+ QV-j2^RIt^RIt^RIHtHt^RIHt!RR]4t!RR]4t !RR ] 4t !R R ] 4t !R R ]4t !RR]4t R#)N)DatasetIterableDataset) ModelOutputc2a]tRt^toRtRtRtRtVtR#)PipelineDatasetc*WnW nW0nR#Ndatasetprocessparams)selfr r r s&&&&p/Volumes/fast/ai/experiments/ui-tars-smoke/.venv/lib/python3.14/site-packages/transformers/pipelines/pt_utils.py__init__PipelineDataset.__init__ s   c,\VP4#r lenr rs&r__len__PipelineDataset.__len__4<<  rcfVPV,pVP!V3/VPBpV#r r )riitem processeds&& r __getitem__PipelineDataset.__getitem__s,||ALL55 r)r r r N __name__ __module__ __qualname____firstlineno__rrr__static_attributes____classdictcell__ __classdict__s@rrrs !rrcBa]tRt^toRRltRtRtRtRtRt Vt R#) PipelineIteratorNc dWnW nW0nV^8XdRpW@nRVnRVnR#)a Roughly equivalent to ``` for item in loader: yield infer(item, **params) ``` Arguments: loader (`torch.utils.data.DataLoader` or `Iterable`): The iterator that will be used to apply `infer` on. infer (any function): The function to apply of each element of `loader`. params (`dict`): The parameters passed to `infer` along with every item loader_batch_size (`int`, *optional*): If specified, the items of `loader` are supposed to come as batch, and are loader_batched here making it roughly behave as ``` for items in loader: for i in loader_batch_size: item = items[i] yield infer(item, **params) ```N)loaderinferr loader_batch_size_loader_batch_index_loader_batch_data)rr,r-r r.s&&&&&rrPipelineIterator.__init__s96    ! $ !2$( "&rc,\VP4#r )rr,rs&rrPipelineIterator.__len__?s4;;rc:\VP4VnV#r iterr,iteratorrs&r__iter__PipelineIterator.__iter__BT[[)  rc a\SP\P4'd/SPSP,P ^4pEM/pSPP 4EFwr4\V\4'dVP4p\V^,\P4'd6\;QJd.V3RlV4FNK 5M!V3RlV44W#&M[\V^,\P4'd5\;QJd.V3RlV4FNK 5M!V3RlV44W#&KVR9d\V\4'd\V^,\P4'd6\;QJd.V3RlV4FNK 5M!V3RlV44W#&M[\V^,\P4'd5\;QJd.V3RlV4FNK 5M!V3RlV44W#&EKVR8XdEKVfRW#&EK\VSP,\P4'd(VSP,P ^4W#&EK)\VSP,\P4'd.\P!VSP,^4W#&EKVSP,W#&EK SPPV4pS;P^, unV#)zU Return item located at `loader_batch_index` within the current `loader_batch_data`. c3f<"TF&qSP,P^4xK( R#5irNr/ unsqueeze.0elrs& r 5PipelineIterator.loader_batch_item..U,1nfm`bT5M5M2N2X2XYZ2[2[fm.1c3t<"TF-p\P!VSP,^4xK/ R#5ir=np expand_dimsr/r@s& rrCrDW.1tlsfh"..DD\D\A]_`2a2als58c3f<"TF&qSP,P^4xK( R#5ir=r>r@s& rrCrD\rErFc3t<"TF-p\P!VSP,^4xK/ R#5ir=rHr@s& rrCrD^rKrLpast_key_valuesN> attentions hidden_states) isinstancer0torchTensorr/r?itemsrto_tupletuplerIndarrayrJ __class__)rresultloader_batchedkelementsf rloader_batch_item"PipelineIterator.loader_batch_itemFsO d--u|| < <,,T-E-EFPPQRSF N"55;;= g{33%..0G!'!*ell;;,1E1nfm1nEE1nfm1n,n)#GAJ ;;,1E1tls1tEE1tls1t,t)77JwPUF,,66~FF   A%  rcVPe,VPVP8dVP4#\VP4pVP !V3/VP BpVPEe\V\P4'dTpMH\V\4'd V^,pM(\VP44^,pW$,p\V\4'd \V4pMVP^,p^Tu;8dVP8d MMWPn\V\4'd V^,MTVn^VnVP4#V#r )r/r.r^nextr7r-r rRrSrTrWlistkeysrshaper0)rrr first_tensorkeyobserved_batch_sizes& r__next__PipelineIterator.__next__ws-  # # /D4L4LtOeOe4e))+ +DMM"JJt3t{{3  ! ! -)U\\22( Iu--(| 9>>+,Q/(~ ,--&),&7#&2&8&8&;#&?)?)??*=&6@E6R6RilXaD #'(D $))+ + r)r0r/r-r7r,r.r r ) r!r"r#r$rrr8r^rhr%r&r's@rr*r*s%%'N /b""rr*cBaa]tRt^toRV3RlltRtRtRtVtV;t #)PipelineChunkIteratorc (<\SV`WV4R#)a Roughly equivalent to ``` for iterator in loader: for item in iterator: yield infer(item, **params) ``` Arguments: loader (`torch.utils.data.DataLoader` or `Iterable`): The iterator that will be used to apply `infer` on. infer (any function): The function to apply of each element of `loader`. params (`dict`): The parameters passed to `infer` along with every item N)superr)rr,r-r r.rYs&&&&&rrPipelineChunkIterator.__init__s$ /rcH\VP4VnRVnV#r )r6r,r7 subiteratorrs&rr8PipelineChunkIterator.__iter__sT[[)  rcrVPf7VP!\VP43/VPBVn\VP4pV# \ dOTP!\TP43/TPBTn\TP4pT#i;ir )rpr-rar7r StopIteration)rrs& rrhPipelineChunkIterator.__next__s    # \#zz$t}}*=MMD  /T--.I / $zz$t}}*=MMD T--.I /sAAB65B6)r7rpr ) r!r"r#r$rr8rhr%r& __classcell__)rYr(s@@rrkrks0( rrkc0a]tRt^toRtRtRtRtVtR#)PipelinePackIteratorah Roughly equivalent to ``` packed = [] for item in loader: packed.append(item) if item["is_last"]: yield packed packed = [] ``` but it also handles cases where `item` are batched (meaning it's a dict of Tensor with first dimension > 1. In that case it does ``` packed = [] for batch in loader: # item is batched for item in batch: packed.append(item) if item["is_last"]: yield packed packed = [] ``` Arguments: loader (`torch.utils.data.DataLoader` or `Iterable`): The iterator that will be used to apply `infer` on. infer (any function): The function to apply of each element of `loader`. params (`dict`): The parameters passed to `infer` along with every item loader_batch_size (`int`, *optional*): If specified, the items of `loader` are supposed to come as batch, and are loader_batched here making it roughly behave as ``` for items in loader: for i in loader_batch_size: item = items[i] yield infer(item, **params) ```c:\VP4VnV#r r5rs&rr8PipelinePackIterator.__iter__r:rcRp.pVPeuVPVP8dZVPVP8d?VP4pVPR4pVP V4V'gKWV#V'EgrVP !\ VP43/VPBpVPEe \V\P4'dTpM(\VP44^,pWF,p\V\4'd \V4pMVP^,p^Tu;8dVP8d MMWpnW@n^VnVPVP8d?VP4pVPR4pVP V4V'gKWV#EKSTpVPR4pVP V4EKzV#)Fis_last)r/r.r^popappendr-rar7r rRrSrTrbrcrrdr0)rr{ accumulatorrrrerfrgs& rrhPipelinePackIterator.__next__s  # # /D4L4LtOeOe4e**T-C-CC--/((9-""4(7&&' 4 #6F$++FI%%1i66#,Ly~~/03C#,>LlD11*-l*;'*6*<*a]tRtRtoV3RlRltRtRtRtVtR#) KeyDataseti-c&<V^8dQhRS[RS[/#)r rfrstr)formatr(s"r __annotate__KeyDataset.__annotate__.scrcWnW nR#r r rf)rr rfs&&&rrKeyDataset.__init__.s  rc,\VP4#r rrs&rrKeyDataset.__len__2rrcJVPV,VP,#r rrrs&&rrKeyDataset.__getitem__5s||Atxx((rrNr r's@rrr-s!))rrc>a]tRtRtoV3RlRltRtRtRtVtR#)KeyPairDataseti9c,<V^8dQhRS[RS[RS[/#)rr key1key2r)rr(s"rrKeyPairDataset.__annotate__:s"s#rc*WnW nW0nR#r r rr)rr rrs&&&&rrKeyPairDataset.__init__:s   rc,\VP4#r rrs&rrKeyPairDataset.__len__?rrcRVPV,VP,RVPV,VP,/#)text text_pairrrs&&rrKeyPairDataset.__getitem__Bs6 Q 2KaQUQZQZA[\\rrNr r's@rrr9s  !]]rr)numpyrIrStorch.utils.datarr utils.genericrrr*rkrwrrrrrsd 5' g BBJ*,*Za+aH ) ) ]W ]r