+ j`^RIt^RIHt^RIHt^RIt^RIHt^RIHuH t ^RI H t ^RIHt^RIHtHt^RIHt^RIHtHtHtHt^R IHt^R IHtHtHt^R I H!t!^R I"H#t#H$t$^R I%H&t&H't'^RI(H)t)^RI*H+t+H,t,H-t-H.t.H/t/^RI0H1t1^RI2H3t3^RI4H5t5H6t6H7t7H8t8]-!RR7]!RR]#444t9]-!RR7]!RR]+444t:]!R4!RR]Pv44t<!RR]Pv4t=RR lt>RRR!R"llt?!R#R$]Pv4t@!R%R&]Pv4tA]!R'R(]Pv44tB!R)R*]Pv4tC!R+R,]Pv4tD!R-R.]Pv4tER/tF]!R0R14RSR2R3ll4tG]!R4R14RSR5R6ll4tH]!]G]H.4!R7R8]Pv44tI!R9R:]!4tJ]-!R;R<]'44tK]-!R=R>]K44tL]-!R?R@]K]44tM!RARB]Pv4tN]-!RCRD]K44tO!RERF]Pv4tPRGRHltQRTRIRJlltR]-!RKRL]K44tS]-!RMR7!RNRO]K44tT]-!RMR7!RPRQ]K]44tU.RUOtVR#)VN)Callable) dataclass)initialization)ACT2FN)Cache DynamicCache)GenerationMixin)use_experts_implementationuse_kernel_forward_from_hub&use_kernel_func_from_hub_with_fallbackuse_kernelized_func)force_accelerate_hooks)create_causal_maskcreate_recurrent_attention_mask!create_sliding_window_causal_mask)GradientCheckpointingLayer)BaseModelOutputWithPastBaseModelOutputWithPooling)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack) ModelOutputTransformersKwargsauto_docstringcan_return_tupletorch_compilable_check)merge_with_config_defaults)capture_outputs)InklingAudioConfig InklingConfigInklingTextConfigInklingVisionConfigzL Base class for Inkling outputs, with hidden states and attentions.  custom_introc4a]tRt^4toRtRtV3RltRtVtR#)InklingModelOutputWithPasta image_hidden_states (`torch.FloatTensor`, *optional*): A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`. image_hidden_states of the model produced by the vision encoder and after projecting the last hidden state. NcH<V^8dQh/S[PR,;R&#)Nimage_hidden_states)torch FloatTensor)format __classdict__s"m/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/inkling/modeling_inkling.py __annotate__'InklingModelOutputWithPast.__annotate__4s**T18) __name__ __module__ __qualname____firstlineno____doc__r)__annotate_func____static_attributes____classdictcell__r-s@r.r&r&4s 59r1r&zS Base class for Inkling causal language model (or autoregressive) outputs. cHa]tRt^DtoRtRtRtRtRtRt Rt V3Rlt Rt Vt R#)InklingCausalLMOutputWithPasta loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Language modeling loss (for next-token prediction). logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.text_config.vocab_size)`): Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax). past_key_values (`Cache`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`): It is a [`~cache_utils.Cache`] instance. For more details, see our [kv cache guide](https://huggingface.co/docs/transformers/en/kv_cache). Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. image_hidden_states (`torch.FloatTensor`, *optional*): A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`. image_hidden_states of the model produced by the vision encoder after projecting last hidden state. Nc:<V^8dQh/S[PR,;R&S[PR,;R&S[R,;R&S[S[P,R,;R&S[S[P,R,;R&S[PR,;R&#)r(Nlosslogitspast_key_values hidden_states attentionsr))r*r+rtuple)r,r-s"r.r/*InklingCausalLMOutputWithPast.__annotate__Ds,   d ")-.    $+/0T\(12**+d2934e''(4/656**T187r1r2)r3r4r5r6r7r?r@rArBrCr)r8r9r:r;s@r.r=r=Ds2 &*D'+F$(O59M26J487r1r=RMSNormcZaa]tRt^btoRV3RlV3RllltV3RlRltRtRtVtV;t #)InklingRMSNormc$<V^8dQhRS[RR/#)r(epsreturnN)float)r,r-s"r.r/InklingRMSNorm.__annotate__ds$$$$$r1c <\SV`4\P!\P !V44VnW nR#)z- InklingRMSNorm is equivalent to T5LayerNorm N)super__init__nn Parameterr*onesweightvariance_epsilon)self hidden_sizerJ __class__s&&&r.rPInklingRMSNorm.__init__ds/ ll5::k#:; #r1cN<V^8dQhRS[PRS[P/#r(rBrKr*Tensor)r,r-s"r.r/rMls#;;U\\;ell;r1c:VPpVP\P4pVP ^4P RRR7pV\P !W0P,4,pVPVPV4,#)r(T)keepdim) dtypetor*float32powmeanrsqrtrUrT)rVrB input_dtypevariances&& r.forwardInklingRMSNorm.forwardlsw#)) %((7  $$Q',,R,>% H?T?T4T(UU {{]--k:::r1c^\VPP4 RVP 2#)z, eps=)rDrTshaperU)rVs&r. extra_reprInklingRMSNorm.extra_reprss* ))*+6$2G2G1HIIr1)rUrT)ư>) r3r4r5r6rPrirmr9r: __classcell__rXr-s@@r.rHrHbs%$$;;JJr1rHcTaa]tRt^wtoRtV3RlV3RlltV3RlRltRtVtV;t #)InklingRelativeLogitsalhidden states conditioned relative position bias. `proj` is a trained bank of bias-vs-distance profiles; each token's `relative_states` mixes them into one bias value per backward distance (`sglang RelLogitsProj` + the FA4 `score_mod`, materialized densely). The bias is zero outside `0 <= distance < rel_extent`; causality and padding stay in the attention mask. c&<V^8dQhRS[RS[/#)r(d_rel rel_extentint)r,r-s"r.r/"InklingRelativeLogits.__annotate__~sAAcAsAr1c<\SV`4W n\P!\ P !W44VnR#N)rOrPrvrQrRr*emptyproj)rVrurvrXs&&&r.rPInklingRelativeLogits.__init__~s, $LLU!?@ r1c<V^8dQhRS[PRS[PRS[PRS[P/#)r(relative_statesquery_positions key_positionsrKr\)r,r-s"r.r/rysI ^ ^ ^ ^|| ^  ^r1cWP,P^^4pVR,VR,, R,pVP^VP^, 4P!.VP R,ORNRN5!pVP RV4pVPV^8WPP8,R4#):Nr(N)NNNN)Nr)NNrrr`)r} transposeclamprvexpandrlgather masked_fill)rVrrr rel_logitsdistance gather_index position_biass&&&& r.riInklingRelativeLogits.forwards& 1<