
    rhgo                     P   d dl m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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 d dlmZmZmZmZ d d	lmZm Z  d
Z!dZ" G d dejF                        Z$ G d dejF                        Z% G d dejF                        Z& G d dejF                        Z' G d dejF                        Z( G d dejF                        Z) G d dejF                        Z* G d dejF                        Z+ G d dejF                        Z, G d dejF                        Z- G d  d!ejF                        Z. G d" d#ejF                        Z/ G d$ d%ejF                        Z0 G d& d'ejF                        Z1 G d( d)e      Z2 G d* d+ejF                        Z3 ed,e!       G d- d.e2             Z4d/Z5 ee4e5        ee4ee0        G d1 d2ejF                        Z6 G d3 d4ejF                        Z7 ed5e!       G d6 d7e2             Z8d8Z9 ee8e9        ee8ee0       g d9Z:y):    )partial)OptionalN)
FrozenDictfreezeunfreeze)flatten_dictunflatten_dict)RegNetConfig)"FlaxBaseModelOutputWithNoAttentionFlaxBaseModelOutputWithPooling,FlaxBaseModelOutputWithPoolingAndNoAttention(FlaxImageClassifierOutputWithNoAttention)ACT2FNFlaxPreTrainedModel append_replace_return_docstringsoverwrite_call_docstring)add_start_docstrings%add_start_docstrings_to_model_forwarda  

    This model inherits from [`FlaxPreTrainedModel`]. Check the superclass documentation for the generic methods the
    library implements for all its model (such as downloading, saving and converting weights from PyTorch models)

    This model is also a
    [flax.linen.Module](https://flax.readthedocs.io/en/latest/api_reference/flax.linen/module.html) subclass. Use it as
    a regular Flax linen Module and refer to the Flax documentation for all matter related to general usage and
    behavior.

    Finally, this model supports inherent JAX features such as:

    - [Just-In-Time (JIT) compilation](https://jax.readthedocs.io/en/latest/jax.html#just-in-time-compilation-jit)
    - [Automatic Differentiation](https://jax.readthedocs.io/en/latest/jax.html#automatic-differentiation)
    - [Vectorization](https://jax.readthedocs.io/en/latest/jax.html#vectorization-vmap)
    - [Parallelization](https://jax.readthedocs.io/en/latest/jax.html#parallelization-pmap)

    Parameters:
        config ([`RegNetConfig`]): Model configuration class with all the parameters of the model.
            Initializing with a config file does not load the weights associated with the model, only the
            configuration. Check out the [`~FlaxPreTrainedModel.from_pretrained`] method to load the model weights.
        dtype (`jax.numpy.dtype`, *optional*, defaults to `jax.numpy.float32`):
            The data type of the computation. Can be one of `jax.numpy.float32`, `jax.numpy.float16` (on GPUs) and
            `jax.numpy.bfloat16` (on TPUs).

            This can be used to enable mixed-precision training or half-precision inference on GPUs or TPUs. If
            specified all the computation will be performed with the given `dtype`.

            **Note that this only specifies the dtype of the computation and does not influence the dtype of model
            parameters.**

            If you wish to change the dtype of the model parameters, see [`~FlaxPreTrainedModel.to_fp16`] and
            [`~FlaxPreTrainedModel.to_bf16`].
a@  
    Args:
        pixel_values (`numpy.ndarray` of shape `(batch_size, num_channels, height, width)`):
            Pixel values. Pixel values can be obtained using [`AutoImageProcessor`]. See
            [`RegNetImageProcessor.__call__`] for details.

        output_hidden_states (`bool`, *optional*):
            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
            more detail.
        return_dict (`bool`, *optional*):
            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
c                   4    e Zd ZdZej
                  d        Zy)IdentityzIdentity function.c                     |S N )selfxkwargss      /var/www/html/ai-insurance-compliance-backend/venv/lib/python3.12/site-packages/transformers/models/regnet/modeling_flax_regnet.py__call__zIdentity.__call__b   s        N)__name__
__module____qualname____doc__nncompactr   r   r   r   r   r   _   s    ZZ r   r   c                       e Zd ZU eed<   dZeed<   dZeed<   dZeed<   dZe	e
   ed<   ej                  Zej                  ed	<   d
 Zddej                  dedej                  fdZy)FlaxRegNetConvLayerout_channels   kernel_size   stridegroupsrelu
activationdtypec                    t        j                  | j                  | j                  | j                  f| j                  | j                  dz  | j
                  dt         j                  j                  ddd      | j                        | _	        t        j                  dd	| j                  
      | _        | j                  t        | j                     | _        y t               | _        y )N   F       @fan_outtruncated_normalmodedistribution)r*   stridespaddingfeature_group_countuse_biaskernel_initr0   ?h㈵>momentumepsilonr0   )r$   Convr(   r*   r,   r-   initializersvariance_scalingr0   convolution	BatchNormnormalizationr/   r   r   activation_funcr   s    r   setupzFlaxRegNetConvLayer.setupo   s    77))4+;+;<KK$$) $889[m8n**	
  \\3TZZX:>//:Uvdoo6[c[er   hidden_statedeterministicreturnc                 p    | j                  |      }| j                  ||      }| j                  |      }|S N)use_running_average)rF   rH   rI   )r   rL   rM   s      r   r   zFlaxRegNetConvLayer.__call__}   s=    ''5)),M)Z++L9r   NT)r    r!   r"   int__annotations__r*   r,   r-   r/   r   strjnpfloat32r0   rK   ndarrayboolr   r   r   r   r'   r'   g   so    KFCOFCO &J&{{E399"fS[[  QTQ\Q\ r   r'   c                       e Zd ZU eed<   ej                  Zej                  ed<   d Zd	dej                  de
dej                  fdZy)
FlaxRegNetEmbeddingsconfigr0   c                     t        | j                  j                  dd| j                  j                  | j                        | _        y )Nr)   r2   )r*   r,   r/   r0   )r'   r\   embedding_size
hidden_actr0   embedderrJ   s    r   rK   zFlaxRegNetEmbeddings.setup   s5    +KK&&{{--**
r   pixel_valuesrM   rN   c                     |j                   d   }|| j                  j                  k7  rt        d      | j	                  ||      }|S )NzeMake sure that the channel dimension of the pixel values match with the one set in the configuration.rM   )shaper\   num_channels
ValueErrorr`   )r   ra   rM   rf   rL   s        r   r   zFlaxRegNetEmbeddings.__call__   sN    #))"-4;;333w  }}\}Or   NrR   )r    r!   r"   r
   rT   rV   rW   r0   rK   rX   rY   r   r   r   r   r[   r[      sD    {{E399"
S[[  QTQ\Q\ r   r[   c                       e Zd ZU dZeed<   dZeed<   ej                  Z	ej                  ed<   d Z
ddej                  ded	ej                  fd
Zy)FlaxRegNetShortCutz
    RegNet shortcut, used to project the residual features to the correct size. If needed, it is also used to
    downsample the input using `stride=2`.
    r(   r2   r,   r0   c                    t        j                  | j                  d| j                  dt         j                  j                  ddd      | j                        | _        t        j                  dd	| j                  
      | _	        y )Nr+   r+   Fr3   r4   r5   r6   )r*   r9   r<   r=   r0   r>   r?   r@   )
r$   rC   r(   r,   rD   rE   r0   rF   rG   rH   rJ   s    r   rK   zFlaxRegNetShortCut.setup   se    77KK889[m8n**
  \\3TZZXr   r   rM   rN   c                 N    | j                  |      }| j                  ||      }|S rP   )rF   rH   )r   r   rM   rL   s       r   r   zFlaxRegNetShortCut.__call__   s-    ''*)),M)Zr   NrR   )r    r!   r"   r#   rS   rT   r,   rV   rW   r0   rK   rX   rY   r   r   r   r   ri   ri      sR    
 FCO{{E399"	Y#++ d ckk r   ri   c                       e Zd ZU eed<   eed<   ej                  Zej                  ed<   d Zdej                  dej                  fdZ
y)	FlaxRegNetSELayerCollectionin_channelsreduced_channelsr0   c           	      P   t        j                  | j                  dt         j                  j	                  ddd      | j
                  d      | _        t        j                  | j                  dt         j                  j	                  ddd      | j
                  d      | _        y )	Nrk   r3   r4   r5   r6   0)r*   r=   r0   name2)	r$   rC   rp   rD   rE   r0   conv_1ro   conv_2rJ   s    r   rK   z!FlaxRegNetSELayerCollection.setup   s    gg!!889[m8n**
 gg889[m8n**
r   rL   rN   c                     | j                  |      }t        j                  |      }| j                  |      }t        j                  |      }|S r   )ru   r$   r.   rv   sigmoid)r   rL   	attentions      r   r   z$FlaxRegNetSELayerCollection.__call__   s@    {{<0ww|,{{<0JJ|,	r   N)r    r!   r"   rS   rT   rV   rW   r0   rK   rX   r   r   r   r   rn   rn      s@    {{E399"
 S[[ S[[ r   rn   c                       e Zd ZU dZeed<   eed<   ej                  Zej                  ed<   d Z	dej                  dej                  fdZy	)
FlaxRegNetSELayerz
    Squeeze and Excitation layer (SE) proposed in [Squeeze-and-Excitation Networks](https://huggingface.co/papers/1709.01507).
    ro   rp   r0   c                     t        t        j                  d      | _        t	        | j
                  | j                  | j                        | _        y )Nr   r   r~   r:   r0   )	r   r$   avg_poolpoolerrn   ro   rp   r0   ry   rJ   s    r   rK   zFlaxRegNetSELayer.setup   s8    bkk3CD4T5E5EtG\G\dhdndnor   rL   rN   c                     | j                  ||j                  d   |j                  d   f|j                  d   |j                  d   f      }| j                  |      }||z  }|S )Nr+   r2   window_shaper9   )r   re   ry   )r   rL   pooledry   s       r   r   zFlaxRegNetSELayer.__call__   st    &,,Q/1C1CA1FG!''*L,>,>q,AB  

 NN6*	#i/r   N)r    r!   r"   r#   rS   rT   rV   rW   r0   rK   rX   r   r   r   r   r{   r{      sH     {{E399"pS[[ S[[ r   r{   c                       e Zd ZU eed<   eed<   dZeed<   ej                  Z	ej                  ed<   d Z
ddej                  ded	ej                  fd
Zy)FlaxRegNetXLayerCollectionr\   r(   r+   r,   r0   c           	         t        d| j                  | j                  j                  z        }t	        | j                  d| j                  j
                  | j                  d      t	        | j                  | j                  || j                  j
                  | j                  d      t	        | j                  dd | j                  d      g| _        y )Nr+   rr   r*   r/   r0   rs   1r,   r-   r/   r0   rs   rt   )	maxr(   r\   groups_widthr'   r_   r0   r,   layerr   r-   s     r   rK   z FlaxRegNetXLayerCollection.setup   s    Q))T[[-E-EEF  !!;;11jj  !!{{;;11jj  !!jj!

r   rL   rM   rN   c                 <    | j                   D ]  } |||      } |S Nrd   r   )r   rL   rM   r   s       r   r   z#FlaxRegNetXLayerCollection.__call__  s)    ZZ 	LE ]KL	Lr   NrR   )r    r!   r"   r
   rT   rS   r,   rV   rW   r0   rK   rX   rY   r   r   r   r   r   r      sS    FCO{{E399"
8S[[  QTQ\Q\ r   r   c                       e Zd ZU dZeed<   eed<   eed<   dZeed<   ej                  Z
ej                  ed<   d Zdd	ej                  d
edej                  fdZy)FlaxRegNetXLayerzt
    RegNet's layer composed by three `3x3` convolutions, same as a ResNet bottleneck layer with reduction = 1.
    r\   ro   r(   r+   r,   r0   c                    | j                   | j                  k7  xs | j                  dk7  }|r,t        | j                  | j                  | j                        n	t               | _        t        | j                  | j                   | j                  | j                  | j                        | _	        t        | j                  j                     | _        y Nr+   )r,   r0   )ro   r(   r,   r0   )ro   r(   r,   ri   r0   r   shortcutr   r\   r   r   r_   rI   r   should_apply_shortcuts     r   rK   zFlaxRegNetXLayer.setup   s     $ 0 0D4E4E E YXYIY % !!{{jj  	 0KK((**;;**

  &dkk&<&<=r   rL   rM   rN   c                 ~    |}| j                  |      }| j                  ||      }||z  }| j                  |      }|S r   r   r   rI   r   rL   rM   residuals       r   r   zFlaxRegNetXLayer.__call__4  G    zz,/===G ++L9r   NrR   r    r!   r"   r#   r
   rT   rS   r,   rV   rW   r0   rK   rX   rY   r   r   r   r   r   r     s`     FCO{{E399">(S[[  QTQ\Q\ r   r   c                       e Zd ZU eed<   eed<   eed<   dZeed<   ej                  Z	ej                  ed<   d Z
dej                  d	ej                  fd
Zy)FlaxRegNetYLayerCollectionr\   ro   r(   r+   r,   r0   c                 &   t        d| j                  | j                  j                  z        }t	        | j                  d| j                  j
                  | j                  d      t	        | j                  | j                  || j                  j
                  | j                  d      t        | j                  t        t        | j                  dz              | j                  d      t	        | j                  dd | j                  d	      g| _        y )
Nr+   rr   r   r   r      rt   )rp   r0   rs   3)r   r(   r\   r   r'   r_   r0   r,   r{   rS   roundro   r   r   s     r   rK   z FlaxRegNetYLayerCollection.setupD  s    Q))T[[-E-EEF  !!;;11jj  !!{{;;11jj !!!$U4+;+;a+?%@!Ajj	  !!jj-

r   rL   rN   c                 8    | j                   D ]
  } ||      } |S r   r   )r   rL   r   s      r   r   z#FlaxRegNetYLayerCollection.__call__f  s%    ZZ 	/E .L	/r   N)r    r!   r"   r
   rT   rS   r,   rV   rW   r0   rK   rX   r   r   r   r   r   r   =  sP    FCO{{E399" 
DS[[ S[[ r   r   c                       e Zd ZU dZeed<   eed<   eed<   dZeed<   ej                  Z
ej                  ed<   d Zdd	ej                  d
edej                  fdZy)FlaxRegNetYLayerzC
    RegNet's Y layer: an X layer with Squeeze and Excitation.
    r\   ro   r(   r+   r,   r0   c                    | j                   | j                  k7  xs | j                  dk7  }|r,t        | j                  | j                  | j                        n	t               | _        t        | j                  | j                   | j                  | j                  | j                        | _	        t        | j                  j                     | _        y r   )ro   r(   r,   ri   r0   r   r   r   r\   r   r   r_   rI   r   s     r   rK   zFlaxRegNetYLayer.setupw  s     $ 0 0D4E4E E YXYIY % !!{{jj  	 0KK((**;;**

  &dkk&<&<=r   rL   rM   rN   c                 ~    |}| j                  |      }| j                  ||      }||z  }| j                  |      }|S r   r   r   s       r   r   zFlaxRegNetYLayer.__call__  r   r   NrR   r   r   r   r   r   r   l  s`     FCO{{E399">*S[[  QTQ\Q\ r   r   c                       e Zd ZU dZeed<   eed<   eed<   dZeed<   dZeed<   e	j                  Ze	j                  ed<   d	 Zdd
e	j                  dede	j                  fdZy)FlaxRegNetStageLayersCollection4
    A RegNet stage composed by stacked layers.
    r\   ro   r(   r2   r,   depthr0   c                    | j                   j                  dk(  rt        nt        } || j                   | j                  | j
                  | j                  | j                  d      g}t        | j                  dz
        D ]R  }|j                   || j                   | j
                  | j
                  | j                  t        |dz                      T || _        y )Nr   rr   )r,   r0   rs   r+   r0   rs   )r\   
layer_typer   r   ro   r(   r,   r0   ranger   appendrU   layers)r   r   r   is       r   rK   z%FlaxRegNetStageLayersCollection.setup  s    $(KK$:$:c$A GW   !!{{jj

 tzzA~& 		AMMKK%%%%**QU		 r   r   rM   rN   c                 @    |}| j                   D ]  } |||      } |S r   r   )r   r   rM   rL   r   s        r   r   z(FlaxRegNetStageLayersCollection.__call__  s.    [[ 	LE ]KL	Lr   NrR   r    r!   r"   r#   r
   rT   rS   r,   r   rV   rW   r0   rK   rX   rY   r   r   r   r   r   r     sf     FCOE3N{{E399"8#++ d ckk r   r   c                       e Zd ZU dZeed<   eed<   eed<   dZeed<   dZeed<   e	j                  Ze	j                  ed<   d	 Zdd
e	j                  dede	j                  fdZy)FlaxRegNetStager   r\   ro   r(   r2   r,   r   r0   c                     t        | j                  | j                  | j                  | j                  | j
                  | j                        | _        y )N)ro   r(   r,   r   r0   )r   r\   ro   r(   r,   r   r0   r   rJ   s    r   rK   zFlaxRegNetStage.setup  s<    5KK((**;;****
r   r   rM   rN   c                 (    | j                  ||      S r   r   )r   r   rM   s      r   r   zFlaxRegNetStage.__call__  s    {{1M{::r   NrR   r   r   r   r   r   r     sf     FCOE3N{{E399"
;#++ ;d ;ckk ;r   r   c            	           e Zd ZU eed<   ej                  Zej                  ed<   d Z	 	 d
dej                  de
de
defdZy	)FlaxRegNetStageCollectionr\   r0   c                 v   t        | j                  j                  | j                  j                  dd        }t        | j                  | j                  j                  | j                  j                  d   | j                  j
                  rdnd| j                  j                  d   | j                  d      g}t        t        || j                  j                  dd              D ]K  \  }\  \  }}}|j                  t        | j                  |||| j                  t        |dz                      M || _        y )Nr+   r   r2   rr   )r,   r   r0   rs   )r   r0   rs   )zipr\   hidden_sizesr   r^   downsample_in_first_stagedepthsr0   	enumerater   rU   stages)r   in_out_channelsr   r   ro   r(   r   s          r   rK   zFlaxRegNetStageCollection.setup  s   dkk668P8PQRQS8TU**((+ KKAAqqkk((+jj

 8A_VZVaVaVhVhijikVlAm7n 	3A3+lUMM[,e[_[e[elopqtupulvw	
 r   rL   output_hidden_statesrM   rN   c                     |rdnd }| j                   D ]&  }|r||j                  dddd      fz   } |||      }( ||fS )Nr   r   r)   r+   r2   rd   )r   	transpose)r   rL   r   rM   hidden_statesstage_modules         r   r   z"FlaxRegNetStageCollection.__call__  s]     3 KK 	SL# -1G1G1aQR1S0U U'MRL		S ]**r   N)FTr    r!   r"   r
   rT   rV   rW   r0   rK   rX   rY   r   r   r   r   r   r   r     sV    {{E399"0 &+"	+kk+ #+ 	+
 
,+r   r   c                       e Zd ZU eed<   ej                  Zej                  ed<   d Z	 	 	 ddej                  de
de
de
def
d	Zy
)FlaxRegNetEncoderr\   r0   c                 P    t        | j                  | j                        | _        y )Nr   )r   r\   r0   r   rJ   s    r   rK   zFlaxRegNetEncoder.setup  s    /4::Nr   rL   r   return_dictrM   rN   c                     | j                  |||      \  }}|r||j                  dddd      fz   }|st        d ||fD              S t        ||      S )N)r   rM   r   r)   r+   r2   c              3   &   K   | ]	  }||  y wr   r   ).0vs     r   	<genexpr>z-FlaxRegNetEncoder.__call__.<locals>.<genexpr>!  s     SqQ]Ss   )last_hidden_stater   )r   r   tupler   )r   rL   r   r   rM   r   s         r   r   zFlaxRegNetEncoder.__call__  st     '+kk/CS` '2 '
#m  )\-C-CAq!Q-O,QQMS\=$ASSS1*'
 	
r   N)FTTr   r   r   r   r   r     sd    {{E399"O &+ "
kk
 #
 	

 
 
,
r   r   c                   "    e Zd ZU dZeZdZdZdZe	j                  ed<   ddej                  dfd	ed
edej                  def fdZddej&                  j(                  dededefdZ ee      	 	 	 	 ddee   dedee   dee   fd       Z xZS )FlaxRegNetPreTrainedModelz
    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
    models.
    regnetra   Nmodule_class)r+      r   r)   r   Tr\   seedr0   _do_initc                      | j                   d||d|}|$d|j                  |j                  |j                  f}t        |   ||||||       y )Nr\   r0   r+   )input_shaper   r0   r   r   )r   
image_sizerf   super__init__)	r   r\   r   r   r0   r   r   module	__class__s	           r   r   z"FlaxRegNetPreTrainedModel.__init__5  sc     #""H&HHf//1B1BFDWDWXK[tSXcklr   rngr   paramsrN   c                 X   t        j                  || j                        }d|i}| j                  j	                  ||d      }|dt        t        |            }t        t        |            }| j                  D ]
  }||   ||<    t               | _        t        t        |            S |S )Nr   r   F)r   )rV   zerosr0   r   initr   r   _missing_keyssetr   r	   )r   r   r   r   ra   rngsrandom_paramsmissing_keys           r   init_weightsz&FlaxRegNetPreTrainedModel.init_weightsC  s    yyDJJ?#((|(O(-)@AM!(6"23F#11 A&3K&@{#A!$D.011  r   trainr   r   c           	         ||n| j                   j                  }||n| j                   j                  }t        j                  |d      }i }| j
                  j                  ||d   n| j                  d   ||d   n| j                  d   dt        j                  |t        j                        | ||||rdg      S d      S )N)r   r2   r)   r+   r   batch_stats)r   r   r   F)r   mutable)
r\   r   r   rV   r   r   applyr   arrayrW   )r   ra   r   r   r   r   r   s          r   r   z"FlaxRegNetPreTrainedModel.__call__U  s     %9$D $++JjJj 	 &1%<k$++BYBY}}\<@ {{  .4.@&*dkkRZF[8>8Jvm4PTP[P[\iPj IIl#++6I ',]O ! 
 	
 38 ! 
 	
r   r   )NFNN)r    r!   r"   r#   r
   config_classbase_model_prefixmain_input_namer   r$   ModulerT   rV   rW   rS   r0   rY   r   jaxrandomPRNGKeyr   r   r   r   REGNET_INPUTS_DOCSTRINGr   dictr   __classcell__)r   s   @r   r   r   *  s    
  L $O"L"))"
 %;;mm 	m
 yym m!

 2 2 ! !PZ !fp !$ ++BC "&/3&*
 
 	

 'tn
 d^
 D
r   r   c            	       t    e Zd ZU eed<   ej                  Zej                  ed<   d Z	 	 	 d
de	de	de	de
fdZy	)FlaxRegNetModuler\   r0   c                     t        | j                  | j                        | _        t	        | j                  | j                        | _        t        t        j                  d      | _	        y )Nr   r}   r   )
r[   r\   r0   r`   r   encoderr   r$   r   r   rJ   s    r   rK   zFlaxRegNetModule.setup{  sF    ,T[[

K(DJJG KK$
r   rM   r   r   rN   c                    ||n| j                   j                  }||n| j                   j                  }| j                  ||      }| j	                  ||||      }|d   }| j                  ||j                  d   |j                  d   f|j                  d   |j                  d   f      j                  dddd      }|j                  dddd      }|s
||f|dd  z   S t        |||j                        S )	Nrd   )r   r   rM   r   r+   r2   r   r)   )r   pooler_outputr   )
r\   r   use_return_dictr`   r  r   re   r   r   r   )	r   ra   rM   r   r   embedding_outputencoder_outputsr   pooled_outputs	            r   r   zFlaxRegNetModule.__call__  s-    %9$D $++JjJj 	 &1%<k$++B]B]==]=S,,!5#'	 ' 
 ,A.+11!46G6M6Ma6PQ&,,Q/1B1H1H1KL $ 
 )Aq!Q
	 	 .771aC%}58KKK;/')77
 	
r   N)TFT)r    r!   r"   r
   rT   rV   rW   r0   rK   rY   r   r   r   r   r   r  r  w  sW    {{E399"
 #%* &
 &
 #	&

 &
 
6&
r   r  zOThe bare RegNet model outputting raw features without any specific head on top.c                       e Zd ZeZy)FlaxRegNetModelN)r    r!   r"   r  r   r   r   r   r  r    s	    
 $Lr   r  at  
    Returns:

    Examples:

    ```python
    >>> from transformers import AutoImageProcessor, FlaxRegNetModel
    >>> from PIL import Image
    >>> import requests

    >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
    >>> image = Image.open(requests.get(url, stream=True).raw)

    >>> image_processor = AutoImageProcessor.from_pretrained("facebook/regnet-y-040")
    >>> model = FlaxRegNetModel.from_pretrained("facebook/regnet-y-040")

    >>> inputs = image_processor(images=image, return_tensors="np")
    >>> outputs = model(**inputs)
    >>> last_hidden_states = outputs.last_hidden_state
    ```
)output_typer   c                       e Zd ZU eed<   ej                  Zej                  ed<   d Zdej                  dej                  fdZ
y)FlaxRegNetClassifierCollectionr\   r0   c                 z    t        j                  | j                  j                  | j                  d      | _        y )Nr   r   )r$   Denser\   
num_labelsr0   
classifierrJ   s    r   rK   z$FlaxRegNetClassifierCollection.setup  s%    ((4;;#9#9RUVr   r   rN   c                 $    | j                  |      S r   )r  )r   r   s     r   r   z'FlaxRegNetClassifierCollection.__call__  s    q!!r   N)r    r!   r"   r
   rT   rV   rW   r0   rK   rX   r   r   r   r   r  r    s;    {{E399"W"#++ "#++ "r   r  c                   j    e Zd ZU eed<   ej                  Zej                  ed<   d Z	 	 	 	 dde	fdZ
y)&FlaxRegNetForImageClassificationModuler\   r0   c                     t        | j                  | j                        | _        | j                  j                  dkD  r't        | j                  | j                        | _        y t               | _        y )Nr   r   r   )r  r\   r0   r   r  r  r  r   rJ   s    r   rK   z,FlaxRegNetForImageClassificationModule.setup  sL    &dkkL;;!!A%<T[[PTPZPZ[DO&jDOr   NrM   c                    ||n| j                   j                  }| j                  ||||      }|r|j                  n|d   }| j	                  |d d d d ddf         }|s|f|dd  z   }|S t        ||j                        S )N)rM   r   r   r+   r   r2   )logitsr   )r\   r	  r   r  r  r   r   )	r   ra   rM   r   r   outputsr  r  outputs	            r   r   z/FlaxRegNetForImageClassificationModule.__call__  s     &1%<k$++B]B]++'!5#	  
 2=--'!*q!Qz!:;Y,FM7vU\UjUjkkr   )NTNN)r    r!   r"   r
   rT   rV   rW   r0   rK   rY   r   r   r   r   r  r    s>    {{E399") "!l lr   r  z
    RegNet Model with an image classification head on top (a linear layer on top of the pooled features), e.g. for
    ImageNet.
    c                       e Zd ZeZy) FlaxRegNetForImageClassificationN)r    r!   r"   r  r   r   r   r   r  r    s	     :Lr   r  aa  
    Returns:

    Example:

    ```python
    >>> from transformers import AutoImageProcessor, FlaxRegNetForImageClassification
    >>> from PIL import Image
    >>> import jax
    >>> import requests

    >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
    >>> image = Image.open(requests.get(url, stream=True).raw)

    >>> image_processor = AutoImageProcessor.from_pretrained("facebook/regnet-y-040")
    >>> model = FlaxRegNetForImageClassification.from_pretrained("facebook/regnet-y-040")

    >>> inputs = image_processor(images=image, return_tensors="np")
    >>> outputs = model(**inputs)
    >>> logits = outputs.logits

    >>> # model predicts one of the 1000 ImageNet classes
    >>> predicted_class_idx = jax.numpy.argmax(logits, axis=-1)
    >>> print("Predicted class:", model.config.id2label[predicted_class_idx.item()])
    ```
)r  r  r   );	functoolsr   typingr   
flax.linenlinenr$   r   	jax.numpynumpyrV   flax.core.frozen_dictr   r   r   flax.traverse_utilr   r	   transformersr
   "transformers.modeling_flax_outputsr   r   r   r    transformers.modeling_flax_utilsr   r   r   r   transformers.utilsr   r   REGNET_START_DOCSTRINGr   r   r   r'   r[   ri   rn   r{   r   r   r   r   r   r   r   r   r   r  r  FLAX_VISION_MODEL_DOCSTRINGr  r  r  FLAX_VISION_CLASSIF_DOCSTRING__all__r   r   r   <module>r0     s  "    
  > > ; %  ! F ryy ")) :299 0 6")) <		 0% %P%ryy %P, ,^&ryy &R,bii ,`;bii ;6'+		 '+V
		 
>I
 3 I
Z4
ryy 4
n U$/ $	$ , *E F  ."RYY "$lRYY $lN  :'@ ::! 6 9;X Y  $8 _r   