rH   c                 ó  — |                       |¦  «        }|                     dd¬¦  «        \  }}|                     ¦   «         }|                     ¦   «         }t          j        |¦  «                             d¬¦  «        }t          j        ||¦  «        }t          j                             || j	        dd¬¦  «         
                    dd¦  «        }|| j        d d d …d f         z  }|j        d         }|dz
  | j        z  | j	        z   }t          j        |d|fd| j	        fd| j        f¬	¦  «        d d …d
d
| j        | j         …f         }	t          j        | j                             ¦   «                              d|d¦  «         
                    dd¦  «        d|fd| j	        fd| j        f¬	¦  «                             ¦   «         | j        | j         …         }
|
                     d¬¦  «        }
|	|
z  }	|	                     d¦  «        S )Nr   rQ   rR   g      Y@)ÚmaxÚbackward)rS   Únormr   )Úoutput_sizerâ   rõ   r   g•dyáý¥=)Úmin)r>  Úchunkr¸   r(   rÒ   ÚclampÚpolarÚfftÚirfftr=  r   rB  rT   r?  rÿ   Úfoldr@  Úsquarerg   Úsqueezer\   )rE   rG   Ú	stft_predÚ	magnitudeÚphaseÚspectrogram_complexÚtime_framesÚ
num_framesrG  ÚaudioÚwindow_envelopes              r/   rL   zNeuCodecISTFTHead.forward<  sú  € Ø—K’K Ñ.Ô.ˆ	à$Ÿ?š?¨1°"˜?Ñ5Ô5Ñˆ	�5à—O’OÑ%Ô%ˆ	Ø—’‘”ˆå”I˜iÑ(Ô(×.Ò.°3Ð.Ñ7Ô7ˆ	Ý#œk¨)°UÑ;Ô;Ðõ ”i—o’oÐ&9¸4¼:È2ÐT^�oÑ_Ô_×iÒiÐjkÐmnÑoÔoˆØ! D¤K°°a°a°a¸°Ô$>Ñ>ˆØ(Ô.¨qÔ1ˆ
Ø! A‘~¨¬Ñ8¸4¼:ÑEˆÝ”ØØ˜KÐ(Ø˜DœJ˜Ø�t”Ð'ð	
ñ 
ô 
ð
 ˆ!ˆ!ˆQ��4”< 4¤< -Ð/Ð
/ô1ˆõ œ&ØŒK×ÒÑ Ô ×'Ò'¨¨:°rÑ:Ô:×DÒDÀQÈÑJÔJØ˜KÐ(Ø˜DœJ˜Ø�t”Ð'ð	
ñ 
ô 
÷
 Š'‰)Œ)�D”L D¤L =Ð0ô2ˆð *×/Ò/°EÐ/Ñ:Ô:ˆØ˜Ñ'ˆØ�Š˜qÑ!Ô!Ð!r.   ©
r$   r%   r&   r'   r   r;   r(   r,   rL   rN   rO   s   @r/   r;  r;  +  st   ø€ € € € € ðð ð:˜~ð :ð :ð :ð :ð :ð :ð"" U¤\ð ""°e´lð ""ð ""ð ""ð ""ð ""ð ""ð ""ð ""r.   r;  c                   óf   ‡ — e Zd ZU eed<   dZdZdZdZdgZ	dZ
dZdZdZdZdZdZeedœZˆ fd	„Zˆ xZS )
ÚNeuCodecPreTrainedModelr6   Úneucodec)rW  TNr™   Úinput_values)rG   Ú
attentionsc                 óÒ  •— t          ¦   «                              |¦  «         t          |t          ¦  «        r4t	          j        |j        ¦  «         t	          j        |j        ¦  «         d S t          |t          ¦  «        r5t          j
        |j        ¦  «        }t	          j        |j        |¦  «         d S t          |t          ¦  «        rt|                     |j        j        ¬¦  «        \  }}}t	          j        |j        |¦  «         t	          j        |j        |¦  «         t	          j        |j        |¦  «         d S t          |t(          ¦  «        rBt+          d|j        z  d|j        z  |j        ¦  «        }t	          j        |j        |¦  «         d S t          |t2          ¦  «        r<t+          |j        |j        |j        ¦  «        }t	          j        |j        |¦  «         d S d S )Nr  rØ   rî   )r:   Ú_init_weightsr2  rÉ   ÚinitÚzeros_rÎ   rÏ   r;  r(   rA  r=  Úcopy_rB  r  r  r  r  r  r  r  rê   rø   râ   r÷   rì   rà   rá   )rE   ro   rB  r  r  r  Úfilter_tensorrF   s          €r/   r`  z%NeuCodecPreTrainedModel._init_weightsu  s¯  ø€ Ý‰Œ×Ò˜fÑ%Ô%Ð%Ý�fÕ/Ñ0Ô0ð 	5ÝŒK˜œÑ%Ô%Ð%ÝŒK˜œÑ$Ô$Ð$Ð$Ð$Ý˜Õ 1Ñ2Ô2ð 	5ÝÔ& v¤|Ñ4Ô4ˆFÝŒJ�v”} fÑ-Ô-Ð-Ð-Ð-Ý˜Õ @ÑAÔAð 
	5Ø&,×&=Ò&=ÀVÄ]ÔEYÐ&=Ñ&ZÔ&ZÑ#ˆF�E˜8ÝŒJ�v”} fÑ-Ô-Ð-ÝŒJ�v”| UÑ+Ô+Ð+ÝŒJ�v”¨Ñ1Ô1Ð1Ð1Ð1Ý˜Õ 2Ñ3Ô3ð 	5Ý0°°v´|Ñ1CÀSÈ6Ì<ÑEWÐY_ÔYkÑlÔlˆMÝŒJ�v”} mÑ4Ô4Ð4Ð4Ð4Ý˜Õ 4Ñ5Ô5ð 	5Ý0°´ÀÔ@QÐSYÔSeÑfÔfˆMÝŒJ�v”} mÑ4Ô4Ð4Ð4Ð4ð	5ð 	5r.   )r$   r%   r&   r   r*   Úbase_model_prefixÚinput_modalitiesÚsupports_gradient_checkpointingÚ_no_split_modulesÚ_skip_keys_device_placementÚ_supports_flash_attnÚ_supports_sdpaÚ_supports_flex_attnÚ_supports_cache_classÚ_supports_attention_backendÚ_can_compile_fullgraphÚmain_input_namerº   Ú_can_record_outputsr`  rN   rO   s   @r/   r[  r[  a  s¢   ø€ € € € € € àÐÐÑØ"ÐØ!ÐØ&*Ð#ØÐØ#4Ð"5ÐØÐØ€NØÐØ ÐØ"&ÐØ!ÐØ$€Oà-Ø*ðð Ðð
5ð 5ð 5ð 5ð 5ð 5ð 5ð 5ð 5r.   r[  c            	       óî   ‡ — e Zd Z edd¬¦  «        d
defˆ fd„¦   «         Ze edd¬¦  «        d
dedeej	        e
f         fd„¦   «         ¦   «         Z ej        ¦   «         ed	„ ¦   «         ¦   «         Zˆ xZS )ÚNeuCodecRotaryEmbeddingr  z5.18)ÚversionNr6   c                 ó¾  •— t          ¦   «                              ¦   «          |j        | _        |j        | _        || _        | j        j        d         | _        | j        }| j        dk    rt          | j                 } || j        |¦  «        \  }| _
        t          j        |d¬¦  «        | _        t          j        |                     ¦   «         d¬¦  «        | _        d S )NÚ	rope_typeÚdefaultFrï   )r:   r;   Úmax_position_embeddingsÚmax_seq_len_cachedÚoriginal_max_seq_lenr6   Úrope_parametersrv  Úcompute_default_rope_parametersr   Úattention_scalingr>   rö   Úinv_freqÚcloneÚoriginal_inv_freq)rE   r6   r  Úrope_init_fnr~  rF   s        €r/   r;   z NeuCodecRotaryEmbedding.__init__‹  sÀ   ø€ å‰Œ×ÒÑÔÐØ"(Ô"@ˆÔØ$*Ô$BˆÔ!àˆŒàœÔ4°[ÔAˆŒØ!%Ô!EˆØŒ>˜YÒ&Ð&Ý.¨t¬~Ô>ˆLØ+7¨<¸¼ÀVÑ+LÔ+LÑ(ˆ�$Ô(åœ	 (°uÐ=Ñ=Ô=ˆŒÝ!#¤¨8¯>ª>Ñ+;Ô+;ÈÐ!NÑ!NÔ!NˆÔÐÐr.   rH   c                 óä   — | j         d         }t          | dd¦  «        p| j        | j        z  }d}d|t	          j        d|dt          j        ¬¦  «        |z  z  z  }|                     |¦  «        |fS )a¾  
        Computes the inverse frequencies according to the original RoPE implementation
        Args:
            config ([`~transformers.PreTrainedConfig`]):
                The model configuration.
        Returns:
            Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the
            post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE).
        Ú
rope_thetarl   NrÊ   r   r   rÙ   )r{  rŽ   r@   r�   r(   rÝ   r¸   rƒ   )r6   r  rv   ÚbaserS   Úattention_factorr~  s          r/   r|  z7NeuCodecRotaryEmbedding.compute_default_rope_parametersœ  sz   € ð Ô% lÔ3ˆÝ�f˜j¨$Ñ/Ô/Ðc°6Ô3EÈÔIcÑ3cˆàÐà˜$¥5¤<°°3¸ÅÄÐ#MÑ#MÔ#MÐPSÑ#SÑTÑUˆØ�{Š{˜6Ñ"Ô"Ð$4Ð4Ð4r.   c                 óú  — | j         d d d …d f                              |j        d         dd¦  «                             t          j        |j        ¬¦  «        }|d d …d d d …f                              ¦   «         }t          |j        j        t          ¦  «        r|j        j        dk    r|j        j        nd}t          |d¬¦  «        5  ||z                       dd	¦  «        }t	          j        ||fd¬
¦  «        }|                     ¦   «         | j        z  }|                     ¦   «         | j        z  }	d d d ¦  «         n# 1 swxY w Y   |                     |j        ¬¦  «        |	                     |j        ¬¦  «        fS )Nr   rQ   r   r  r-  r.  Fr/  r   rR   rÙ   )r~  rg   rT   rƒ   r(   r¸   r  r2  r3  r4  r   r   rU   r_   r}  r`   ry   )
rE   rV   rÂ   Úinv_freq_expandedÚposition_ids_expandedr0  ÚfreqsÚembr_   r`   s
             r/   rL   zNeuCodecRotaryEmbedding.forward°  s›  € ð ŒM˜$    4˜-Ô(×/Ò/°Ô0BÀ1Ô0EÀrÈ1ÑMÔM×PÒPÕW\ÔWbÐklÔksÐPÑtÔtð 	ð !-¨Q¨Q¨Q°°a°a°a¨ZÔ 8× >Ò >Ñ @Ô @Ðå'1°!´(´-ÅÑ'EÔ'EÐkÈ!Ì(Ì-Ð[`ÒJ`ÐJ`�a”h”m�mÐfkˆå¨¸UÐCÑCÔCð 	5ð 	5Ø&Ð)>Ñ>×IÒIÈ!ÈQÑOÔOˆEÝ”)˜U E˜N°Ð3Ñ3Ô3ˆCØ—'’'‘)”)˜dÔ4Ñ4ˆCØ—'’'‘)”)˜dÔ4Ñ4ˆCð		5ð 	5ð 	5ñ 	5ô 	5ð 	5ð 	5ð 	5ð 	5ð 	5ð 	5øøøð 	5ð 	5ð 	5ð 	5ð �vŠv˜AœGˆvÑ$Ô$ c§f¢f°1´7 fÑ&;Ô&;Ð;Ð;s   ÃA*D<Ä<E ÅE rJ   )r$   r%   r&   r   r   r;   Ústaticmethodr¥   r(   r,   r¸   r|  Úno_gradr   rL   rN   rO   s   @r/   rs  rs  Š  sö   ø€ € € € € Ø€_�X vÐ.Ñ.Ô.ðOð O˜~ð Oð Oð Oð Oð Oñ /Ô.ðOð  Ø€_�X vÐ.Ñ.Ô.ð5ð 5°ð 5ÐZ_Ð`eÔ`lÐnsÐ`sÔZtð 5ð 5ð 5ñ /Ô.ñ „\ð5ð$ €U„]�_„_Øð<ð <ñ Ôñ „_ð<ð <ð <ð <ð <r.   rs  c            	       ó@   ‡ — e Zd Z	 	 	 	 d	dedededefˆ fd„Zd„ Zˆ xZS )
ÚNeuCodecAntiAliasedActivation1dr   é   Úup_ratioÚ
down_ratioÚup_kernel_sizeÚdown_kernel_sizec                 óæ   •— t          ¦   «                              ¦   «          t          |¦  «        st          d¦  «        ‚|| _        t          ||¦  «        | _        t          ||¦  «        | _        d S )Nz$Activation function must be callable)	r:   r;   ÚcallableÚ	TypeErrorÚactr  Úupsamplerì   Ú
downsample)rE   Ú
activationr�  r‘  r’  r“  rF   s         €r/   r;   z(NeuCodecAntiAliasedActivation1d.__init__Ä  si   ø€ õ 	‰Œ×ÒÑÔÐÝ˜
Ñ#Ô#ð 	DÝÐBÑCÔCÐCØˆŒÝ*¨8°^ÑDÔDˆŒÝ.¨zÐ;KÑLÔLˆŒˆˆr.   c                 ó„   — |                       |¦  «        }|                      |¦  «        }|                      |¦  «        }|S rJ   )r˜  r—  r™  rK   s     r/   rL   z'NeuCodecAntiAliasedActivation1d.forwardÓ  s;   € ØŸš mÑ4Ô4ˆØŸš Ñ/Ô/ˆØŸš¨Ñ6Ô6ˆàÐr.   )r   r   r�  r�  )r$   r%   r&   r¤   r;   rL   rN   rO   s   @r/   rŽ  rŽ  Ã  s’   ø€ € € € € ð ØØ Ø "ðMð Mð ðMð ð	Mð
 ðMð ðMð Mð Mð Mð Mð Mðð ð ð ð ð ð r.   rŽ  c                   ó(   ‡ — e Zd ZdZˆ fd„Zd„ Zˆ xZS )ÚNeuCodecResidualUnitza
    A residual unit composed of Snake1d and weight-normalized Conv1d layers with dilations.
    c                 óT  •— t          ¦   «                              ¦   «          d|z  dz  }t          t          |¦  «        ¬¦  «        | _        t          j        ||d||¬¦  «        | _        t          t          |¦  «        ¬¦  «        | _        t          j        ||d¬¦  «        | _	        d S )Nr	  r   ©rš  é   )râ   Údilationr@  r   )râ   )
r:   r;   rŽ  rÉ   Úsnake1r>   ÚConv1dÚconv1Úsnake2Úconv2)rE   Ú	dimensionr¡  r   rF   s       €r/   r;   zNeuCodecResidualUnit.__init__à  s™   ø€ Ý‰Œ×ÒÑÔÐØ˜Ñ! aÑ'ˆÝ5ÕARÐS\ÑA]ÔA]Ð^Ñ^Ô^ˆŒÝ”Y˜y¨)ÀÈXÐ_bÐcÑcÔcˆŒ
Ý5ÕARÐS\ÑA]ÔA]Ð^Ñ^Ô^ˆŒÝ”Y˜y¨)ÀÐCÑCÔCˆŒ
ˆ
ˆ
r.   c                 ó  — |}|                       |                      |¦  «        ¦  «        }|                      |                      |¦  «        ¦  «        }|j        d         |j        d         z
  dz  }|dk    r|d|| …f         }||z   }|S )ar  
        Forward pass through the residual unit.

        Args:
            hidden_state (`torch.Tensor` of shape `(batch_size, channels, time_steps)`):
                Input tensor .

        Returns:
            output_tensor (`torch.Tensor` of shape `(batch_size, channels, time_steps)`):
                Input tensor after passing through the residual unit.
        rQ   r   r   .)r¤  r¢  r¦  r¥  rT   )rE   Úhidden_stateÚoutput_tensorr@  s       r/   rL   zNeuCodecResidualUnit.forwardè  s‘   € ð %ˆØŸ
š
 4§;¢;¨}Ñ#=Ô#=Ñ>Ô>ˆØŸ
š
 4§;¢;¨}Ñ#=Ô#=Ñ>Ô>ˆàÔ% bÔ)¨MÔ,?ÀÔ,CÑCÈÑIˆØ�QŠ;ˆ;Ø'¨¨W°g°XÐ-=Ð(=Ô>ˆLØ$ }Ñ4ˆØÐr.   rÓ   rO   s   @r/   r�  r�  Û  sV   ø€ € € € € ðð ðDð Dð Dð Dð Dðð ð ð ð ð ð r.   r�  c                   ó8   ‡ — e Zd ZdZddededefˆ fd„Zd„ Zˆ xZS )	ÚNeuCodecEncoderBlockz'Encoder block used in NEUCODEC encoder.r   r6   rõ   Ústride_indexc           
      ó´  •— t          ¦   «                              ¦   «          |j        d|z  z  }t          |dz  d¬¦  «        | _        t          |dz  d¬¦  «        | _        t          |dz  d¬¦  «        | _        t          t          |dz  ¦  «        ¬¦  «        | _	        t          j        |dz  |d|z  |t          j        |dz  ¦  «        ¬¦  «        | _        d S )Nr   r   )r¡  r   é	   rŸ  ©râ   rõ   r@  )r:   r;   Úencoder_hidden_sizer�  Ú	res_unit1Ú	res_unit2Ú	res_unit3rŽ  rÉ   r¢  r>   r£  rÚ   Úceilr¤  )rE   r6   rõ   r­  r§  rF   s        €r/   r;   zNeuCodecEncoderBlock.__init__  sÖ   ø€ Ý‰Œ×ÒÑÔÐØÔ.°°L±Ñ@ˆ	Ý-¨i¸1©nÀqÐIÑIÔIˆŒÝ-¨i¸1©nÀqÐIÑIÔIˆŒÝ-¨i¸1©nÀqÐIÑIÔIˆŒÝ5ÕARÐS\Ð`aÑSaÑAbÔAbÐcÑcÔcˆŒÝ”YØ˜‰N˜I°1°v±:ÀfÕVZÔV_Ð`fÐijÑ`jÑVkÔVkð
ñ 
ô 
ˆŒ
ˆ
ˆ
r.   c                 óÔ   — |                       |¦  «        }|                      |¦  «        }|                      |                      |¦  «        ¦  «        }|                      |¦  «        }|S rJ   )r²  r³  r¢  r´  r¤  )rE   r©  s     r/   rL   zNeuCodecEncoderBlock.forward  sX   € Ø—~’~ lÑ3Ô3ˆØ—~’~ lÑ3Ô3ˆØ—{’{ 4§>¢>°,Ñ#?Ô#?Ñ@Ô@ˆØ—z’z ,Ñ/Ô/ˆàÐr.   )r   r   )	r$   r%   r&   r'   r   r¤   r;   rL   rN   rO   s   @r/   r¬  r¬  ÿ  sl   ø€ € € € € Ø1Ð1ð	
ð 	
˜~ð 	
°sð 	
Ècð 	
ð 	
ð 	
ð 	
ð 	
ð 	
ðð ð ð ð ð ð r.   r¬  c                   ó.   ‡ — e Zd ZdZdefˆ fd„Zd„ Zˆ xZS )ÚNeuCodecEncoderzNEUCODEC Encoderr6   c                 ó  •— t          ¦   «                              ¦   «          t          j        d|j        dd¬¦  «        | _        g | _        t          |j        ¦  «        D ]+\  }}|dz   }| xj        t          |||¬¦  «        gz  c_        Œ,t          j
        | j        ¦  «        | _        |j        dt          |j        ¦  «        z  z  }t          t          |¦  «        ¬¦  «        | _        t          j        ||j        dd¬¦  «        | _        d S )Nr   r   r   ©râ   r@  )rõ   r­  r   rŸ  )r:   r;   r>   r£  r±  r¤  ÚblockÚ	enumerateÚdownsampling_ratiosr¬  Ú
ModuleListÚlenrŽ  rÉ   r¢  r@   r¦  )rE   r6   r­  rõ   Úd_modelrF   s        €r/   r;   zNeuCodecEncoder.__init__  sÿ   ø€ Ý‰Œ×ÒÑÔÐõ ”Y˜q &Ô"<È!ÐUVÐWÑWÔWˆŒ
àˆŒ
å$-¨fÔ.HÑ$IÔ$Ið 	cð 	cÑ ˆL˜&Ø'¨!Ñ+ˆLØˆJŒJÕ/°¸vÐT`ÐaÑaÔaÐbÑbˆJŒJˆJå”] 4¤:Ñ.Ô.ˆŒ
ØÔ,¨qµC¸Ô8RÑ4SÔ4SÑ/SÑSˆÝ5ÕARÐSZÑA[ÔA[Ð\Ñ\Ô\ˆŒÝ”Y˜w¨Ô(:ÈÐSTÐUÑUÔUˆŒ
ˆ
ˆ
r.   c                 ó®   — |                       |¦  «        }| j        D ]} ||¦  «        }Œ|                      |¦  «        }|                      |¦  «        }|S rJ   )r¤  r»  r¢  r¦  )rE   r©  ro   s      r/   rL   zNeuCodecEncoder.forward*  s]   € Ø—z’z ,Ñ/Ô/ˆà”jð 	0ð 	0ˆFØ!˜6 ,Ñ/Ô/ˆLˆLà—{’{ <Ñ0Ô0ˆØ—z’z ,Ñ/Ô/ˆàÐr.   )r$   r%   r&   r'   r   r;   rL   rN   rO   s   @r/   r¸  r¸    s`   ø€ € € € € ØÐðV˜~ð Vð Vð Vð Vð Vð Vð"	ð 	ð 	ð 	ð 	ð 	ð 	r.   r¸  c                   óH   ‡ — e Zd Zdefˆ fd„Zdej        dej        fd„Zˆ xZS )ÚNeuCodecResNetBlockr6   c                 óè  •— t          ¦   «                              ¦   «          t          j        d|j        dd¬¦  «        | _        t          j        ¦   «         | _        t          j        |j        |j        ddd¬¦  «        | _	        t          j        d|j        dd¬¦  «        | _
        t          j        ¦   «         | _        |j        | _        t          j        |j        |j        ddd¬¦  «        | _        d S )Né    r©   T)Ú
num_groupsÚnum_channelsrª   Úaffiner   r   r°  )r:   r;   r>   Ú	GroupNormr@   Únorm1ÚSiLUÚactivation1r£  r¤  Únorm2Úactivation2Úactivation_dropoutr¦  rD   s     €r/   r;   zNeuCodecResNetBlock.__init__7  sÌ   ø€ Ý‰Œ×ÒÑÔÐÝ”\¨R¸fÔ>PÐVZÐcgÐhÑhÔhˆŒ
Ýœ7™9œ9ˆÔÝ”Y˜vÔ1°6Ô3EÐSTÐ]^ÐhiÐjÑjÔjˆŒ
Ý”\¨R¸fÔ>PÐVZÐcgÐhÑhÔhˆŒ
Ýœ7™9œ9ˆÔØ"(Ô";ˆÔÝ”Y˜vÔ1°6Ô3EÐSTÐ]^ÐhiÐjÑjÔjˆŒ
ˆ
ˆ
r.   rG   rH   c                 ó¸  — |                      dd¦  «        }|}|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }t          j                             || j	        | j
        ¬¦  «        }|                      |¦  «        }||z                         dd¦  «        S )Nr   r   rz   )r   rÊ  rÌ  r¤  rÍ  rÎ  r>   r€   ru   rÏ  r|   r¦  ©rE   rG   rÅ   s      r/   rL   zNeuCodecResNetBlock.forwardA  sÄ   € Ø%×/Ò/°°1Ñ5Ô5ˆØ ˆØŸ
š
 =Ñ1Ô1ˆØ×(Ò(¨Ñ7Ô7ˆØŸ
š
 =Ñ1Ô1ˆØŸ
š
 =Ñ1Ô1ˆØ×(Ò(¨Ñ7Ô7ˆÝœ×-Ò-¨m¸tÔ?VÐaeÔanÐ-ÑoÔoˆØŸ
š
 =Ñ1Ô1ˆØ Ñ(×3Ò3°A°qÑ9Ô9Ð9r.   rM   rO   s   @r/   rÃ  rÃ  6  sq   ø€ € € € € ðk˜~ð kð kð kð kð kð kð
: U¤\ð 
:°e´lð 
:ð 
:ð 
:ð 
:ð 
:ð 
:ð 
:ð 
:r.   rÃ  c                   ó†   ‡ — e Zd Zdefˆ fd„Zdej        dej        fd„Zdej        deej        ej        f         fd„Z	ˆ xZ
S )ÚNeuCodecQuantizerr6   c                 ó4  •— t          ¦   «                              ¦   «          t          |¦  «        | _        t	          j        |j        t          |j        ¦  «        ¦  «        | _	        t	          j        t          |j        ¦  «        |j        ¦  «        | _
        d S rJ   )r:   r;   r  Ú	quantizerr>   r?   Úquantization_dimr¿  r  Ú
project_inÚproject_outrD   s     €r/   r;   zNeuCodecQuantizer.__init__O  sq   ø€ Ý‰Œ×ÒÑÔÐÝ9¸&ÑAÔAˆŒÝœ) FÔ$;½SÀÔA[Ñ=\Ô=\Ñ]Ô]ˆŒÝœ9¥S¨Ô)CÑ%DÔ%DÀfÔF]Ñ^Ô^ˆÔÐÐr.   r  rH   c                 óz   — |                      d¦  «        }| j        j        |         }|                      |¦  «        S ©NrQ   )rP  rÕ  r  rØ  )rE   r  r!  s      r/   Ú
from_codeszNeuCodecQuantizer.from_codesU  s6   € Ø—/’/ "Ñ%Ô%ˆØ”Ô'¨Ô0ˆØ×Ò Ñ&Ô&Ð&r.   rG   c                 ó   — |                       |¦  «        }|j        }| j                             |¦  «        }|                      |¦  «        \  }}|                      |                     |¦  «        ¦  «        }|                     d¦  «        }||fS rÚ  )r×  ry   rÕ  r+  rØ  rƒ   r\   )rE   rG   r7  Úquantized_outr  s        r/   rL   zNeuCodecQuantizer.forwardZ  s…   € ØŸš¨Ñ6Ô6ˆØ&Ô,ˆØœ×,Ò,¨]Ñ;Ô;ˆØ!%§¢°Ñ!>Ô!>Ñˆ�wØ×(Ò(¨×)9Ò)9¸.Ñ)IÔ)IÑJÔJˆØ×#Ò# BÑ'Ô'ˆØ˜gÐ%Ð%r.   )r$   r%   r&   r   r;   r(   r,   rÛ  r¥   rL   rN   rO   s   @r/   rÓ  rÓ  N  s£   ø€ € € € € ð_˜~ð _ð _ð _ð _ð _ð _ð' %¤,ð '°5´<ð 'ð 'ð 'ð 'ð
& U¤\ð &°e¸E¼LÈ%Ì,Ð<VÔ6Wð &ð &ð &ð &ð &ð &ð &ð &r.   rÓ  c                   óL   ‡ — e Zd ZdZdefˆ fd„Zdej        dej        fd„Zˆ xZ	S )ÚNeuCodecDecoderzVVocos-based decoder with ResNet, Transformer, and ISTFT head for audio reconstruction.r6   c                 óæ  •‡— t          ¦   «                              ¦   «          t          j        ‰j        ‰j        j        z   ‰j        ¦  «        | _        t          j        ‰j        ‰j        dd¬¦  «        | _        t          j	        t          ‰¦  «        t          ‰¦  «        g¦  «        | _        ‰j        | _        t          ‰¬¦  «        | _        t          j	        ˆfd„t          ‰j        ¦  «        D ¦   «         ¦  «        | _        t          j	        t          ‰¦  «        t          ‰¦  «        g¦  «        | _        t          j        ‰j        d¬¦  «        | _        t+          ‰¦  «        | _        d S )Nr   r   rº  )r6   c                 ó0   •— g | ]}t          ‰|¦  «        ‘ŒS r-   )rº   )Ú.0rŒ   r6   s     €r/   ú
<listcomp>z,NeuCodecDecoder.__init__.<locals>.<listcomp>o  s$   ø€ ÐfÐfÐf¸Õ! &¨)Ñ4Ô4ÐfÐfÐfr.   r©   r¼   )r:   r;   r>   r?   r@   Úsemantic_model_configÚfcr£  Úembedr¾  rÃ  Ú	prior_netr�   rs  Ú
rotary_embÚrangeÚnum_hidden_layersÚlayersÚpost_netÚ	LayerNormrF  r;  ÚheadrD   s    `€r/   r;   zNeuCodecDecoder.__init__g  s3  øø€ Ý‰Œ×ÒÑÔÐÝ”)˜FÔ.°Ô1MÔ1YÑYÐ[aÔ[mÑnÔnˆŒÝ”Y˜vÔ1°6Ô3EÐSTÐ^_Ð`Ñ`Ô`ˆŒ
ÝœÕ(;¸FÑ(CÔ(CÕEXÐY_ÑE`ÔE`Ð'aÑbÔbˆŒØ#)Ô#=ˆÔ Ý1¸Ð@Ñ@Ô@ˆŒÝ”mØfÐfÐfÐfÅeÈFÔLdÑFeÔFeÐfÑfÔfñ
ô 
ˆŒõ œÕ':¸6Ñ'BÔ'BÕDWÐX^ÑD_ÔD_Ð&`ÑaÔaˆŒÝ”L Ô!3¸Ð>Ñ>Ô>ˆŒ	Ý% fÑ-Ô-ˆŒ	ˆ	ˆ	r.   rG   rH   c                 ó  — |                       |¦  «        }|                     dd¦  «        }|                      |¦  «        }|                     dd¦  «        }| j        D ]} ||¦  «        }Œt	          j        | j        |j        ¬¦  «                             d¦  «        }|  	                    ||¦  «        }| j
        D ]} ||fd|i|¤Ž}Œ| j        D ]} ||¦  «        }Œ|                      |                      |¦  «        ¦  «        S )Nr   r   r  r   r˜   )rå  r   ræ  rç  r(   rÝ   r�   r  r\   rè  rë  rì  rî  rF  )rE   rG   rv   ÚlayerrÂ   r˜   s         r/   rL   zNeuCodecDecoder.forwardu  s&  € ØŸš Ñ.Ô.ˆØ%×/Ò/°°1Ñ5Ô5ˆØŸ
š
 =Ñ1Ô1ˆØ%×/Ò/°°1Ñ5Ô5ˆð ”^ð 	1ð 	1ˆEØ!˜E -Ñ0Ô0ˆMˆMõ ”| DÔ$<À]ÔEYÐZÑZÔZ×dÒdÐefÑgÔgˆØ"Ÿošo¨m¸\ÑJÔJÐØ”[ð 	dð 	dˆEØ!˜E -ÐcÐcÐEXÐcÐ\bÐcÐcˆMˆMð ”]ð 	1ð 	1ˆEØ!˜E -Ñ0Ô0ˆMˆMà�yŠy˜Ÿš =Ñ1Ô1Ñ2Ô2Ð2r.   rY  rO   s   @r/   rß  rß  d  sp   ø€ € € € € Ø`Ð`ð.˜~ð .ð .ð .ð .ð .ð .ð3 U¤\ð 3ÀÄð 3ð 3ð 3ð 3ð 3ð 3ð 3ð 3r.   rß  c                   óH   ‡ — e Zd Zdefˆ fd„Zdej        dej        fd„Zˆ xZS )ÚNeuCodecSemanticAdapterr6   c                 ó8  •— t          ¦   «                              ¦   «          t          j        |j        j        |j        j        ddd¬¦  «        | _        t          j        ¦   «         | _        t          j        |j        j        |j        j        ddd¬¦  «        | _	        t          j        ¦   «         | _
        t          j        |j        j        |j        j        ddd¬¦  «        | _        t          j        |j        j        |j        j        ddd¬¦  «        | _        d S )Nr   r   F)Úin_channelsÚout_channelsrâ   r@  r9   T)râ   r@  r9   )r:   r;   r>   r£  rä  r@   r¤  ÚReLUÚact1r¦  Úact2Úconv3Úconv4rD   s     €r/   r;   z NeuCodecSemanticAdapter.__init__‘  s	  ø€ Ý‰Œ×ÒÑÔÐÝ”YØÔ4Ô@ØÔ5ÔAØØØð
ñ 
ô 
ˆŒ
õ ”G‘I”IˆŒ	Ý”YØÔ(Ô4ØÔ(Ô4ØØØð
ñ 
ô 
ˆŒ
õ ”G‘I”IˆŒ	Ý”YØÔ(Ô4ØÔ(Ô4ØØØð
ñ 
ô 
ˆŒ
õ ”YØÔ4Ô@ØÔ5ÔAØØØð
ñ 
ô 
ˆŒ
ˆ
ˆ
r.   rG   rH   c                 ó  — |                       |¦  «        }|                      |¦  «        }|}|                      |¦  «        }|                      |¦  «        }|                      |¦  «        }||z   }|                      |¦  «        }|S rJ   )r¤  r÷  r¦  rø  rù  rú  rÑ  s      r/   rL   zNeuCodecSemanticAdapter.forward²  s}   € ØŸ
š
 =Ñ1Ô1ˆØŸ	š	 -Ñ0Ô0ˆØ ˆØŸ
š
 =Ñ1Ô1ˆØŸ	š	 -Ñ0Ô0ˆØŸ
š
 =Ñ1Ô1ˆØ%¨Ñ0ˆØŸ
š
 =Ñ1Ô1ˆØÐr.   rM   rO   s   @r/   rò  rò  �  sk   ø€ € € € € ð
˜~ð 
ð 
ð 
ð 
ð 
ð 
ðB	 U¤\ð 	°e´lð 	ð 	ð 	ð 	ð 	ð 	ð 	ð 	r.   rò  z"NeuCodec neural audio codec model.)Úcustom_introc                   óÒ  ‡ — e Zd ZeZdefˆ fd„Zee	 	 	 ddej	        dej	        dej	        dz  dej	        dz  d	e
d
ee         deez  fd„¦   «         ¦   «         Zee	 	 	 ddej	        dz  dej	        dz  dej	        dz  d
ee         deez  f
d„¦   «         ¦   «         Zee	 	 	 ddej	        dej	        dej	        dz  dej	        dz  d	e
d
ee         deez  fd„¦   «         ¦   «         Zˆ xZS )ÚNeuCodecModelr6   c                 ó
  •— t          ¦   «                              |¦  «         |j        | _        t	          j        |j        ¦  «        | _        t          |¦  «        | _	        t          |¦  «        | _        t          j        |j        |j        j        z   |j        |j        j        z   ¦  «        | _        t!          |¦  «        | _        t%          |¦  «        | _        |j        |j        z  | _        |                      ¦   «          d S rJ   )r:   r;   Úencoder_hop_lengthr?  r   Úfrom_configrä  Úsemantic_encoderrò  Úsemantic_adapterr¸  Úacoustic_encoderr>   r?   r@   Ú
fc_encoderrÓ  rÕ  rß  Úacoustic_decoderÚoutput_sampling_rateÚinput_sampling_rateÚsample_rate_conversion_factorÚ	post_initrD   s     €r/   r;   zNeuCodecModel.__init__Â  sÜ   ø€ Ý‰Œ×Ò˜Ñ Ô Ð ð !Ô3ˆŒÝ )Ô 5°fÔ6RÑ SÔ SˆÔÝ 7¸Ñ ?Ô ?ˆÔÝ /°Ñ 7Ô 7ˆÔÝœ)ØÔ Ô!=Ô!IÑIØÔ Ô!=Ô!IÑIñ
ô 
ˆŒõ +¨6Ñ2Ô2ˆŒÝ /°Ñ 7Ô 7ˆÔØ-3Ô-HÈ6ÔKeÑ-eˆÔ*à�ŠÑÔÐÐÐr.   NFr]  Úinput_featuresÚpadding_maskÚinput_features_maskÚoutput_latentsrv   rH   c                 óÐ  — t          j        ¦   «         5  |                      ||¬¦  «        }ddd¦  «         n# 1 swxY w Y   |j                             dd¦  «        }|                      |¦  «        }|                      |¦  «        }	t          |	j        d         |j        d         ¦  «        }
|	dd|
…f         }	|dd|
…f         }t          j	        ||	gd¬¦  «        }|  
                    |                     dd¦  «        ¦  «        }|                      |¦  «        \  }}|                     dd¦  «        }|                     dd¦  «        }d}|�ƒ|                     dd¬	¦  «        }|| j        z  }|| j        z  }t          j        |j        d         |j        ¬
¦  «                             dd¦  «        }||k                          |j        ¦  «        }t)          ||r|nd|¬¦  «        S )a  
        input_values (`torch.Tensor` of shape `(batch_size, 1, sequence_length)`):
            Input audio waveform.
        input_features (`torch.Tensor` of shape `(batch_size, mel_bins, time_steps)`):
            Input audio mel spectrogram for semantic encoding.
        padding_mask (`torch.Tensor` of shape `(batch_size, 1, sequence_length)`):
            Padding mask used to pad `input_values`.
        input_features_mask (`torch.Tensor` of shape `(batch_size, time_steps)`, *optional*):
            Attention mask for the spectrogram input to the semantic encoder. `1` for valid frames, `0` for padding.
        output_latents (`bool`, *optional*, defaults to `False`):
            Whether to return the continuous latent representation from the quantizer.
        )rs   Nr   r   rQ   .rR   T)rS   r°   r  )r!   r"   r#   )r(   rŒ  r  Úlast_hidden_stater   r  r  rH  rT   rU   r  rÕ  rß   r	  r?  rÝ   r  r›   rƒ   ry   r1   )rE   r]  r  r  r  r  rv   Úsemantic_outputÚsemantic_hidden_statesÚacoustic_hidden_statesÚ
min_lengthrG   r"   r!   r#   Úaudio_lengthÚtoken_lengthÚidxs                     r/   ÚencodezNeuCodecModel.encodeÕ  sT  € õ2 Œ]‰_Œ_ð 	hð 	hØ"×3Ò3°NÐSfÐ3ÑgÔgˆOð	hð 	hð 	hñ 	hô 	hð 	hð 	hð 	hð 	hð 	hð 	høøøð 	hð 	hð 	hð 	hà!0Ô!B×!LÒ!LÈQÐPQÑ!RÔ!RÐØ!%×!6Ò!6Ð7MÑ!NÔ!NÐð "&×!6Ò!6°|Ñ!DÔ!DÐõ Ð/Ô5°bÔ9Ð;QÔ;WÐXZÔ;[Ñ\Ô\ˆ
Ø!7¸¸[¸j¸[Ð8HÔ!IÐØ!7¸¸[¸j¸[Ð8HÔ!IÐåœ	Ð#9Ð;QÐ"RÐXYÐZÑZÔZˆØŸš¨×(?Ò(?ÀÀ1Ñ(EÔ(EÑFÔFˆð  $Ÿ~š~¨mÑ<Ô<Ñˆ�Ø×#Ò# A qÑ)Ô)ˆØ!×+Ò+¨A¨qÑ1Ô1ˆð  ÐØÐ#Ø'×+Ò+°¸DÐ+ÑAÔAˆLØ'¨$Ô*LÑLˆLØ'¨4¬?Ñ:ˆLÝ”,˜{Ô0°Ô4¸\Ô=PÐQÑQÔQ×VÒVÐWXÐZ\Ñ]Ô]ˆCØ # lÒ 2×6Ò6°|Ô7IÑJÔJÐå$Ø#Ø-Ð7�G�G°4Ø-ð
ñ 
ô 
ð 	
s   ”8¸<¿<r!   r"   r#   c                 ó$  — |€|€t          d¦  «        ‚|�/| j                             |                     dd¦  «        ¦  «        }n|                     dd¦  «        }t	          | j        ||¬¦  «        } | j        |fd|i|¤Ž}t          |¬¦  «        S )a1  
        audio_codes (`torch.LongTensor`  of shape `(batch_size, 1, codes_length)`):
            Discrete code indices computed using `model.encode`.
        latents (torch.Tensor of shape `(batch_size, dimension, time_steps)`, *optional*):
            Quantized continuous representation of input.
        audio_codes_mask (`torch.Tensor` of shape `(batch_size, codes_length)`, *optional*):
            Mask marking valid (non-padding) positions, as returned by `model.encode`. Needed for correct
            batched decoding of variable-length audio.
        Nz3Either `latents` or `audio_codes` must be provided.r   r   )r6   Úinputs_embedsrs   rs   )r    )rñ   rÕ  rÛ  r   r   r6   r  r3   )rE   r!   r"   r#   rv   rs   Úrecon_audios          r/   ÚdecodezNeuCodecModel.decode  s³   € ð$ ˆ?˜{Ð2ÝÐRÑSÔSÐSàÐ"Ø”n×/Ò/°×0EÒ0EÀaÈÑ0KÔ0KÑLÔLˆGˆGà×'Ò'¨¨1Ñ-Ô-ˆGõ 3Ø”;Ø!Ø+ð
ñ 
ô 
ˆð ,�dÔ+¨GÐ]Ð]ÀNÐ]ÐV\Ð]Ð]ˆÝ$°+Ð>Ñ>Ô>Ð>r.   c                 ó"  — |j         d         }t          || j        z  ¦  «        }|                      ||||dd¬¦  «        }	 | j        d	|	j        |	j        ddœ|¤Žd         dd|…f         }
t          |
|	j        |r|	j        nd|	j        ¬¦  «        S )
aC  
        input_values (`torch.Tensor` of shape `(batch_size, 1, sequence_length)`):
            Input audio waveform, sampled at `config.input_sampling_rate`.
        input_features (`torch.Tensor` of shape `(batch_size, mel_bins, time_steps)`):
            Input audio mel spectrogram for semantic encoding.
        padding_mask (`torch.Tensor` of shape `(batch_size, 1, sequence_length)`):
            Padding mask used to pad `input_values`.
        input_features_mask (`torch.Tensor` of shape `(batch_size, time_steps)`, *optional*):
            Attention mask for the spectrogram input to the semantic encoder. `1` for valid frames, `0` for padding.
        output_latents (`bool`, *optional*, defaults to `False`):
            Whether to return the continuous latent representation from the quantizer.

        Examples:

        ```python
        >>> from datasets import load_dataset
        >>> from transformers import AutoFeatureExtractor, NeuCodecModel

        >>> dataset = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
        >>> audio = dataset["train"]["audio"][0]["array"]

        >>> model_id = "neuphonic/neucodec"
        >>> model = NeuCodecModel.from_pretrained(model_id)
        >>> feature_extractor = AutoFeatureExtractor.from_pretrained(model_id)

        >>> inputs = feature_extractor(audio=audio, sampling_rate=feature_extractor.sampling_rate, return_tensors="pt")

        >>> outputs = model(**inputs)
        >>> audio_codes = outputs.audio_codes
        >>> audio_values = outputs.audio_values  # sampled at 24kHz
        ```rQ   T)r  r  r  r  Úreturn_dict)r"   r#   r  r   .N)r    r!   r"   r#   r-   )	rT   r¤   r	  r  r  r"   r#   r   r!   )rE   r]  r  r  r  r  rv   Úinput_lengthÚoutput_lengthÚencoder_outputsr    s              r/   rL   zNeuCodecModel.forward7  sÞ   € ðV $Ô)¨"Ô-ˆÝ˜L¨4Ô+MÑMÑNÔNˆàŸ+š+ØØ)Ø%Ø 3ØØð &ñ 
ô 
ˆð #�t”{ð 
Ø#Ô+Ø,Ô=Øð
ð 
ð ð	
ð 
ð
 ôð
 �.�=�.Ð ô"ˆõ Ø%Ø'Ô3Ø/=ÐG�OÔ+Ð+À4Ø,Ô=ð	
ñ 
ô 
ð 	
r.   )NNFr£   )r$   r%   r&   r   Úconfig_classr;   r   r   r(   r,   rÇ   r   r   r¥   r1   r  r3   r  r   rL   rN   rO   s   @r/   rþ  rþ  ¾  s   ø€ € € € € à!€Lð˜~ð ð ð ð ð ð ð& Øð
 -1Ø37Ø$ð:
ð :
à”lð:
ð œð:
ð ”l TÑ)ð	:
ð
 #œ\¨DÑ0ð:
ð ð:
ð Ð+Ô,ð:
ð 
Ð&Ñ	&ð:
ð :
ð :
ñ Ôñ „^ð:
ðx Øð ,0Ø'+Ø04ð	 ?ð  ?à”\ DÑ(ð ?ð ” Ñ$ð ?ð  œ,¨Ñ-ð	 ?ð
 Ð+Ô,ð ?ð 
Ð&Ñ	&ð ?ð  ?ð  ?ñ Ôñ „^ð ?ðD Øð
 -1Ø37Ø$ð@
ð @
à”lð@
ð œð@
ð ”l TÑ)ð	@
ð
 #œ\¨DÑ0ð@
ð ð@
ð Ð+Ô,ð@
ð 
�Ñ	ð@
ð @
ð @
ñ Ôñ „^ð@
ð @
ð @
ð @
ð @
r.   rþ  )r   )rn   )SrÚ   Úcollections.abcr   Údataclassesr   Únumpyr  Útorch.nnr>   r   Ú r   ra  Úactivationsr   Úcache_utilsr	   Úintegrationsr
   r   Úmasking_utilsr   Úmodeling_layersr   Úmodeling_rope_utilsr   r   Úmodeling_utilsr   r   Úprocessing_utilsr   Úutilsr   r   r   r   r   Úutils.deprecationr   Úutils.genericr   Úautor   Úconfiguration_neucodecr   r(   Útorch.nn.functionalr€   rÿ   r   r1   r3   ÚModuler5   rY   rd   r,   r¤   rm   r¸   r‰   r‹   r¨   rº   rÉ   rê   rì   r  r  r;  r[  rs  rŽ  r�  r¬  r¸  rÃ  rÓ  rß  rò  rþ  Ú__all__r-   r.   r/   ú<module>r8     sƒ  ðð* €€€Ø $Ð $Ð $Ð $Ð $Ð $Ø !Ð !Ð !Ð !Ð !Ð !à Ð Ð Ð Ø Ð Ð Ð Ð Ð Ø Ð Ð Ð Ð Ð à &Ð &Ð &Ð &Ð &Ð &Ø !Ð !Ð !Ð !Ð !Ð !Ø  Ð  Ð  Ð  Ð  Ð  Ø LÐ LÐ LÐ LÐ LÐ LÐ LÐ LØ 6Ð 6Ð 6Ð 6Ð 6Ð 6Ø 9Ð 9Ð 9Ð 9Ð 9Ð 9Ø KÐ KÐ KÐ KÐ KÐ KÐ KÐ KØ FÐ FÐ FÐ FÐ FÐ FÐ FÐ FØ &Ð &Ð &Ð &Ð &Ð &Ø jÐ jÐ jÐ jÐ jÐ jÐ jÐ jÐ jÐ jÐ jÐ jÐ jÐ jØ 0Ð 0Ð 0Ð 0Ð 0Ð 0Ø +Ð +Ð +Ð +Ð +Ð +Ø Ð Ð Ð Ð Ð Ø 2Ð 2Ð 2Ð 2Ð 2Ð 2ð ÐÑÔð $Ø€L€L€LØ#Ð#Ð#Ð#Ð#Ð#Ð#Ð#Ð#ð Ø
ð1ð 1ð 1ð 1ð 1�[ñ 1ô 1ñ „ñ „ð1ð( Ø
ð1ð 1ð 1ð 1ð 1˜Kñ 1ô 1ñ „ñ „ð1ð" Ø
ð2ð 2ð 2ð 2ð 2˜Kñ 2ô 2ñ „ñ „ð2ðð ð ð ð �"”)ñ ô ð ð(ð (ð (ð ÐÐ-Ñ.Ô.ðð ð ñ /Ô.ðð2	U˜Uœ\ð 	U°#ð 	U¸%¼,ð 	Uð 	Uð 	Uð 	Uð& ð%ð %ØŒIð%àŒ<ð%ð 
Œð%ð Œ<ð	%ð
 ”L 4Ñ'ð%ð ð%ð ð%ð Ð'Ô(ð%ð %ð %ð %ð2 ÐÐ)Ñ*Ô*ðC)ð C)ð C)ð C)ð C)˜œ	ñ C)ô C)ñ +Ô*ðC)ðL Ð˜YÑ'Ô'ðJð Jð Jð Jð J�b”iñ Jô Jñ (Ô'ðJð((ð (ð (ð (ð (Ð5ñ (ô (ð (ðV&ð &ð &ð &ð &˜œ	ñ &ô &ð &ðR+5ð +5ð +5ð\ð ð ð ð ˜2œ9ñ ô ð ðDð ð ð ð ˜œñ ô ð ð6P1ð P1ð P1ð P1ð P1 r¤yñ P1ô P1ð P1ðf3"ð 3"ð 3"ð 3"ð 3"˜œ	ñ 3"ô 3"ð 3"ðl ð%5ð %5ð %5ð %5ð %5˜oñ %5ô %5ñ „ð%5ðP6<ð 6<ð 6<ð 6<ð 6<˜bœiñ 6<ô 6<ð 6<ðrð ð ð ð  b¤iñ ô ð ð0!ð !ð !ð !ð !˜2œ9ñ !ô !ð !ðHð ð ð ð ˜2œ9ñ ô ð ð.ð ð ð ð �b”iñ ô ð ð@:ð :ð :ð :ð :˜"œ)ñ :ô :ð :ð0&ð &ð &ð &ð &˜œ	ñ &ô &ð &ð,)3ð )3ð )3ð )3ð )3�b”iñ )3ô )3ð )3ðX+ð +ð +ð +ð +˜bœiñ +ô +ð +ð\ €ÐAÐBÑBÔBðz
ð z
ð z
ð z
ð z
Ð+ñ z
ô z
ñ CÔBðz
ðz Ð5Ð
6€€€r.   