Ph% 2ddlZddlZddlmZddlmZmZmZddlm Z ddl Z ddl m Z ddl m Z ddlmZdd lmZ ed Zed Zed Zed dZdZeee ZedeeeZedeeeedej4e dZedeee ZedeGddeZGddeZed e j@jBfd!Z"ed e j@jBfd"Z#ed e j@jBfd#Z$ed e j@jBfd$Z%ejLGd%d&Z'd e j@jBfd'Z(Gd(d)Z)y)*N) import_module)AnyListOptional)#min_cut_rematerialization_partition)_guards) ts_compile) aot_autograd)register_debug_backendc|SNgmfake_tensor_inputss kC:\Users\daisl\Desktop\realtime-object-detection\venv\Lib\site-packages\torch/_dynamo/backends/debugging.pyeagerrs IcXddlm}fd}||d|}|j|S)Nr)make_fxcZtjjj|Srtorchfx Interpreterrun)argsrs r runnable_gmz'pre_dispatch_eager..runnable_gms#xx##B'++T22rT) pre_dispatch)"torch.fx.experimental.proxy_tensorrprint_readable)rrrrpre_dispatch_gms` rpre_dispatch_eagerr$s3:3>gk=?QRO""$ rc"ddlmfd}|S)Nr)SchemaCheckModec5tjjj|cdddS#1swYyxYwrr)rr&rs rinnerzeager_debug..inner,s2  88''+//6  s +>A)#torch._subclasses.schema_check_moder&)rrr(r&s` @r eager_debugr*%sC 7 Lrts)namec@tjj|Sr)rjitscriptrs r torchscriptr03s 99  B rc fd}d|_|S)Nc`tjjj|Sr)rrr boxed_run)rfx_gs rrzboxed_nop..run:s#xx##D)33D99rT) _boxed_call)r4example_inputsrs` r boxed_nopr79s:CO Jr) fw_compiler partition_fn aot_eager)r, compiler_fn)r8aot_eager_default_partitionerc4tdjS)Nztorch._inductor.compile_fx)rselect_decomp_tablerrrr?Vs=$rinductor)compiler)r8 bw_compilerdecompositionsr9aot_eager_decomp_partitionaot_tsc eZdZy)ReluCompileErrorN__name__ __module__ __qualname__rrrrGrGkrrGc eZdZy)TestingOnlyCompileErrorNrHrrrrNrNorLrrNrc|jjD])}|jtjk(s!t |Sr)graphnodestargetrrelurGrr6nodes rrelu_compile_error_TESTING_ONLYrVss4 ;;%** $"$ $ Irc|jjD]<}|jtjk(s!tj |_d|_>|j|S)N)FReluRuntimeError)rPrQrRrrS_assertr recompilerTs rrelu_runtime_error_TESTING_ONLYr[{sF ;;%** $--DK3DILLN Irc|jjD]K}|jtjk(s!tj |_|j ddf|_M|j|S)Nrr )rPrQrRrrSaddrrZrTs r relu_accuracy_error_TESTING_ONLYr^sS ;;%** $))DK1q)DILLN Irc|jjD]}|jdk(sn|S|D]}|jrt |S)N call_function)rPrQopis_leafrN)rr6rUts r#non_leaf_compile_error_TESTING_ONLYrdsM 77o %  yy)+ + IrceZdZUdZeej jed<e ed<e ed<ee ed<e ed<dZ e eej jed<dZe eej ed <dZe eed <d Zy) ExplainOutputzu This is the output of :func:`torch._dynamo.explain()` There is no reason to create this class directly. graphs graph_countgraph_break_count break_reasonsop_countN ops_per_graph out_guards compile_timescd|jd}|d|jdz }|d|jdz }|dz }t|jD]C\}}|d|dzdz }|d |j dz }|d z }|j D] }|d |dz } E|j>|d z }t|jD]!\}}|d |dzdz }|D] }|d|dz } #|j?|dz }t|jD]"\}}|d|dzdz }|dt|z }$|j|d|jdz }|S)Nz Graph Count:  zGraph Break Count: z Op Count: zBreak Reasons: z Break Reason r z: z Reason: z User Stack: z zOps per Graph: z Ops z z Out Guards: z Guard zCompile Times: ) rhrirk enumeraterjreason user_stackrlrmstrrn) selfoutputidx break_reason frame_summaryopsraiguards r__str__zExplainOutput.__str__s !1!1 2"5'(>(>'?rBBJt}}oR00$$!*4+=+=!> C Awc2 2F  \%8%8$9< >F r)rIrJrK__doc__rrr GraphModule__annotations__intrrlrNodermrGuardrnrtr}rrrrfrfs %% && M37M8D/0704Jgmm,-4#'M8C='rrfcV|j||jjDcgc]}|jdk(s|j }}|t |z }|j||j jr|j|j |||||fScc}w)a This function is a utility which processes a torch.fx.GraphModule and accumulates information about its ops, graph breaks, and other details. It is intended to be used by the ExplainWithBackend class and `torch._dynamo.explain()` to provide details from Dynamo's graph capture. Parameters: gm (torch.fx.GraphModule): The GraphModule to be processed. graphs (list): A list that accumulates all the GraphModules processed. op_count (int): The total count of operations in all GraphModules processed so far. ops_per_graph (list): A list that accumulates the operations of each GraphModule. break_reasons (list): A list that accumulates the reasons for breaks in each GraphModule. Returns: tuple: A tuple containing the processed GraphModule, the updated lists of graphs, operations per graph, and break reasons, and the updated operation count. r`)appendrPrQrarRlencompile_subgraph_reason graph_break)rrgrkrlrjrUrzs r_explain_graph_detailrs( MM"#%88>> P>4TWW5O4;;>C P CH !!--R778 vx == Qs B&B&cVeZdZdZdZdej jfdZde fdZ y)ExplainWithBackenda This class is intended to be used as a backend for `torch.compile`. It is composable with other backends. When used in this way, it accumulates information about graph breaks, ops, and other info and provides a string representation summarizing this information. Attributes: backend (str): The name of the backend to use for optimization. graphs (list): A list of the graphs captured by TorchDynamo. op_count (int): The total number of operations in all optimized graphs. break_reasons (list): A list of graph break reasons with stack traces. Example Usage: def fn(x): x = torch.sigmoid(x) return x torch._dynamo.reset() eb = ExplainWithBackend("inductor") optimized_fn = torch.compile(fn, backend=eb) result = optimized_fn(torch.randn(5)) print(eb.output()) cTddlm}|||_g|_d|_g|_y)Nr )lookup_backendr)registryrbackendrgrkrj)rurrs r__init__zExplainWithBackend.__init__s',%g.   rrct||j|jg|j\}|_|_}|_|j ||Sr)rrgrkrjr)rurr6_s r__call__zExplainWithBackend.__call__sK@U  T]]B0B0BA =DK4+=||B//rreturnct|j}t|j||dz |j|j}|S)Nr )rrgrfrjrk)rurhrvs rrvzExplainWithBackend.outputsB$++&  KK  !O    MM   rN) rIrJrKr~rrrrrrfrvrrrrrs.0 0588//0  rr)* dataclasses functools importlibrtypingrrrfunctorch.compilerrrtorch._functorch.compilersr commonr rr register_backendrr$r*r0r7r:r<partialrDrE ExceptionrGrNrrrVr[r^rd dataclassrfrrrrrrs#&&A 1 @     t   (K  ky9 , C (6S*#""+j  %3M * -hF3  y  i (<(<(<(<)=)= EHH,@,@   ---`> ><11r