3j'SrSSKrSSKrSSKrSSKJrJrJr SSK J r SSK J r SSK JrJrJrJr SSKJr \"1Sk5r"S S \ 5r\R.r\R0r\R2r\R6rg) z Implements ONNX's backend API. N) ModelProtohelperversion)Backend) check_model)InferenceSessionSessionOptionsget_available_providers get_device)OnnxRuntimeBackendRep> logidexecution_modeexecution_orderenable_mem_reuseenable_mem_patternlog_severity_levellog_verbosity_levelenable_cpu_mem_arenainter_op_num_threadsintra_op_num_threadsuse_per_session_threadsgraph_optimization_leveluse_deterministic_computec\rSrSrSr\"\R"SS5S:H5r\ S Sj5r \ S5r \ S5r \ S S j5r \ S S j5r\ SS j5rS rg)OnnxRuntimeBackend,a Implements `ONNX's backend API `_ with *ONNX Runtime*. The backend is mostly used when you need to switch between multiple runtimes with the same API. `Importing models from ONNX to Caffe2 `_ shows how to use *caffe2* as a backend for a converted model. Note: This is not the official Python API. ALLOW_RELEASED_ONNX_OPSET_ONLY1Nc >Uc [5nURU5$)z Return whether the model is compatible with the backend. :param model: unused :param device: None to use the default device or a string (ex: `'CPU'`) :return: boolean )r supports_device)clsmodeldevicekwargss U/home/wildlama/miniconda3/lib/python3.13/site-packages/onnxruntime/backend/backend.py is_compatible OnnxRuntimeBackend.is_compatible:s! >\F""6**cUR(amURH]nUR(a UROSnX2R4nU[R ;aSUSURS3nSU4s $M_ g ![ aN US:XaURS:dUS:Xa.URS:aSUSURS3nSU4ss $MMf=f) aS Return whether the opset for the model is supported by the backend. When By default only released onnx opsets are allowed by the backend To test new opsets env variable ALLOW_RELEASED_ONNX_OPSET_ONLY should be set to 0 :param model: Model whose opsets needed to be verified. :return: boolean and error message if opset is not supported. zai.onnxzSkipping this test as only released onnx opsets are supported.To run this test set env variable ALLOW_RELEASED_ONNX_OPSET_ONLY to 0. Got Domain 'z ' version 'z'.F z ai.ommx.ml)T)allowReleasedOpsetsOnly opset_importdomainrrOP_SET_ID_VERSION_MAPAttributeError)r!r"opsetr/key error_messages r%is_opset_supported%OnnxRuntimeBackend.is_opset_supportedGs  & &++).94!==1C&">">>,,28;u}}oRQ& %m33 ? ,,& 4)+ 0BR^H^chcpcpstct,,28;u}}oRQ& %m33 duH^ 4s7BA CCcUS:XaSnSU-[5;=(d$ US-[5;=(d U[5:H$)zu Check whether the backend is compiled with particular device support. In particular it's used in the testing suite. CUDAGPU-)r )r!r#s r%r "OnnxRuntimeBackend.supports_devicejs@ V FV|z|+ev|z|/KevYcYeOeer(c [U[5(aU$[U[5(a [U5$[U[[45(Ga[ 5nUR 5HZupVU[;a[XEU5 M[XE5(dM/[SUSSR[[5535e [R"SSS9RS5n[!5Vs/sH oU;dM UPM n n[XU S9n U R#5 Ub0UR%U5(d[S US ['5S 35eUR("X40UD6$[*R,R/[,R,5=(d [*R,R1S 5n U R2S :n U (aUR55OUn [7U 5 UR9U5upU(d[:R<"U5eU n[U[[45(dUR55nUR("UU40UD6$s snf)u Load the model and creates an :class:`onnxruntime.backend.backend_rep.OnnxRuntimeBackendRep` ready to be used as a backend. :param model: the model to prepare — accepts a file path (str), serialized model (bytes), :class:`onnx.ModelProto`, :class:`onnxruntime.InferenceSession`, or :class:`onnxruntime.backend.backend_rep.OnnxRuntimeBackendRep` (returned as-is) :param device: requested device for the computation, None means the default one which depends on the compilation settings :param kwargs: only a safe subset of :class:`onnxruntime.SessionOptions` attributes are accepted; see ``_ALLOWED_SESSION_OPTIONS`` for the list :return: :class:`onnxruntime.backend.backend_rep.OnnxRuntimeBackendRep` zSessionOptions attribute 'z<' is not permitted via the backend API. Allowed attributes: z, "ORT_ONNX_BACKEND_EXCLUDE_PROVIDERSr,)default,) sess_options providerszIncompatible device expected 'z', got ''0) r) isinstancer rstrbytesr items_ALLOWED_SESSION_OPTIONSsetattrhasattr RuntimeErrorjoinsortedosgetenvsplitr disable_fallbackr r prepare packagingrparseVersionreleaseSerializeToStringrr5unittestSkipTest)r!r"r#r$optionskvexcluded_providersxrAinf onnx_version$onnx_supports_serialized_model_check bin_or_modelopset_supportedr4bins r%rTOnnxRuntimeBackend.preparets e2 3 3L / 0 0(/ / U| , ,$&G 00G*W((&4QC8//3yy@X9Y/Z.[] '"$+OY[!\!b!bcf!g $;$=_$=qK]B]$=I_"5)TC  "!#*=*=f*E*E"%CF88T^T`Saab#cdd;;s5f5 5 %,,227??CeyGXGXG`G`adGeL3?3G3G:3U 08\5224bgL  %-0-C-CE-J *O"'' 66CcC<00++-;;sF5f5 55`s < I( I(c NUR"X40UD6nUR"U40UD6$)u Compute the prediction. :param model: the model to run — accepts a file path (str), serialized model (bytes), :class:`onnx.ModelProto`, :class:`onnxruntime.InferenceSession`, or :class:`onnxruntime.backend.backend_rep.OnnxRuntimeBackendRep` :param inputs: inputs :param device: requested device for the computation, None means the default one which depends on the compilation settings :param kwargs: ``run_model()`` forwards kwargs to both ``prepare()`` and ``rep.run()``. ``prepare()`` validates and applies ``_ALLOWED_SESSION_OPTIONS`` only when creating a new session from a model path or bytes; if ``model`` is already an ``InferenceSession`` or ``OnnxRuntimeBackendRep``, session-option kwargs are silently ignored. ``rep.run()`` always validates against ``_ALLOWED_RUN_OPTIONS`` and raises ``RuntimeError`` for known-but-blocked run attributes. Logging-related kwargs (``log_severity_level``, ``log_verbosity_level``, ``logid``) appear in both allowlists. :return: predictions )rTrun)r!r"inputsr#r$reps r% run_modelOnnxRuntimeBackend.run_models+,kk%262wwv(((r(c [S5e)zq This method is not implemented as it is much more efficient to run a whole model than every node independently. zMIt is much more efficient to run a whole model than every node independently.)NotImplementedError)r!noderjr# outputs_infor$s r%run_nodeOnnxRuntimeBackend.run_nodes ""qrrr()N)NN)__name__ __module__ __qualname____firstlineno____doc__boolrPrQr- classmethodr&r5r rTrlrr__static_attributes__rtr(r%rr,s #299-Ms#SWZ#Z[ + +  Dff:6:6x))0ssr(r)ryrPrZpackaging.versionrUonnxrrronnx.backend.baser onnx.checkerr onnxruntimerr r r onnxruntime.backend.backend_repr frozensetrJrr&rTrlrir rtr(r%rs{  ,,%$]]A%&dsdsN#00  $ $""$44r(