VhdZddlZddlmZmZmZmZddlmZm Z m Z ddl m Z m Z mZddlmZddlmZdd lmZmZmZmZmZmZmZmZmZmZmZmZmZmZdd lm Z m!Z!m"Z"m#Z#m$Z$m%Z%m&Z&m'Z'm(Z(dd l)m*Z*dd l+m,Z,dd l-m.Z.ddl/m0Z0ddl1m2Z2m3Z3m4Z4m5Z5ddl6m7Z8gdZ9ejtjwejxj"jzgej|ej~j|urUddl@Zejej|e_>ejjjej|dddZEddZFy)a TorchDynamo is a Python-level JIT compiler designed to make unmodified PyTorch programs faster. TorchDynamo hooks into the frame evaluation API in CPython (PEP 523) to dynamically modify Python bytecode right before it is executed. It rewrites Python bytecode in order to extract sequences of PyTorch operations into an FX Graph which is then just-in-time compiled with a customizable backend. It creates this FX Graph through bytecode analysis and is designed to mix Python execution with compiled backends to get the best of both worlds: usability and performance. This allows it to seamlessly optimize PyTorch programs, including those using modern Python features. N)config convert_frame eval_frameresume_execution) list_backendslookup_backendregister_backend)callback_handleron_compile_endon_compile_start) code_context)replay)allow_in_graphassume_constant_resultdisabledisallow_in_graphforbid_in_graph graph_break mark_dynamic mark_staticmark_static_addressmaybe_mark_dynamicnonstrict_tracerun set_stancesubstitute_in_graph) _reset_guarded_backend_cacheexplainexportis_dynamo_supportedis_inductor_supportedoptimizeoptimize_assertOptimizedModule reset_code) is_compiling)GenerationTracker)reset_code_state)TensorifyState)graph_break_reasonsguard_failures orig_code_mapreset_frame_count)loader)rrrrrrrrrrrr#r$r rrrrrresetr%r'r rr rzaten::manual_seedcfddl}|jt}|jdtj 5t tjjttjjtjtjtjtjj jt#t%t&j(j*j-dt_tj0jt3jt5jt7jt&j(j8j:jt&j(j8j<jt&j>j@jCddddy#1swYyxYw)a Clear all compile caches and restore initial state. This function is intended to reset Dynamo's state *as if* you had started a fresh process invocation, which makes it good for testing scenarios where you want to behave as if you started a new process. It does NOT affect any file system caches. NB: this does NOT reset logging state. Don't use this to test logging initialization/reinitialization. rNztorch._dynamo.resetF)"logging getLogger__name__infor compile_lockreset_code_caches input_codesclearr) output_codesr-r,r+rContinueExecutionCachecacherr.torch_dynamocompiled_autogradr0 FRAME_COUNTERFRAME_COMPILE_COUNTERr r(r*utilswarn_once_cacheuser_obj_id_to_weakref_C _autograd _saved_tensors_hooks_set_tracing)r2logs F/home/dcms/DCMS/lib/python3.12/site-packages/torch/_dynamo/__init__.pyr0r0bsb   H %CHH "#  # #C!!'')""((*!!#//55;;=$& ''--/&' #++113 ! ++113 2288: ;;EB'CCCs G#H''H0c~ddl}|jt}|jd tj 5t tjjtjjzD]}|}|s t|tjdddy#1swYyxYw)z Clears in-memory code cache, which is what stores compiled products. This resets less state than :func:`reset` and is mostly only used for testing purposes. rNztorch._dynamo.reset_code_caches) r2r3r4r5rr6r)r8seenr:r&rr9)r2rH weak_codecodes rIr7r7s   H %CHH ./=  # #  % % * *]-G-G-L-L L !I;D4  ! sA B3 !B33B<)returnN)G__doc__r=rrrrbackends.registryrr r callbackr r r rr decoratorsrrrrrrrrrrrrrrrrr r!r"r#r$r%r&external_utilsr'mutation_guardr(pgor)symbolic_convertr*rBr+r,r-r. polyfillsr/___all__ serializationadd_safe_globalsr> _DimRange manual_seedrandomtorch.jit._builtins_disable_dynamojit _builtins_register_builtinr0r7rIrgs AANNHH&!   )-!,XX# <$$emm&>&>&H&H%IJ 000.--e.?.?@E II))%*;*;=PQ"CJrf