qi{ddlmZddlmZddlZddlmcmZddlmZddl m Z ddl m Z ddlmZmZdd lmZdd lmZmZmZmZdd lmZdd lmZdd lmZmZddlm Z m!Z!ddl"m#Z#m$Z$ddl%m&Z&ddl'm(Z(m)Z)m*Z*ddl+m,Z,m-Z-ddl.m/Z/m0Z0ddl1m2Z2edGddejfZ4GddejfZ5GddejfZ6dZ7eddAdZ8d ejrd!e:d"ejrfd#Z; dBd$ejfd%ejrd&ejrd'ejrd(ejrdzd)eeGd/d0ejfZ?Gd1d2ejfZ@Gd3d4ejfZAGd5d6eZBe)Gd7d8e$ZCe)Gd9d:eCZD dCd;ejreEejrzdzdd?eCeZGgd@ZHy)D)Callable)OptionalN)nn)initialization)ACT2FN)Cache DynamicCache)GenerationMixin)use_experts_implementationuse_kernel_forward_from_hubuse_kernel_func_from_hubuse_kernelized_func)create_causal_mask)GradientCheckpointingLayer)MoeCausalLMOutputWithPastMoeModelOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tuple)maybe_autocastmerge_with_config_defaults)OutputRecordercapture_outputs) OlmoeConfigRMSNormc`eZdZddfd Zdej dej fdZdZxZS) OlmoeRMSNormreturnct|tjt j ||_||_y)z; OlmoeRMSNorm is equivalent to T5LayerNorm N)super__init__r Parametertorchonesweightvariance_epsilon)self hidden_sizeeps __class__s Z/opt/pipecat/venv/lib/python3.12/site-packages/transformers/models/olmoe/modeling_olmoe.pyr(zOlmoeRMSNorm.__init__2s1 ll5::k#:; # hidden_statesc"|j}|jtj}|j dj dd}|tj ||jzz}|j|j|zS)NT)keepdim) dtypetor*float32powmeanrsqrtr-r,)r.r4 input_dtypevariances r2forwardzOlmoeRMSNorm.forward:sy#)) %((7  $$Q',,R,>% Ht?T?T4T(UU {{]--k:::r3c^t|jjd|jS)Nz, eps=)tupler,shaper-)r.s r2 extra_reprzOlmoeRMSNorm.extra_reprAs* ))*+6$2G2G1HIIr3)gh㈵>)r%N) __name__ __module__ __qualname__r(r*TensorrArE __classcell__r1s@r2r$r$0s)$;U\\;ell;Jr3r$c eZdZUejed<d deffd Ze d dedzde dde dzde d e ffd Z ejed ZxZS)OlmoeRotaryEmbeddinginv_freqNconfigct||j|_|j|_||_|j j d|_|j}|jdk7rt|j}||j |\}|_ |jd|d|jd|jdy)N rope_typedefaultrNF) persistentoriginal_inv_freq) r'r(max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenrOrope_parametersrQcompute_default_rope_parametersrattention_scalingregister_bufferclone)r.rOdevice rope_init_fnrNr1s r2r(zOlmoeRotaryEmbedding.__init__Hs "("@"@$*$B$B! 44[A!%!E!E >>Y &.t~~>L+7 V+L($( ZeD 0(..2BuUr3r]z torch.deviceseq_lenr%z torch.Tensorc|jd}t|ddxs|j|jz}d}d|t j d|dtj j|tj|z zz }||fS) a Computes the inverse frequencies according to the original RoPE implementation Args: config ([`~transformers.PreTrainedConfig`]): The model configuration. device (`torch.device`): The device to use for initialization of the inverse frequencies. seq_len (`int`, *optional*): The current sequence length. Unused for this type of RoPE. Returns: Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE). rope_thetahead_dimNg?rr6r9)r]r9) rXgetattrr/num_attention_headsr*arangeint64r:float)rOr]r_basedimattention_factorrNs r2rYz4OlmoeRotaryEmbedding.compute_default_rope_parametersXs&%%l3fj$/c63E3EIcIc3c U\\!S!5;;?BB&X]XcXcBdgjj k )))r3cN|jddddfjj|jdddj |j }|dddddfj}t |j jtr/|j jdk7r|j jnd}t|d5|j|jzjdd}tj||fd }|j|jz}|j|jz} dddj |j  j |j fS#1swYAxYw) Nrr7r mpscpuF) device_typeenabledr6rjrc)rNrhexpandrDr:r] isinstancetypestrr transposer*catcosrZsinr9) r.x position_idsinv_freq_expandedposition_ids_expandedrofreqsembrxrys r2rAzOlmoeRotaryEmbedding.forwardvsR!MM$4-8>>@GG HZHZ[\H]_acdehhijiqiqr ,QaZ 8 > > @'1!((--'E!((--[`J`ahhmmfk  U C 5&,,.1F1L1L1NNYYZ[]^_E))UEN3C'')d444C'')d444C  5 vvAGGv$cff177f&;;;  5 5s BFF$N)NNN)rFrGrHr*rI__annotations__r!r( staticmethodrintrCrhrYno_gradrrArJrKs@r2rMrMEsllV{V %)+/"*d"*(*t* ~u$ % **:U]]_ <