qyiRSSKJr SSKJr SSKrSSKJr SSKJr SSKJ r J r SSK J r SS K JrJrJr SS KJr SS KJrJrJrJr SS KJrJr SS KJrJr SSKJrJ r SSK!J"r" SSK#J$r$J%r%J&r&J'r' SSK(J)r)J*r* SSK+J,r, SSK-J.r. \'R^"\05r1\"S5"SS\Rd55r3"SS\Rd5r4Sr5\"S5S=Sj5r6"SS\Rd5r7S\RpS \9S!\Rp4S"jr:S>S#\RdS$\RpS%\RpS&\RpS'\RpS-S(\;S)\;S*\"\$4S+jjr<\"\65"S,S-\Rd55r="S.S/\5r>\%"S0S1\ 55r?\%"S2S3\?55r@\%"S4S5\?\ 55rA"S6S7\\?5rB"S8S9\\?5rC"S:S;\\?5rD/S[TU]5 [R"[R "U55UlX lg)z+ LlamaRMSNorm is equivalent to T5LayerNorm N)super__init__r Parametertorchonesweightvariance_epsilon)self hidden_sizer' __class__s e/home/mande/repo/quber/.venv/lib/python3.13/site-packages/transformers/models/llama/modeling_llama.pyr+LlamaRMSNorm.__init__6s/ ll5::k#:; # hidden_statescURnUR[R5nUR S5R SSS9nU[R "X0R-5-nURURU5-$)NT)keepdim) dtypetor-float32powmeanrsqrtr0r/)r1r7 input_dtypevariances r4forwardLlamaRMSNorm.forward>sw#)) %((7  $$Q',,R,>% H?T?T4T(UU {{]--k:::r6c^[URR5SUR3$)Nz, eps=)tupler/shaper0)r1s r4 extra_reprLlamaRMSNorm.extra_reprEs* ))*+6$2G2G1HIIr6)r0r/)gư>) __name__ __module__ __qualname____firstlineno__floatr+r-TensorrDrI__static_attributes__ __classcell__r3s@r4r%r%4sB$$$$$;U\\;ell;JJr6r%c ^\rSrSr%\R \S'SS\4U4Sjjjr\ SS\S-S\ SS\ S-S \ S \ 44S jj5r\R"5\S 55rS rU=r$)LlamaRotaryEmbeddingIinv_freqNconfigc>[TU]5 URUlURUlXlUR R SUlURnURS:wa[URnU"UR U5uo@l URSUSS9 URSUR5SS9 g)N rope_typedefaultrWF) persistentoriginal_inv_freq) r*r+max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenrXrope_parametersrZcompute_default_rope_parametersrattention_scalingregister_bufferclone)r1rXdevice rope_init_fnrWr3s r4r+LlamaRotaryEmbedding.__init__Ls "("@"@$*$B$B! 44[A!%!E!E >>Y &.t~~>L+7 V+L(( ZeD 0(..2BuUr6rfz torch.deviceseq_lenr(z torch.Tensorc URSn[USS5=(d URUR-nSnSU[R "SUS[R S9RU[RS9U- -- nXe4$) aH Computes the inverse frequencies according to the original RoPE implementation Args: config ([`~transformers.PreTrainedConfig`]): The model configuration. device (`torch.device`): The device to use for initialization of the inverse frequencies. seq_len (`int`, *optional*): The current sequence length. Unused for this type of RoPE. Returns: Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE). rope_thetahead_dimNg?rr9r<)rfr<) ragetattrr2num_attention_headsr-arangeint64r=rO)rXrfribasedimattention_factorrWs r4rb4LlamaRotaryEmbedding.compute_default_rope_parameters\s&%%l3fj$/c63E3EIcIc3c U\\!S!5;;?BB&X]XcXcBdgjj k ))r6cLURSSS2S4R5RURSSS5R UR 5nUSS2SSS24R5n[ UR R[5(a0UR RS:waUR ROSn[USS9 UR5UR5-RSS5n[R"Xf4SS 9nUR5UR-nUR5UR-n SSS5 WR UR S 9W R UR S 94$!,(df  N@=f) Nrr:r!mpscpuF) device_typeenabledr9rsrm)rWrOexpandrHr=rf isinstancetypestrr transposer-catcosrcsinr<) r1x position_idsinv_freq_expandedposition_ids_expandedryfreqsembrrs r4rDLlamaRotaryEmbedding.forwardzsN!MM$4-8>>@GG HZHZ[\H]_acdehhijiqiqr ,QaZ 8 > > @'1!((--'E'E!((--[`J`ahhmmfk  U C&,,.1F1L1L1NNYYZ[]^_E))UN3C'')d444C'')d444C D vvAGGv$cff177f&;;; D Cs BF F#)rcrXr_r`rZNNNN)rKrLrMrNr-rP__annotations__r"r+ staticmethodrintrGrOrbno_gradrrDrQrRrSs@r4rUrUIsllV{VV %)+/"*d"*(*t* ~u$ % **: ]]_ <