Zj{PSrSSKrSSKJr SSKJr SSKrSSKJs J r SSKJ r J r Jr SSK Jr SSKJr SS KJrJrJr SS KJr SS KJr SS KJrJr SS KJrJrJ r J!r! SSK"J#r#J$r$ SSK%J&r&J'r' SSK(J)r) SSK*J+r+ SSK,J-r-J.r.J/r/J0r0 SSK1J2r2J3r3 SSK4J5r5 SSK6J7r7 \0Rp"\95r:Sr;"SS\Rx5r="SS\R|5r?Sr@S:SjrA"SS\R|5rBS \RS!\CS"\R4S#jrD"S$S%\R|5rE"S&S'\E5rF"S(S)\E5rG\E\F\GS*.rH"S+S,\!5rI\."S-S.\)55rJ\."S/S0\J55rK"S1S2\J\5rL"S3S4\\J5rM"S5S6\\J5rN"S7S8\ \J5rO/S9QrPg);zPyTorch Nemotron model.N)Callable)Optional)SizeTensornn)initialization)ACT2FN)Cache DynamicCache StaticCache)GenerationMixin)create_causal_mask)_flash_attention_forward!flash_attn_supports_top_left_mask)GenericForQuestionAnswering GenericForSequenceClassificationGenericForTokenClassificationGradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)PreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tuplelogging)maybe_autocastmerge_with_config_defaults)capture_outputs)NemotronConfigc[R"5(dU$[R"U5n[RRR XU5$N)torchis_autocast_enabledget_autocast_dtypeamp autocast_mode_cast) device_typeargs target_dtypes /root/GenerationalWealth/GenerationalWealth/venv/lib/python3.13/site-packages/transformers/models/nemotron/modeling_nemotron.py_cast_if_autocast_enabledr18sC  $ $ & & // < yy&&,,T MMc n^\rSrSrS S\\\-\-S\S\S\4U4Sjjjr S\ S\ 4S jr S r U=r $) NemotronLayerNorm1P@normalized_shapeepselementwise_affinebiasc(>[TU]XX4XV5 gr&)super__init__)selfr6r7r8r9devicedtype __class__s r0r<NemotronLayerNorm1P.__init__As )0B&Xr2inputreturncrURRS:waURROSn[X!URURS-UR UR 5n[URRSS9 [R"U6sSSS5 $!,(df  g=f)Nmpscpur#Fr-enabled) r>typer1r6weightr9r7r F layer_norm)r=rBr-r.s r0forwardNemotronLayerNorm1P.forwardLs+0<<+<+<+Eell''5 (  5 5t{{Q SWS[S[  (9(95 I<<&J I Is B(( B6)gh㈵>TTNN)__name__ __module__ __qualname____firstlineno__intlistrfloatboolr<rrM__static_attributes__ __classcell__r@s@r0r4r4@so#' YS /D0 Y Y! Y  Y Y'V'''r2r4c ^\rSrSr%\R \S'SS\4U4Sjjjr\ SS\S-S\ SS\ S-S \ S \ 44S jj5r\R"5\S 55rS rU=r$)NemotronRotaryEmbeddingVinv_freqNconfigc>[TU]5 URUlURUlXlUR R SUlURnURS:wa[URnU"UR U5uo@l URSUSS9 URSUR5SS9 g)N rope_typedefaultr^F) persistentoriginal_inv_freq) r;r<max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenr_rope_parametersracompute_default_rope_parametersrattention_scalingregister_bufferclone)r=r_r> rope_init_fnr^r@s r0r< NemotronRotaryEmbedding.__init__Ys "("@"@$*$B$B! 44[A!%!E!E >>Y &.t~~>L+7 V+L(( ZeD 0(..2BuUr2r>z torch.deviceseq_lenrCz torch.Tensorc jURSnURRSS5n[USS5=(d URUR-n[ XT-5nSnSU[ R"SUS[ RS9RU[ RS 9U- -- nX4$) 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_thetapartial_rotary_factorg?head_dimNrr?)r>r?) rhgetgetattr hidden_sizenum_attention_headsrTr'arangeint64torV) r_r>robaserrrsdimattention_factorr^s r0ri7NemotronRotaryEmbedding.compute_default_rope_parametersis(%%l3 & 6 6 : :;RTW X6:t4h8J8JfNhNh8h(23 U\\!S!5;;?BB&X]XcXcBdgjj k ))r2cLURSSS2S4R5RURSSS5R 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#rErFFrGrtr~ru)r^rVexpandshaper|r> isinstancerIstrr transposer'catcosrjsinr?) r=x position_idsinv_freq_expandedposition_ids_expandedr-freqsembrrs r0rMNemotronRotaryEmbedding.forwardsN!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#)rjr_rfrgrar&)NNN)rPrQrRrSr'r__annotations__r$r< staticmethodrrTtuplerVrino_gradrrMrXrYrZs@r0r\r\VsllV~VV )-+/"*%*(*t* ~u$ % **> ]]_ <