^j`XddlmZddlmZddlZddlmZddlmZddl m Z m Z ddl m Z ddlmZmZdd lmZdd lmZmZdd lmZmZdd lmZmZdd lmZddlmZm Z m!Z!ddl"m#Z#m$Z$ddl%m&Z&ddl'm(Z(GddejRZ*GddejRZ+dejXde-dejXfdZ. d3dejRdejXdejXdejXdejXdzd e/d!e/d"eefd#Z0d$Z1d4d%Z2Gd&d'ejRZ3Gd(d)ejRZ4Gd*d+eZ5e Gd,d-eZ6e Gd.d/e6Z7e Gd0d1e6e Z8gd2Z9y)5)Callable)OptionalN)ACT2FN)Cache DynamicCache)GenerationMixin)create_causal_mask!create_sliding_window_causal_mask)GradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tuple)maybe_autocastmerge_with_config_defaults)capture_outputs) Cohere2Configc eZdZUejed<d deffd Ze d dedzde dde dzde d e ffd Z ejed ZxZS)Cohere2RotaryEmbeddinginv_freqNconfigct||j|_|j|_||_|j j d|_|j}|jdk7rt|j}||j |\}|_ |jd|d|jd|jdy)N rope_typedefaultrF) persistentoriginal_inv_freq) super__init__max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenrrope_parametersr!compute_default_rope_parametersrattention_scalingregister_bufferclone)selfrdevice rope_init_fnr __class__s w/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/cohere2/modeling_cohere2.pyr&zCohere2RotaryEmbedding.__init__-s "("@"@$*$B$B! 44[A!%!E!E >>Y &.t~~>L+7 V+L($( ZeD 0(..2BuUr0z torch.deviceseq_lenreturnz torch.Tensorc|jd}t|ddxs|j|jz}d}d|t j d|dtj j|tj|z zz }||fS) a Computes the inverse frequencies according to the original RoPE implementation Args: config ([`~transformers.PreTrainedConfig`]): The model configuration. device (`torch.device`): The device to use for initialization of the inverse frequencies. seq_len (`int`, *optional*): The current sequence length. Unused for this type of RoPE. Returns: Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE). rope_thetahead_dimNg?rdtype)r0r<) r*getattr hidden_sizenum_attention_headstorcharangeint64tofloat)rr0r5basedimattention_factorrs r3r+z6Cohere2RotaryEmbedding.compute_default_rope_parameters=s&%%l3fj$/c63E3EIcIc3c U\\!S!5;;?BB&X]XcXcBdgjj k )))r4c|jddddfjj|jddd}|dddddfj}t |j j tr/|j j dk7r|j j nd}t|d5|j|jzjdd}tj|dd }|j|jz}|j|jz} dddj|j  j|j fS#1swYAxYw) NrrmpscpuF) device_typeenabledr:rFr;)rrDexpandshape isinstancer0typestrr transposer@repeat_interleavecosr,sinrCr<) r/x position_idsinv_freq_expandedposition_ids_expandedrLfreqsembrVrWs r3forwardzCohere2RotaryEmbedding.forward[s@!MM$4-8>>@GG HZHZ[\H]_acde ,QaZ 8 > > @'1!((--'E!((--[`J`ahhmmfk  U C 5&,,.1F1L1L1NNYYZ[]^_E))%;C'')d444C'')d444C  5 vvAGGv$cff177f&;;;  5 5s =BFF N)NNN)__name__ __module__ __qualname__r@Tensor__annotations__rr& staticmethodrinttuplerDr+no_gradrr^ __classcell__r2s@r3rr*sllV}V '++/"*$*(*t* ~u$ % **:U]]_ <