+
    ÒÜ³jp  ã                   óœ   € ^ RI t ^ RIHtHt ^RIHt ^ RIHt ^ RIH	t	 ^ RI
Ht ^ RIHt ^ RIHt ^ R	IHt  ! R
 R]4      t ! R R4      tR# )é    N)ÚListÚTuple)ÚBaseAWQForCausalLM)Úfuse_qkv)ÚLlamaLikeBlock)ÚLlamaLikeModel)ÚLlamaDecoderLayer)ÚLlavaForConditionalGeneration)ÚFasterTransformerRMSNormc                   ó´   a € ] tR t^t o RtRt]V 3R lR l4       t]V 3R lR l4       t]V 3R lR l4       t	]V 3R	 lR
 l4       t
]V 3R lR l4       tRtV tR# )ÚLlavaAWQForCausalLMr	   Úmax_position_embeddingsc                ó    <€ V ^8„  d   QhRS[ /# ©é   Úmodel©Ú OldLlavaForConditionalGeneration)ÚformatÚ__classdict__s   "€ÚQ/home/luca/screen-checking/.venv/lib/python3.14/site-packages/awq/models/llava.pyÚ__annotate__Ú LlavaAWQForCausalLM.__annotate__   s   ø€ ÷ !ñ !Ñ;ñ !ó    c                ó<   € \        V 4      pVP                  4        R # ©N)Ú
LlavaFuserÚfuse_transformer)r   Úfusers   & r   Úfuse_layersÚLlavaAWQForCausalLM.fuse_layers   s   € ä˜5Ó!ˆØ×ÑÖ r   c                ó    <€ V ^8„  d   QhRS[ /# r   r   )r   r   s   "€r   r   r      s   ø€ ÷ 1ñ 1Ñ @ñ 1r   c                óB   € V P                   P                  P                  # r   )Úlanguage_modelr   Úlayers)r   s   &r   Úget_model_layersÚ$LlavaAWQForCausalLM.get_model_layers   s   € à×#Ñ#×)Ñ)×0Ñ0Ð0r   c                ó    <€ V ^8„  d   QhRS[ /# ©r   Úmodule©ÚOldLlamaDecoderLayer)r   r   s   "€r   r   r      s   ø€ ÷ 'ñ 'Ñ$8ñ 'r   c                ó   € \        R R7      # )F)Úis_scalable)Údict)r*   s   &r   Úget_act_for_scalingÚ'LlavaAWQForCausalLM.get_act_for_scaling   s   € ä Ô&Ð&r   c                ó&   <€ V ^8„  d   QhRS[ RS[/# )r   r   Údevice)r   Ústr)r   r   s   "€r   r   r   "   s    ø€ ÷ eñ eÑ:ð eÁCñ er   c                óV  € V P                  4       P                  V4      V P                  P                  n        \        V P                  P                  R 4      '       dK   V P                  P                  P                  P                  V4      V P                  P                  n        R# R# )Ú
rotary_embN)Úget_input_embeddingsÚtor$   r   Úembed_tokensÚhasattrr6   )r   r3   s   &&r   Ú
move_embedÚLlavaAWQForCausalLM.move_embed!   s   € à27×2LÑ2LÓ2N×2QÑ2QØó3
ˆ×Ñ×"Ñ"Ô/ô �5×'Ñ'×-Ñ-¨|×<Ò<Ø49×4HÑ4H×4NÑ4N×4YÑ4Y×4\Ñ4\Ð]cÓ4dˆE× Ñ ×&Ñ&Ö1ñ =r   c                ó    <€ V ^8„  d   QhRS[ /# r)   r+   )r   r   s   "€r   r   r   *   s   ø€ ÷ 0ñ 0Ñ';ñ 0r   c           
     ó¤  € . pVP                  \        V P                  V P                  P                  V P                  P
                  V P                  P                  .VR ,          V P                  VR7      4       V P                  P                  P                  P                  V P                  P                  P                  P                  8X  dN   VP                  \        V P                  P                  V P                  P                  .VR,          R7      4       VP                  \        V P                  V P                  P                  V P                  P                  .VR,          V P                  R7      4       VP                  \        V P                  P                  V P                  P                  .VR,          R7      4       V# )zself_attn.q_proj)Úprev_opr%   ÚinpÚmodule2inspectÚkwargszself_attn.o_proj)r?   r%   r@   zmlp.gate_proj)r?   r%   r@   rA   zmlp.down_proj)Úappendr/   Úinput_layernormÚ	self_attnÚq_projÚk_projÚv_projÚweightÚshapeÚo_projÚpost_attention_layernormÚmlpÚ	gate_projÚup_projÚ	down_proj)r*   Ú
input_featÚmodule_kwargsr%   s   &&& r   Úget_layers_for_scalingÚ*LlavaAWQForCausalLM.get_layers_for_scaling)   ss  € àˆð 	�‰ÜØ×.Ñ.à×$Ñ$×+Ñ+Ø×$Ñ$×+Ñ+Ø×$Ñ$×+Ñ+ðð
 Ð1Õ2Ø%×/Ñ/Ø$ô
ô	
ð  ×Ñ×"Ñ"×)Ñ)×/Ñ/°6×3CÑ3C×3JÑ3J×3QÑ3Q×3WÑ3WÔWØ�M‰MÜØ"×,Ñ,×3Ñ3Ø"×,Ñ,×3Ñ3Ð4Ø"Ð#5Õ6ôôð 	�‰ÜØ×7Ñ7ØŸ
™
×,Ñ,¨f¯j©j×.@Ñ.@ÐAØ˜Õ/Ø%Ÿz™zô	ô	
ð 	�‰ÜØŸ
™
×*Ñ*ØŸ
™
×,Ñ,Ð-Ø˜Õ/ôô	
ð ˆr   © N)Ú__name__Ú
__module__Ú__qualname__Ú__firstlineno__Ú
layer_typeÚmax_seq_len_keyÚstaticmethodr    r&   r0   r;   rS   Ú__static_attributes__Ú__classdictcell__©r   s   @r   r   r      ss   ø‡ € Ø$€JØ/€Oà÷!ó ð!ð ÷1ó ð1ð ÷'ó ð'ð ÷eó ðeð ÷0ó ö0r   r   c                   ó8   a € ] tR t^]t o V 3R lR ltR tRtV tR# )r   c                ó    <€ V ^8„  d   QhRS[ /# r   r   )r   r   s   "€r   r   ÚLlavaFuser.__annotate__^   s   ø€ ÷ 
ñ 
Ñ>ñ 
r   c                ó  € VP                   V n        V P                  P                  4        UUu. uFA  w  r#R P                  4       VP                  P
                  P                  4       9   g   K>  W#3NKC  	  uppV n        R# u uppi )r	   N)r$   r   Únamed_modulesÚlowerÚ	__class__rV   Úllama_blocks)Úselfr   Únamer*   s   &&  r   Ú__init__ÚLlavaFuser.__init__^   sr   € Ø×)Ñ)ˆŒ
ð !%§
¡
× 8Ñ 8Ô :ôE
á :‘�Ø"×(Ñ(Ó*¨f×.>Ñ.>×.GÑ.G×.MÑ.MÓ.OÑOô ˆT‹NÙ :òE
ˆÖùó E
s   ¯;BÁ/Bc                ó  € . p\         P                   ! V P                  P                  P                  R R7       EF  p\        \	        VP                  4       P                  4       4      4      P                  p\        VVP                  P                  VP                  P                  VP                  P                  4      p\        VP                  P                  VP                  P                   4      p\        VP"                  P                  VP"                  P                   4      p\%        V P                  P&                  R4      '       d"   V P                  P&                  P(                  pM V P                  P&                  P*                  pVP-                  \/        V P                  P&                  P0                  V P                  P&                  P2                  V P                  P&                  P4                  VVP                  P6                  VP8                  VVVVV P                  P&                  P:                  R7      4       EK  	  \=        V P                  P&                  P>                  VV P                  P                  P@                  V P                  P                  PB                  4      V P                  n        \E        V P                  P                  RV P                  P                  PF                  4       R# )zFusing layers...)ÚdescÚmax_seq_len)Úhidden_sizeÚn_headsÚ
n_kv_headsÚ	qkv_layerrK   rM   Únorm_1Únorm_2Údevrn   Ú
rope_thetaÚblocksN)$Útqdmr   r%   ÚnextÚiterÚ
state_dictÚvaluesr3   r   rE   rF   rG   rH   r   rD   rI   Úvariance_epsilonrL   r:   Úconfigrn   r   rC   r   ro   Únum_attention_headsÚnum_key_value_headsrK   rM   rv   r   Ú
vocab_sizer9   ÚnormÚsetattrrw   )rh   rw   r*   r3   Úqkvrs   rt   rn   s   &       r   r   ÚLlavaFuser.fuse_transformerg   s'  € Øˆô —i’i §
¡
× 0Ñ 0× 7Ñ 7Ð>P×QÐQˆFÜœ$˜v×0Ñ0Ó2×9Ñ9Ó;Ó<Ó=×DÑDˆFÜØØ× Ñ ×'Ñ'Ø× Ñ ×'Ñ'Ø× Ñ ×'Ñ'ó	ˆCô .Ø×&Ñ&×-Ñ-¨v×/EÑ/E×/VÑ/VóˆFô .Ø×/Ñ/×6Ñ6Ø×/Ñ/×@Ñ@óˆFô �t—z‘z×(Ñ(¨-×8Ò8Ø"Ÿj™j×/Ñ/×;Ñ;‘à"Ÿj™j×/Ñ/×GÑG�Ø�M‰MÜØ $§
¡
× 1Ñ 1× =Ñ =Ø ŸJ™J×-Ñ-×AÑAØ#Ÿz™z×0Ñ0×DÑDØ!Ø!×+Ñ+×2Ñ2ØŸ
™
Ø!Ø!ØØ +Ø#Ÿz™z×0Ñ0×;Ñ;ô÷ñ' RôF *Ø�J‰J×Ñ×(Ñ(ØØ�J‰J×Ñ×)Ñ)Ø�J‰J×Ñ×!Ñ!ó	
ˆ�
‰
Ôô 	�—
‘
× Ñ  (¨D¯J©J×,<Ñ,<×,CÑ,CÖDr   )rg   r   N)rV   rW   rX   rY   rj   r   r]   r^   r_   s   @r   r   r   ]   s   ø‡ € ÷
ð 
÷-Eð -Er   r   )rx   Útypingr   r   Úbaser   Úawq.utils.fused_utilsr   Úawq.modules.fused.blockr   Úawq.modules.fused.modelr   Ú(transformers.models.llama.modeling_llamar	   r,   Ú(transformers.models.llava.modeling_llavar
   r   Úawq.modules.fused.normr   r   r   rU   r   r   Ú<module>rŽ      sB   ðÛ ß Ý $Ý *Ý 2Ý 2õõõ <ôJÐ,ô J÷Z7Eó 7Er   