§
    9²¸j£  ã                   óÎ   — d dl Z ddlmZ ddlmZ  ej        e¦  «        Z ed¬¦  «        	 dde j        d	e j        d
e j        de j        de	de
de j        fd„¦   «         Z	 dd„ZdS )é    Né   )Úlogging)Úrequires)Ú
torchaudio)ÚbackendsÚmean_volumeÚlogitsÚtargetsÚlogit_lengthsÚtarget_lengthsÚblank_token_idÚ	reductionÚreturnc           	      óP  — ddl }d}||vr3t          d|› dd                     d„ |D ¦   «         ¦  «        › d�¦  «        ‚|                     | j        ¦  «        }|j                             |                      ¦   «                              ¦   «         |                     | j        ¦  «         	                    ¦   «         |                     | j        ¦  «         	                    ¦   «         | 	                    ¦   «         |d	¬
¦  «        }|dk    r;| 
                    ¦   «         |                     ¦   «          
                    ¦   «         z  S |dk    r|                     ¦   «         S |dk    r)||                     ¦   «         z                       ¦   «         S |dk    r| 
                    ¦   «         S |S )a  
    Compute standard RNN-T (RNN Transducer) loss (https://huggingface.co/papers/1211.3711).

    Thin wrapper around [`torchaudio.functional.rnnt_loss`]. torchaudio is queried with `reduction="none"` to get
    the per-sample negative log-likelihoods, and the requested reduction is applied here. The reduction names and
    formulas mirror NeMo's `RNNTLoss` (the reference implementation used to train/finetune Parakeet), so that loss
    magnitudes and gradient scaling match when finetuning other RNNT models like Parakeet:

    - `"mean_volume"`: sum of per-sample losses divided by the sum of target lengths (per-token average over the
      whole batch). This is what `nvidia/parakeet-rnnt-0.6b` is trained with (`rnnt_reduction: mean_volume`).
    - `"mean_batch"`: plain average of per-sample losses over the batch (NeMo's default).
    - `"mean"`: per-sample loss divided by its own target length, then averaged over the batch.
    - `"sum"`: sum of per-sample losses.
    - `"none"`: per-sample losses, unreduced.

    Args:
        logits: Joint token logits of shape `(batch, T, U+1, vocab_size)`.
        targets: Target labels of shape `(batch, U)`.
        logit_lengths: Encoder output lengths of shape `(batch,)`.
        target_lengths: Target lengths of shape `(batch,)`.
        blank_token_id: Blank token id.
        reduction: Loss reduction method. One of `"mean_volume"`, `"mean_batch"`, `"mean"`, `"sum"`, or `"none"`.

    Returns:
        Scalar loss tensor (or per-example losses if `reduction="none"`).

    r   N)r   Ú
mean_batchÚmeanÚsumÚnonezInvalid reduction mode "z". Expected one of z, c              3   ó4   K  — | ]}t          |¦  «        V — Œd S )N)Úrepr)Ú.0Úrs     úk/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/transformers/loss/loss_rnnt.pyú	<genexpr>zrnnt_loss.<locals>.<genexpr>B   s*   è è € ÐNqÐNqÐ[\ÍtÐTUÉwÌwÐNqÐNqÐNqÐNqÐNqÐNqó    ú.r   )r	   r
   r   r   Úblankr   r   r   r   r   )r   Ú
ValueErrorÚjoinÚtoÚdeviceÚ
functionalÚ	rnnt_lossÚfloatÚ
contiguousÚintr   r   )	r	   r
   r   r   r   r   r   Úvalid_reductionsÚlossess	            r   r#   r#      s   € ðJ ÐÐÐàKÐØÐ(Ð(Ð(ÝØt yÐtÐtÀTÇYÂYÐNqÐNqÐ`pÐNqÑNqÔNqÑEqÔEqÐtÐtÐtñ
ô 
ð 	
ð $×&Ò& v¤}Ñ5Ô5€NØÔ"×,Ò,Ø�|Š|‰~Œ~×(Ò(Ñ*Ô*Ø—
’
˜6œ=Ñ)Ô)×-Ò-Ñ/Ô/Ø#×&Ò& v¤}Ñ5Ô5×9Ò9Ñ;Ô;Ø%×)Ò)Ñ+Ô+ØØð -ñ ô €Fð �MÒ!Ð!Ø�zŠz‰|Œ|˜n×2Ò2Ñ4Ô4×8Ò8Ñ:Ô:Ñ:Ð:Ø	�lÒ	"Ð	"Ø�{Š{‰}Œ}ÐØ	�fÒ	Ð	Ø˜×-Ò-Ñ/Ô/Ñ/×5Ò5Ñ7Ô7Ð7Ø	�eÒ	Ð	Ø�zŠz‰|Œ|ÐØ€Mr   c                 ó,   — t          | |||||¬¦  «        S )N)r	   r
   r   r   r   r   )r#   )r	   Úlabelsr   Úlabel_lengthsr   r   Úkwargss          r   ÚParakeetForRNNTLossr-   Z   s-   € õ ØØØ#Ø$Ø%Øðñ ô ð r   )r   )ÚtorchÚutilsr   Úutils.import_utilsr   Ú
get_loggerÚ__name__ÚloggerÚTensorr&   Ústrr#   r-   © r   r   ú<module>r7      sð   ðð €€€à Ð Ð Ð Ð Ð Ø )Ð )Ð )Ð )Ð )Ð )ð 
ˆÔ	˜HÑ	%Ô	%€ð 
€�?Ð#Ñ#Ô#ð #ð>ð >ØŒLð>àŒ\ð>ð ”<ð>ð ”Lð	>ð
 ð>ð ð>ð „\ð>ð >ð >ñ $Ô#ð>ðN ðð ð ð ð ð r   