o ki| @s: UddlZddlZddlZddlTddlZddlmmZddl m Z ddl m Z ddl mZgZeeed<iZeeefed<d d Zdd d Z  dddZdddZdeddfddZdededededdf ddZddZdd Zd!d"ZeeeZ de!dBfd#d$Z"d%d&Z#d'd(Z$d)d*Z%d+d,Z&e ej'j(j)j*ej'j(j+j*ej'j(j,j*ej'j(j-j*ej'j(j.j*ej'j(j/j*ej'j(j0j*gd-d.d/Z1e ej'j2j3j*d-d0d1Z4e ej'j(j5j*gd-dd2d3Z6e ej'j(j7j*d-d4d5Z8e ej'j(j7j9d6e8e ej'j(j:j*d6d7d8Z;e ej'j(jj*d<d=d>Z?e ej'j(j@j*d?d@dAZAe ej'j(jBjCdBdCdDZDe ej'j(jEj*d-dEdFZFe ej'j(jGj*dGdHdIZHe ej'j(jIj*d-e$e ej'j(jJj*ej'j(jKj*ej'j(jLj*ej'j(jMj*ej'j(jNj*gd-dJdKZOe ej'j(jPj*dLeOe ej'j(jQj*dMeOe ej'j(jQjRdNeOe ej'j(jSj*d-dOdPZTe ej'j(jUj*dQdRdSZVe ej'j(jWj*dQdTdUZXe ej'j(jYj*dVdWdXZZe ej'j(j[j*dYdZd[Z\e ej'j(j]j*d\d]d^Z^e ej'j(j_j`d_d`daZae ej'j(j_j*dbdcddZbe ej'j(jcjddedfdgZee ej'j(jfj*dhdidjZge ej'j(jhj*dkdldmZhe ej'j(jij*dndodpZje ej'j(jkjldqdrdsZme ej'j(jnj,dtdudvZoe ej'j(jpj*dwdxdyZqe ej'j(jrj*dzd{d|Zse ej'j(jtj*d}d~dZue ej'j(jvj*dddZwe ej'j(jxj*dddZye ej'j(jzj*dddZ{e ej'j(j|j*dddZ|e ej'j(j}j*dddZ}e ej'j(j~jdddZe ej'j(jj*dddZe ej'j(jj*dddZe ej'j(jj*dddZddZe ej'j(jj*dbddZe ej'j(jjdddZe ej'j(jjldddZe ej'j(jj*dddZe ej'j(jj*ej'j(jj*gdddZe ej'j(jj*dddZe ej'j(jj*dddZe ej'j(jjldddZe ej'j(jjddddZe ej'j(jj*de ej'j(jj*dddZe ej'j(jj*dddZe ej'j(jj,dddZe ej'j(jj*dbddZe ej'j(jjdddZe ej'j(jj,dƒddĄZe ej'j(jjdddƄZe ej'j(jj,dƒddȄZe ej'j(jj*ej'j(jj*ej'j(jj*ej'j(jj*gd-ddʄZe ej'j(jj*ej'j(jj*gd-dd̄Ze ej'j(jj,dƒdd΄Ze ej'j(jj,dƒddЄZe ej'j(jj*ddd҄Ze ej'j(jj*dddԄZe ej'j(jj*dddքZe ej'j(jj*ddd؄Ze ej'j(jj*dكddۄZe ej'j(jj*dzdd݄Ze ej'j(jj*dރddZe ej'j(jj*dddZe ej'j(jj*ej'j(jj*gd-ddZe ej'j(jj*d-ddZe ej'j(jj*dddZe ej'j(jj*dddZe ej'j(jj*dddZe ej'j(jj*d-ddZe ej'j(jj*d-ddZe ej'j(jj*d-ddZe ej'j(jj*d-ddZe ej'j(jj*d-ddZe ej'j(jj*dddZe ej'j(jj*dddZe ej'j(jj*ddddZe ej'j(jj*ej'j(jj*ej'j(jj*gdddZe ej'j(jj*ej'j(jj*ej'j(jj*ej'j(jj*ej'j(jj*ej'j(jj*gddd Ze ej'j(jjd d d Ze ej'j(jjd d dZe ej'j(jjdd-ddZe ej'j(jj*dddZe ej'j(jj*dddZejdd$Z(e(dede(dede(dedWddS1swYdS( N)*)normalize_function)#jagged_scaled_dot_product_attention NestedTensor__all__JAGGED_OPS_TABLEcCs>|jr|dkr t|jSt|jS|dkrtdStdS)Nmaxinf-inf)is_floating_pointtorchfinfor minfloat)dtype padding_typerd/var/www/addictedbytheproject.nl/epg/venv/lib/python3.10/site-packages/torch/nested/_internal/ops.py_get_padding_values  rFcsddlm}t|ttfr$t|fdd|D}t|t|S|r+||}|dkr3|ks=td|d|dkrEdS|dS)Nrcanonicalize_dimsc3|] }t|VqdSN_outer_to_inner_dim.0dndim ragged_dimrr #z&_outer_to_inner_dim..zdim z out of range for ndim r) torch._prims_commonr isinstancetuplelisttypedictfromkeysAssertionError)r!dimr" canonicalizeroutputrr rrs  rTc Csvddlm}|||}||kr|st|d|dkr%|s%t|d|r-t|||n|} |r9|dk} | | fS| S)Nrrz0(): not supported for NestedTensor on ragged dimz+(): not supported for NestedTensor on dim=0)r%r RuntimeErrorr) r!r-r"op_nameconvert_to_inner_dimallow_ragged_dimallow_batch_dimrwrappedretoperating_on_batchrrr_wrap_jagged_dim5s   r8c sddlmt|ttfstdt|fdd|D}d|v}|v}tfdd|D}tt fdd|D}||||fS) z For NestedTensor operators, wraps dimensions to non-negative values, and returns metadata related to reduction dimension(s). rrz<_wrap_jagged_dims(): cannot iterate over dimensions of type csg|]}|qSrrr)rr!rr _s z%_wrap_jagged_dims..c3s |] }|dko |kVqdS)rNrr) ragged_idxrrr#ez$_wrap_jagged_dims..c3rrrr)r!r:rrr#ir$) r%rr&r'r(r,r)anyr*r+) r!dimsr1r: wrapped_dimsoperate_on_batchoperate_on_raggedoperate_on_non_batchouter_to_inner_dimr)rr!r:r_wrap_jagged_dimsRs    rC schema_strreturnc s|d}dd|Dd}t||}|ddkr"|dd}n't||kr0t|t|ksItd|jd|d |d t|d t|d d dddddddd}t|D]i\}} | d\} } | d} | rt| ddn| } | |vrtd| |t|kr| std|jd|d| q\|| | ffdd }|||sdddddd}td|jd|d| d || q\dS)!N, cSsg|]}|dqS)?)endswithrxrrrr9qz check_schema..T... NestedTensor (z): expected at least z arguments and at most z arguments, but got: z argumentscSst|tjo t|t Sr)r&rTensorrrJrrrzcheck_schema..cSst|to|jduo|jdkSNr)r&r_lengths _ragged_idxrQrrrrRs  cSs t|tSrr&rrQrrrrRscSdS)NTrrQrrrrR)tjtjt_allr<z: rGzUnknown arg type: z) missing required argument: cs|r |dup |S|Srr)rJ is_optional _check_fnrrcheck_fnszcheck_schema..check_fntensorzoptional tensorz%contiguous jagged layout NestedTensorzjagged layout NestedTensorz )rZzt?r[r\r<z ): expected z to be a )splitcountlen ValueError__name__ enumeraterHr,)rDfuncargskwargsnamed_arg_typesnum_optional_argsmin_argsarg_type_check_fnsinamed_arg_typenamearg_typer]normalized_arg_typer` type_to_descrr^r check_schemaos`       ruaa_namebb_namecCs:|j|j|j|jkrtd|jd|d|ddS)NrNz : expected  and z' to have the same exact offsets tensor.)_sizerVr0rf)rhrvrwrxryrrrcheck_ragged_dim_sames r|cCsL|jd}|jd|}|d|}t|t|ko%tddt||DS)Nrcss$|] \}}||kp |dkVqdS)rLNr)rnssrrrr#"z%raggedness_matches..)rVr{rdallzip)ntsizeend nt_ragged size_raggedrrrraggedness_matchess  rcCsB|dkr|jddkr|d}|dkr|jddks |SNrr)r-shapesqueeze)rZrrrsqueeze_leading_oness rcs4ttsgttsgfdd}|S)Ncs2D]}fdd}D]}||||<q qS)Ncsfdd}|S)Ncs.tg|Ri|g|Ri|Sr)ru)rirj)aten_oprhrDrrinnersz@register_func..wrapper..get_inner..innerr)rr)rhrD)rr get_innersz1register_func..wrapper..get_innerr)rhrrtableaten_opsrDtablesrhrwrappers zregister_func..wrapper)r&r()rrrDrrrr register_funcs  rc Os$t|d}|dur |Stjj|jvrddlm}|D]}||r,td|j d|qt dd|D}|dkrzg}|j j D]}t |jtjrT||jdn ||jd q@|d d |}t||g|Ri|tt|S|d krtd |g|Ri|tt|SdS)Nr is_nested_intrNz: invalid argument css|] }t|tjVqdSr)r&rrPrIrrrr# r$z lookup_jagged..rz: jt_allz: anyrMrFzlhs: any, rhs: any, ...)r getrTag pointwisetags%torch.fx.experimental.symbolic_shapesrr0rfsum_schema argumentsr&r) TensorTypeappendrqjoinru functoolspartialjagged_unary_pointwisejagged_binary_pointwise) rhrirj dispatch_funcrargnum_tensor_args schema_partscheck_schema_strrrr lookup_jaggeds6       rcCs|||j|jd}|S)Noffsetslengths_metadata_cacherVr)rrjrrrextract_kwargs s rcs>tdd|Dt|fdd|Di|fitS)Ncs|] }t|tr|VqdSrrWrrrrrr#,z)jagged_unary_pointwise..c3s"|] }|ur |jn|VqdSr_valuesrnjtrrr#.s nextrrrhrirjrrrr*s rcsR|d|d}}t|tst|tstdd}t|trOt|trOt||jrCt||j|jg|ddRi|fit|St||j |j|jt|t}|rZt|nt|}|rd||fn||f\}| krvt dt |} | dkr|rj| fn| jf\} } t|| | g|ddRi|fi|S| | kr|j d|j dkrt||j |j |j ddlm} j} j}|jj jd}dur||fdd t|j D}| |jj|| |d }|r|fn|f\} } || | g|ddRi|St||j |j |j ) Nrrz4At least one of the arguments must be a NestedTensorzHcannot call binary pointwise function {} with inputs of shapes {} and {}rz+NYI: broadcasting NT with T with larger dimnested_from_paddedcs"g|] \}}|jkr |nqSrrV)rrorJr padded_max_Srrr9n"z+jagged_binary_pointwise..rr:sum_S min_seqlen max_seqlen)r&rr,rr{rrr0formatrfr-NotImplementedErrorrr nested_tensorr_maybe_min_seqlen_maybe_max_seqlenrVexpandrg_offsets)rhrirjrvrxmismatch_error_msga_is_ntextracted_kwargsrZ t_squeezedlhsrhsrrrtotal_Lt_as_ntrrrr3s` " ,  rcs|tjjjurt|i|S|jdkr)||djg|ddRi||dS|jdkrdddd}t|||d d \}}|d }t | |d |j dd d}t | |d|j dd d}||kre|St tj|j||d} g|jd|| |j|ddR} |j| S|jdkr|djrS\} }tj| D]} t| d} | dur| qWdS1swYS|jdkr|djrd S\} }| sd Stfdd| DS|jdkr@dddd}t|||d d \}}|d }|d}| |j d}t||kr#tdtj||i|WdS1s;wYt|)Napply_rrflattenrLrEcSdSrr)input start_dimend_dimrrr _flatten_sigz+jagged_torch_function.._flatten_sigTrirjnormalize_to_only_use_kwargsrrF)r2r share_memory_ is_sharedc3s,|]}t|duot|VqdSr)getattrr)rrqrrrr#s  z(jagged_torch_function..rms_normcSrrr)rnormalized_shapeweightepsrrr _rms_norm_sigrz,jagged_torch_function.._rms_norm_sigrzNrms_norm(): Normalization over the ragged dim not supported for nested tensors)rrLrENNN)r_C_nnscaled_dot_product_attentionrrfrrpopr8r-rVrreduceoperatormulrreshapeis_cuda__tensor_flatten__DisableTorchFunctionSubclassrrrrdrer)rhrirjr_ new_kwargsinprrproduct new_shapenamesrq componentrrmax_normalizablerrrjagged_torch_functions $   *               "rz self: jt_allcOs|tjjjjur dS|tjjjjur|djS|tjjjjur&t|djS|tjjj jtjjj jfvrW|dj durPt t |dj t|djddS|dj S|tjjjjurd|djS|tjjjjurs|djSdS)NFrr)ropsatenis_non_overlapping_and_densedefaultsym_sizer{r-rd sym_numelnumelrUintrmathprodr sym_stride_stridessym_storage_offsetstorage_offsetrrrrtensor_attr_supported_getters  * r cOstjSr)rjaggedrrrrprim_layout_default srcOs|tjjjjur tddS)NznNestedTensor does not support directly calling torch.ops.aten.size; please use `nested_tensor.size()` instead.)rrrrrr0rrrrtensor_attr_unsupported_getters rcOspddlm}t|||dd\}}|d}|durdS|dtj|d<|dtjkr/dS||j fi|S)Nr)is_contiguous_for_memory_formatTrrF memory_format) r%rrrrrrcontiguous_formatpreserve_formatr)rhrirjrrrrrrris_contiguous_generals   rz!self: jt_all, memory_format: any?cOslt|||dd\}}|d}|durdS|dtj|d<|dtjkr)dStjjj j |j fi|S)NTrrFr) rrrrrrrrrsym_is_contiguousrrrhrirjrrrrrrsym_is_contiguous_general5s  rz"input: jt_all, memory_format: any?c Ost|||dd\}}|d}t|}|jdur:|dtjkr:ddlm}|jdkr.t d|| dd\}}|St ||j fi|fi|S) NTrrrr)jagged_from_listz;NJT with ragged_idx != 1 not supported for contiguous clone)r) rrrrUrrrrrVr,unbindrr) rhrirjrrrnew_metarcontigrrr clone_defaultLs    rzinput: jt, weight: t, bias: t?cO@t|||dd\}}|d}t||jfi|fit|SNTrrrrrrrrrrrlinear_defaulti  "r!z6self: jt, grad_output: jt, weight: t, output_mask: anycOst|||dd\}}|d}|d}|d}|d}d\} } } t||d|d|d r>tt|j|fit|} |d r^|jd | d } |jd | d } t| | } |d rt t |jj d }|d krv|j} n tj|j|dd} | | | fS)NTrr grad_outputr output_mask)NNNselfrrrLrrF)keepdim)rrr|rrmatmulrrrrrZr'ranger!cloner)rhrirjrrrr#rr$dsdwdbgrad_2dinput_2d reduce_dimsrrrlinear_backward_defaultts2        r0zinput: jt_all, dtype: anycOrrr rrrrto_dtyper"r1cOsddlm}t|||dd\}}|d}|d||jfi|}|jj|jd}d} |jdur:|jj|jd} dd l m } dd l m } m } |jdurP|jn|j} | durY|n| }t|| | fro| |}| | }|j|_n|| ||<t|}||d <| |d <t|fi|}|S) Nr_tensor_symint_registryTrrlayoutdevicer FakeTensorFunctionalTensormb_unwrap_functional_tensorrr)rr3rrrrtor6rUtorch._subclasses.fake_tensorr8#torch._subclasses.functional_tensorr:r;r&nested_int_memorr)rhrirjr3rrr new_values new_offsets new_lengthsr8r:r; ragged_source new_thingtgtsrc inp_kwargsr/rrrto_copy_defaults2       rHz-self: jt_all, src: jt_all, non_blocking: any?c Ost|||dd\}}|d}|d}|j|jkrO|}dd|D}|} dd| D} || kr@td|jd |jt|| D] \} } | | qE|j|j|S) NTrrrFcSg|]}|jqSrrrcrrrr9z copy_default..cSrIrrJrKrrrr9rMzr:r;rr3r&r?r)rhrirjrrrr@rArB output_kwargsr8r:r;r3rCrDrErFrrrlike_factory_defaults6         rXzself: jt_all, fill_value: anyzself: jt_all, high: anyz!self: jt_all, low: any, high: anycOs,t|||dd\}}|d}||j|Srrrrrrrr zero__default>s   rZz*self: jt_all, dim: any, half_to_float: anyc OsZt|||dd\}}t|dtrtd|d}t||dfd|j\|d<}}}|r3td|r>|jdkr>td |rI|jdurItd |dd |d<|rt j j j t j jj|j|jjd d |jg|jgtd d|jd} t j jj| |jg|jjd djd g|jjddR} t| fit|St||jfi|fit|S)NTrr-zHsoftmax(): not supported for dimensions of type 'tuple' for NestedTensorrsoftmaxzRsoftmax(): not supported when reducing across the batch dimension for NestedTensorrzesoftmax(): not supported when reducing along the ragged dimension for ragged_idx > 1 for NestedTensorzksoftmax(): not supported where lengths is not None if reducing across the ragged dimension for NestedTensorrrLr ) max_lengths padding_valuer-r)rr&r'r0rrCr-rVrUrnn functionalr[rr_jagged_to_padded_dense_forwardrrrr _max_seqlenr_padded_dense_to_jagged_forwardrr) rhrirjrrrreduce_on_batchreduce_on_ragged_reduce_on_non_batchpadded_softmax_valuessoftmax_valuesrrr_softmax_defaultIsx     "rjc Ost|||dd\}}t|dtrtd|d}t||dfd|j\|d<}}}|r3td|r9td|dd |d<t||j fi|fit |S) NTrr-zLlog_softmax(): not supported for dimensions of type 'tuple' for NestedTensorr log_softmaxzVlog_softmax(): not supported when reducing across the batch dimension for NestedTensorzVlog_softmax(): not supported when reducing along the ragged dimension for NestedTensorr) rr&r'r0rrCr-rVrrr) rhrirjrrrrerfrgrrr_log_softmax_defaults4  "rlz7grad_output: jt, output: jt, dim: any, input_dtype: anycONt|||dd\}}|d}|d}t||j|jfi|fit|S)NTrr#r/r )rhrirjrrgrad_outr/rrr_softmax_backwards  roz!self: jt, float: any, train: any?cOs\t|||dd\}}|d}||jfi|\}}t|fit|t|fit|fSrrrrrr)rhrirjrrrout1out2rrrnative_dropout_defaults rsz%grad_output: jt, mask: jt, scale: anycOrm)NTrr#maskr )rhrirjrrr#rtrrrnative_dropout_backward_defaults  ruz2self: jt_all, dim: any, keepdim: any?, dtype: any?cOt|ddg|Ri|S)Nrr_apply_reductionrrrr prod_dim_intryzself: jt_all, dtype: any?cO0t|||dd\}}|d}||jfi|SrrYrrrr prod_default  r|z$self: jt, split_size: any, dim: any?cs^t|||dd\}}|dt|djd|d<tfdd|jfi|DS)NTrrr-rbc3s&|]}tdd|itVqdS)valuesNrrrrIrrrr# s  zsplit_tensor..)rrr8r-rVr'rrhrirjrrrrr split_tensors  rz%self: jt, split_sizes: any, dim: any?csZt|||dd\}}|dt|djd|d<fdd|jfi|DS)NTrrr-split_with_sizesc"g|] }tdd|itqSr~rrrIrrrr9 z,split_with_sizes_default..)rrr8r-rVrrrrrsplit_with_sizes_defaults  rz+self: jt, dim: any, start: any, length: anycOsdt|||dd\}}|d}t||d|jd}||j||d|dd}t|fit|S) NTrrr-narrowstartlengthr-rr)rrr8r-rVrrr)rhrirjrrrr-r~rrrr&s rz self: jt, chunks: any, dim: any?c st|||dd\}}|dt|djddd\|d<}|rc|d}j}||}dd |D} d d | D} fd d | Dd d |D} j | fd d t t DSfdd |jfi|DS)NTrrr-chunkr4chunkscSsg|] }tj|ddqS)rr^)rcumsumrIrrrr9Ksz!chunk_default..cSsg|] }tj|dddqS)rrr)value)FpadrIrrrr9Lscsg|]}|jdqS))rrVr)r per_offsetsrrrr9Ms cSsg|]}|qSr)ritemrIrrrr9SrScs&g|]}tdd|i|qSrrrro) chunk_values nested_kwargsrrr9XscrrrrIrrrr9]r) rrr8r-rVrdiffrrrbr(rd) rhrirjrrr7rrchunked_lengthschunked_offsets split_sizesr)rrrr chunk_default9s0        rzself: jt_all, dim: any?c st|||dd\}}|d}|dkrtd|d}||}|}|jddttdttdBd dffd d } |durW| | t j d d Sdkr_td| | | fddt |jdDS)NTrr-rz2unbind(): only supported for NestedTensor on dim=0rrUrrEcsd}jd}tt|D].}t||dkt|||k|||7}|dur=t|||||kddqt||k|durdtt|D]}t||dkt|||kqOdSdS)NrrcSrX)NzMunbind(): nested tensor offsets and lengths do not match ragged_idx dimensionrrrrrrRrYz2unbind_int.._torch_check..)rr(rdr_check)rUr lengths_sumragged_dim_sizero)r:r~rr _torch_checkts&  z unbind_int.._torch_checkrr^zAunbind(): nested tensor ragged_idx out of bounds (should be >= 1)cs*g|]}tjd||dqS)rr)rrrlengths_scalarsoffsets_scalarsr:r~rrr9szunbind_int..r)rr0rr~rrrVr(rrtolistrrbr(r) rhrirjrrr-rrrrrrr unbind_intcs4  *   rzself: jt, dim: anycOsbt|||dd\}}|d}|j}tt|j|d|jd|d<t||fi|fit|S)NTrrr-r) rrrr8rdr{rVrr)rhrirjrrrr~rrr squeeze_dims  rzself: jt_all, dim: anyc Ost|||dd\}}|d}|j}|d}tt|jd||jddd|d<t|}|d|jdkr<|dd7<t||fi|fi|S) NTrrr-r unsqueeze)r3rV) rrrr8rdr{rVrr) rhrirjrrrr~r-rWrrrunsqueeze_defaults  rztensors: any, dim: any?cst|||dd\}}|d}dd|D}t|dkr td|dfdd|D}|d }ttj|jd |d <t|d d|Dfi|fit|dS) NTrtensorscSsg|]}|jr|qSr) is_nestedrrZrrrr9rKzcat_default..rz"At least one tensor must be nestedcs g|] }|jr |n|qSr)r expand_asrfirstrrr9s r-catcSrIrrrrrrr9rM) rrrdr,r8rrVrr)rhrirjrrrnestedr-rrr cat_defaults$   rzself: any, other: anyc s(t||dd\}}|d}|d}fdd}fdd}|jrl|jsl|d krB||krB|j|d krB|||S|d krk||krk|j|d krkt|j|fi|fit|Sn|js|jr|d kr||kr|jd kr|||S|d kr||kr|jd krt||jfi|fit|SnX|jr|jr|d kr|d krt||j rt|j|jfit|S|d kr|d kr|jd kr|jd kr| |j| |jkrt |||St d |j d |j)NTrrothercs fddt||DS)Ncsg|] \}}||qSrr)ra_compb_comprrrr9sz8matmul_default.._unbind_impl..)rr)rvrxrrr _unbind_impls z$matmul_default.._unbind_implc s|jr|}n|}ddlm}|j}|j}|}|jj|jd}|dur%|}g|jd|j||j|jddR}|jd|d} |jrL| |} n|| } || |j |j|||dS)Nrr output_sizer) rrrrrrrrVto_padded_tensorr) rvrxrrrrrr padded_shape padded_ntpadded_trrr _padded_impls:   z$matmul_default.._padded_implrrz1matmul(): not supported between inputs of shapes rz)rrrr-rVrrrrr{rrstackr0r) rhrirjrrrrrrrrrmatmul_defaultsV     %    & &$  rzself: jt_all, mat2: anycOsdt|||dd\}}|d}|d}|dkrtd|dkr(tdttjjjj ||S)NTrrmat2rzbmm(): input must be 3Dzbmm(): mat2 must be 3D) rrr-rerrrrr'rrhrirjrrrrrrr bmm_defaultZs    rz'self: jt_all, size: any, implicit: any?cst|||dd\}}|d|dd|vr |dr tdts0tdjdfd d td D}t|j |fit S) NTrrrimplicitz implicit expand is not supportedzexpand(): cannot expand shape z -> cs"g|] }|jkr dn|qS)rLrrrrrrr9{rz"expand_default..r) rrr,rr0r{r(r-rrr)rhrirjrr expand_argrrrexpand_defaultks  rzself: t, other: jtcOsDt|||dd\}}|d}|d}t|||jfit|S)NTrrrr rrrrexpand_as_defaults   rzself: jt_all, size: anycOslt|||dd\}}|d}|d}t||kr2|gddt|t|D|Std)NTrrrcss|]}dVqdS)rNr)rrrrrr#szbroadcast_to..zwbroadcast_to(): broadcasting to a higher-dim shape is currently not supported for nested tensors with the jagged layout)rrrdr-rr(re)rhrirjrrrrrrr broadcast_tos  ,rz tensors: anyc Ost|||dd\}}|d}t|dkrtdt|dkr#|dSg}tjdd|D}td d|D}|D],} | jrH|| |q:| t|krc|t | |j j fit|q:td t|S) NTrrrz7broadcast_tensors(): expected at least one tensor inputrcss|]}|jVqdSrrJrrrrr#sz$broadcast_tensors..csrrrWrrrrr#rzubroadcast_tensors(): broadcasting nested tensors with dense tensors of equal or higher dim is not currently supported)rrrdrerbroadcast_shapesrrrrr-rrrrr') rhrirjrrroutsbroadcast_shaperrZrrrbroadcast_tensorss,   rz(condition: jt_all, self: any, other: anycOspt|||dd\}}|d}|d}|d}t|||\}}}t||j|j|jfi|fit|S)NTr conditionrr)rrrrrrr)rhrirjrrrrrrrr where_selfs   rzself: jt, device: any?cOrrr rrrr_pin_memory_defaultr"rcOr{rrYrrrris_pinned_defaultr}rzself: jt_all, other: jt_allcOs|dj|djkSr)r{rrrris_same_size_defaultsrcsDt|||dd\}}|d|ddup%t|dttfo%t|ddk}|rU|jfi|}|ddrSt|ttfrMt|fdd |D}|S| j }|S|d} t|dttf} | sg| g} t | |j \} } } }| s|| d} | |d<| rj durt|d dd lm}| r| r|jfi|}|ddr|d d |}|S|rt|d|d| rĈj gnj }||fd|i|S| rt|d|jfi|}t|ddst|dttfr|dn|dg}|D]}|j dkrdd8<q|fdd |S)NTrrr-rr&Fc3s|] }|jVqdSr)rrV)rorrrr#r$z#_apply_reduction..zf(): reducing across the ragged dimension is not supported for non-contiguous nested tensors with holes)tree_mapcSs |dS)Nr)rrrrrrR+s z"_apply_reduction..zW(): reducing along a ragged and non-batch dimension is not supported for nested tensorszg(): reducing along the batch dimension but not the ragged dimension is not supported for nested tensorsrrVcst|fiSrrr) out_kwargsrrrRVs)rrr&r'r(rdrrr)rrVrCr-rUr0torch.utils._pytreerrr)rh func_nameidentity_elementrirjrrfull_reductionoutdim_to_convert is_dimlist converted_dimrerfreduce_on_non_batchr dim_to_passdimlistrr)rrrrxs         rxcOr{rrYrrrr sum_defaultYr}rz3self: jt_all, dim: any?, keepdim: any?, dtype: any?cOrv)Nrrrwrrrrsum_dim_IntListdrzrz"self: jt_all, dim0: any, dim1: anyc Os.t|||dd\}}ddlm}|d}|||d|df\}}||jks.||jkrj|dks6|dkr:td||jkrB|} n|} t|} | | d <t| t t |j ||jt t |j ||jfi| St||d|jd |d<t||d|jd |d<t||jfi|fit|S) NTrrrrdim0dim1z?Transpose is not supported on the batch dimension for jagged NTrV transpose)rr%rrr-rVrerrr~rrrdr{r8r) rhrirjrrrrrrto_dimrGrrr transpose_intls>   "rzself: jt_all, dims: anyc st|||dd\}}|d|d}t}tjt|kr3tdddt|dd d lm}||}t|tt|krLtd j durUtd |d d kr_td | j |d<fdd|ddD} | |d<t |j fi|fi|S)NTrrr=z5permute(): number of dimensions in the tensor input (z) zAdoes not match the length of the desired ordering of dimensions ().rrz*permute(): duplicate dims are not allowed.zBpermute(): not supported on jagged layout nested tensor with holesz=Permute is not supported on the batch dimension for jagged NTrVcsg|] }t|jqSr)rrV)rr-rinp_dimrrr9s z#permute_default..r)rrrrdr{rer%rsetrUindexrVrr) rhrirjrrr=rGrcanonicalized_dims inner_dimsrrrpermute_defaults>           rcst|||dd\}}|d|djdkr.tjtkr.tdjddtd ks9tsDtd jd fd d fddttdD}t  t |j |fitWdS1szwYdS)NTrrrrzVview(): does not support ragged_idx != 1 except when inp._size == size. inp._size is (z) and size is (rrzview(): cannot view shape z as cs&|jdkr j|S|dSrT)rVrr) inner_idxrrrget_inner_sizes  z$view_default..get_inner_sizecsg|]}|qSrrr)rrrr9sz view_default..)rrrVr'r{r0rdrr(rinference_mode is_inferencerrr)rhrirjrr inner_sizer)rrrr view_defaults(  $rzHinput: jt_all, normalized_shape: any, weight: any?, bias: any?, eps: anycOst|||dd\}}|d}|dkrtd|d}|j|j}|t|}|dkr3td||vr@|jdur@td ||vrtj j j |j j |jd |jg|jgd } tj j j tj|j jdd f|j|jd |jg|jgd | j} |jd d | jd} tj| ddd| } | | | } tjt| ddd| }t||d}| |}tj j j||jg|j jddd|j|jd d}t|fit|| |fS||j fi|\}} }t|fit|| |fS)NTrrrzOlayer_norm(): not supported for NestedTensor objects with 2 or fewer dimensionsrrzVlayer_norm(): not supported when normalizing over the batch dimension for NestedTensorzklayer_norm(): not supported where lengths is not None if operating on the ragged dimension for NestedTensorr)r\r)r6r)rr)r-r&rr_rL)rrr-r0rrVrdrUrrrrbrrrrconesr6rrrrrsquaresqrtrd unflattenrr)rhrirjrrrr ragged_sizenum_dims_not_normalized padded_input padded_maskragged_lengthsmeanpadded_normalizedvariancestdpadded_layer_normjagged_layer_norm_valuesr/rrrnative_layer_norm_defaults       rzpgrad_out: jt, input: jt, normalized_shape: any, mean: any, rstd: any, weight: any?, bias: any?, output_mask: anyc Ospt|||dd\}}|d}|d}||j|jfi|\}}} |dur+d|| fSt|fit||| fS)NTrrnrrp) rhrirjrrrnrd_inputd_gammad_betarrr"native_layer_norm_backward_defaultZs   r z"self: jt_all, dim: any, index: anycOst|||dd\}}|d}t||d|jddd\|d<}|r+||dS|jdur4tdt|}|d|jd krI|d d 8<t ||j fi|fi|S) NTrrr-selectrrzSselect(): not yet supported on dim != 0 for non-contiguous nested tensor with holesrrV) rrr8r-rVrrUrerrr)rhrirjrrrr7rrrr select_intks"  rz7self: jt, dim: any?, start: any?, end: any?, step: any?cOs\t|||dd\}}|d}t||d|jd|d<t||jfi|fit|S)NTrrr-slice)rrr8r-rVrrrrrrr slice_tensors "rz8input: jt_all, indices: any, values: t, accumulate: any?cOst|||dd\}}|d}|d}t||kr+tdt|d|dt||jdkr|s    rcOr{rrYrrrr mean_default4r}rz&self: jt_all, dim: any?, keepdim: any?cOrv)Nr<Frwrrrrany_dims?rz%self: jt_all, dim: any, keepdim: any?cO:t|||dd\}}|dg|d<ttjjjjfi|SNTrr-)rrrrrr<r=rrrrany_dimD  r#cOrv)NrTrwrrrrall_dimsOr r%cOr!r")rr%rrrrr=rrrrall_dimTr$r&cOr{rrYrrrrall_any_max_min_default_s   r'cO(t|||dd\}}|d}||jSrrYrrrr_is_true_defaultrs   r)cOBt|||dd\}}|dj}t|d}t|d|g|Ri|S)NTrrr rrrrrxrhrirjrrr dtype_maxrrrmin_dim   r.cOr*)NTrrrr r+rhrirjrrr dtype_minrrrmax_dimr/r2cOr*)NTrrr aminr+r,rrr amin_default   r4cOr*)NTrrramaxr+r0rrr amax_defaultr5r7cOr*)NTrrr argminr+r,rrrargmin_defaultr5r9cOr*)NTrrrargmaxr+r0rrrargmax_defaultr5r;zAgrad: jt_all, dim: any, indices: jt_all, sizes: any, keepdim: anyc sddlmt|||dd\}}|d}|j|d<|d}|j|d<tfddt|d D}tt|d |d |d |d <|d }|j |j d ||<|d d}||d <t |} || d <t |di|fi| S)NrrTrgradrc3s |] \}}|r|VqdSrr)rror~rrrr#r;z=value_selecting_reduction_backward_default..sizesr-"value_selecting_reduction_backwardrrVr) rrrrrrrgr8rdrrVrr) rhrirjrrr<rr:r=rWrrr*value_selecting_reduction_backward_defaults,        r?cOst|||dd\}}|d}|D]%}t|tstd||dkr*tdt||djs6tdqt|dd|d |dj d |d <t|d d |Dfi|fit |dS) NTrrz+stack(): expected all nested tensors inputsrz9stack(): expected all nested tensors to have the same dimzFstack(): expected all nested tensors to have the same nested structurerr-rcSrIrrrrrrr9 rMz!stack_default..) rrr&rr0r-rrr8rVr)rhrirjrrrrZrrr stack_defaults0   r@zQweight: t, indices: jt, padding_idx: any?, scale_grad_by_freq: any?, sparse: any?cOsLt|||dd\}}|d}|d}t|||jfi|fit|S)NTrrrr )rhrirjrrrrrrrembedding_default s  rAzYgrad_output: jt, indices: jt, num_weights: any, padding_idx: any, scale_grad_by_freq: anycOs>t|||dd\}}|d}|d}||j|jfi|S)NTrrr#rY)rhrirjrrrr#rrr embedding_dense_backward_default s   rBcOs(t|||dd\}}|d}|jSr)rrrdetachrrrrvalues_default, s   rDcOr(rrYrrrr all_default? s   rEz-self: jt_all, padding: any, output_size: any?c Oslt|||dd\}}|d}|jdurtd|d}|dur&||j}n|jdur.|jn|jd}| }|jdkrF| |jdd}|j } | dkrV|j dd }n | dkra|d }|jtju} | rr|jrr|tj}tjj||jg|g|d } | r| jr| tj} t| dkr| d | dd} n t| dkr| d } |jdkr| |jd} | S) NTrrz?to_padded_tensor(): not supported for nested tensors with holesrrrrrrLpadding)rrrUr0rV_max_seqlen_tensorrcrrr~rrr-rrrrboolrr<halfrrrbrrdrr) rhrirjrrrr max_seq_lenr~ values_shapeis_bool padded_outrrrto_padded_tensor_defaultJ sR                  rNzcpadded: t, offsets: t, dummy: jt, ragged_idx: any?, min_seqlen: any?, max_seqlen: any?, sum_S: any?cOs`t|||dd\}}|d|d}}|dd}|dkr#||d}|j}|dkr3|jdd }n |dkr>|d }|jtj u} | rO|j rO| tj }tj j||g|d } | rf| j rf| tj } t|dkrw| d |dd} n t|dkr| d } |dkr| |dd } |d } |d} i} | dur| | d <| dur| | d<t| ||| dS)NTrpaddedrr:rrrrrLrrrr)rVr)rrrrr-rrrrrHrr<rIrrrdrdrrr)rhrirjrrrOrr:padded_ragged_dim1_shaperLr~rrmetadata_cacherrr"_nested_from_padded_tensor_default sN               rRzcvalues: t, offsets: t, dummy: jt_all, lengths: t?, ragged_idx: any?, min_seqlen: t?, max_seqlen: t?c Os~t|||dd\}}|d|d|d}}}|d}|d} |d} i} | dur.| | d<| dur6| | d<t||||| d S) NTrrrrr:rr)rrVr)rr) rhrirjrrr~rrr:rrrQrrr _nested_view_from_jagged_default s,  rScO$t|||dd\}}|d}|jSr)rrrrrrr_nested_get_offsets  rUcOrTr)rrrUrrrr_nested_get_lengths rVrWcOrTr)rrrVrrrr_nested_get_ragged_idx rVrXcO,t|||dd\}}|d}|jddS)NTrrrrrrrrrrr_nested_get_min_seqlen  r[cOrY)NTrrrrZrrrr_nested_get_max_seqlen r\r]zself: jt, mask: anyc Ost|||dd\}}|d}|d}|jdkrtd|j|jkr/td|jd|j|j|}t |j d d d }t |}||j |d <t dd |i|S)NTrrrtrz3masked_select only support 2-D selections currentlyzMask with shape z& is not compatible with input's shape rr^rrr~r)rrr!r0rr masked_selectr~rrrrrr) rhrirjrrrrt res_values mask_cumsumrrrmasked_select_default s(    raz2grad_output: t, self: jt_all, dim: any, index: anycOsVt|||dd\}}|d}|d}tj||jd}||d|d||S)NTrrr#)rr-r)rrr zeros_likerr rN)rhrirjrrrr# grad_inputrrr_nested_select_backward_default. s  rdzself: jt_all, s: anycOsF|d}|d}||j|||j||jdur!||j|dSdSr)rrrU)rhrirjrstreamrrrrecord_stream_default@ s   rfzRself: jt_all, size: any, dtype: any?, layout: any?, device: any?, pin_memory: any?cOsHt|||dd\}}|d}t|ddkr ||jfi|Std)NTrrrrz2new_empty() not supported for NJT with shape != ())rrrdrr0rrrrnew_empty_defaultK s  rgzself: jt_all, ...cOs:tdd|D}t|dd|Di|fit|S)NcsrrrWrrrrr#m rz&activation_backward..css$|] }t|tr |jn|VqdSr)r&rrrrrrr#p rr)rhrirjr#rrractivation_backward` s  rhzself: jt_all, value: anycOrrr rrrr fill_Scalarw r"ricOs4t|||dd\}}|d}||jfi||SrrYrrrr fill__Scalar s  rjc OsTt|||dd\}}|d}t|}||j\}}t|fi|t|fi|fSr)rrrrr) rhrirjrrrrWmantissaexponentrrr frexp_Tensor s rmz+grad: any, self: any, other: any, mask: anyc Ost|||dd\}}|d}|d}|d}|d}|dur$dSd} |dr4t||d d } d} |d rDt|d d |} | | fS) NTrr<rrrtrrrLr)rrrr'r) rhrirjrrr<rrgrad_input_mask grad_self grad_otherrrrmatmul_backward_default s     rrz self: anycOsddlm}|S)Nr)_nt_view_dummy)$torch.nested._internal.nested_tensorrs)rhrirjrsrrr_nested_get_jagged_dummy s rurIMPLCPUCUDAMeta)F)TFF)rr)rrrtypingrtorch.nn.functionalr`rartorch.fx.operator_schemasrtorch.nested._internal.sdparrrrr(Any__annotations__r Dictrrr8rCstrrur|rrrrregister_jagged_funcCallablerrrrrrrrrrr-rrr r r primr4rrrrrrrrr)rlinearr!linear_backwardr0r<rr1_to_copyrHrNrUrC empty_like ones_likerb rand_like randn_likerX full_like randint_like low_dtypezero_rZ_softmaxrj _log_softmaxrl_softmax_backward_dataronative_dropoutrsnative_dropout_backwardrurdim_intryr|rbrPrrrrrrrrrrrrrrrr'rbmmrrrrrrrwherer%r _pin_memoryr is_pinnedr is_same_sizerrxrrrrrrpermuterview _unsafe_viewrnative_layer_normrnative_layer_norm_backwardr r rrr index_putr convolutionrrrrr<r=rr#rr%r&r rr' _is_all_true _is_any_truer)r.r2r3r4r6r7r8r9r:r;r>r?rr@ embeddingrAembedding_dense_backwardrBr~_nested_get_valuesrDrErrN_nested_from_padded_tensorrR_nested_view_from_jaggedrSrUrWrXr[r]r^ra_nested_select_backwardrd record_streamrf new_empty new_zerosnew_onesrg elu_backwardhardshrink_backwardhardsigmoid_backwardhardtanh_backwardsoftplus_backwardsoftshrink_backwardrhfillScalarrifill_rjfrexprmmatmul_backwardrrrulibrary_scoped_libraryimplrrrrs      @    )  Of                    %  )        0    H  %                ) H    m            j     * ( /  a        X    )                                   ;  5                        $