U yhs+@sdddlmZmZmZmZmZmZmZmZddl Z ddl m Z ddl Z ddl Z ddlZ ddlmZddlmZdddd d gZeee jee jfZee jefZee jjZee jjZeeZd d d hZeddddZeddeee jj fe jjedddZ!edde jje"dddZ#eddGdd d Z$edde jj%e jj%ddd Z&dS))ListTupleUnionDictAnySetMappingOptionalN) dataclass)_get_qualified_name) compatibilityget_acc_ops_nameget_node_targetis_node_output_tensorFxNetAccFusionsFinderlegalize_graph call_module call_functionZ call_methodF)Zis_backward_compatiblecCsTt|tr|S|jr*d|jkr*d|jS|jdd}|r@|ndd|jSdS)Nacc_opsacc_ops.z torch._opsz torch.ops.) isinstancestr __module____name__replace)kmodulerN/var/www/html/venv/lib/python3.8/site-packages/torch/fx/passes/tools_common.pyr s   ) submodulesnodereturncCs|jtks(tddtd|j|jdkrdt|jtsBt||j}t|dt|}t |S|jdkr|j}|j dk rd|j krd |j St |St|jtst|jSdS) a, Given a `node` returns its target typename. For "call_method" node, return node.target which is the name of that method being called. This could potential lead to conflict but should be okay because normally it's on a tensor. For "call_function" node, return typename of node.target. For "call_module" node, return typename of the module that node.target point to. If seeing "_VariableFunctionsClass" in the target name string, it will be replaced by "torch". e.g. _VariableFunctionsClass.relu would become torch.relu. zExpect op types of z, z , but found rZ_base_class_originrNrr) opCALLABLE_NODE_OPSAssertionErrorjoinrtargetrgetattrtyper rrr )r!r"ZsubmodZ submod_typer(rrr r!s$    )r"r#cCs"|jdd}|dk o t|tjS)aChecks if the node output produces a Tensor or not. NOTE: This requires to run `ShapeProp` on the containing fx graph before calling this function. This is because it works by checking the `type` metadata on the node. This metadata is produced by the `ShapeProp`. r*N)metaget issubclasstorchTensor)r"type_rrr rEsc@speZdZdZejjedddZe GdddZ dde ee fe ed d d Zeejjefd d dZdS)rz Finds groups of connected ACC nodes that pass non-tensor data between each other. Such groups are called fusion groups. )r acc_nodescCs||_t|jj|_||_dSN)rlistgraphnodesr1)selfrr1rrr __init__WszFxNetAccFusionsFinder.__init__c@s6eZdZUeed<eed<eed<eed<ddZdS)!FxNetAccFusionsFinder.FusionGroup top_node_idxr5inputsnodes_need_processcsR|jkrdSj|j|j|jfdd|jDdS)z5 Add a node to fusion group. Ncs$h|]}|jtkr|jkr|qSr)r$r%r5).0nr6rr us z=FxNetAccFusionsFinder.FusionGroup.add_node..)r5r;addr:discardupdateall_input_nodes)r6r"rr>r add_nodejs     z*FxNetAccFusionsFinder.FusionGroup.add_nodeN)rr __qualname__int__annotations__NodeSetrDrrrr FusionGroup\s rINr8) fusion_groupr:visitedcCsz|D]p}|dk r$||krq|||jtkr0q|j||jkrDq||jkrTdS|||j|r||dSqdS)z Start from inputs and going reverse topological order. If any upstream node is in the fusion group, add all the nodes in this path to fusion group. NTF) r@r$r%r5indexr9recursive_add_noderCrD)r6rJr:rKargrrr rM|s     z(FxNetAccFusionsFinder.recursive_add_node)r#cCsi}t|j}|D]l}||kr"q|jtkr.qd|jkr:q||jkrFq|j|j||ht|j |hd}|j rD|j }|j ||j tdd|jkr|jD]:}|jtkrq||jkrq|||j ||j tdq|j D]^}|jtkrqd|jkrq||jkr q||t|j|j||_|j ||j tdqqjt|j|jksh|j|j8_q|jD]}|j||<qnq|S)NZ tensor_meta)r9r5r:r;)rK)r3r1r$r%r+rIr5rLsetrCr;poprMr:usersrDminr9)r6resultr1r"rJuserrNr=rrr __call__st                    zFxNetAccFusionsFinder.__call__)N)rrrE__doc__r.fx GraphModulerHr7r rIrNodeListr rMrNoderUrrrr rPs#  &)gmr#cstjtjtjtjtjtjtjtjtj tj tj tj t jjjjt jjjjt jjjjt jjjjt jjjjg}t|jjdt j}|jjD] }|jD]}|d7<qqt}|jjD]}|dkr| |qit!|dkrd|"}|#|fdd|<|jD]R}|d8<|dkr|j$dkrT|j%|krT|&|n | |qqt!|jt!|jjkrt'dfddD|jj(|_(||_|S) a Replace the graph of the given GraphModule with one that contains the same nodes as the original, but in topologically sorted order. This is used by the merge_matmul transformation below, which disturbs the topologically sorted order of its input GraphModule, so that this order is restored before further transformation. Arguments: gm: The graph module to topologically sort. It is modified in-place. Returns: The graph module in-place sorted rcs|Sr2r)x)envrr !z legalize_graph..rz&Input graph has cycles, unable to add csg|]}|dkr|qS)rr)r<r")indegrr ,s z"legalize_graph..))operatorr@mulsubfloordivtruedivmodleltgegteqner.ZopsZatenZsym_constrain_rangedefaultZsym_constrain_range_for_sizeZ _assert_asyncmsgZ scalar_tensorZ_assert_scalardictfromkeysr4r5rWZGraphrQ collectionsdequeappendlenpopleftZ node_copyr$r( appendleft RuntimeErrorZ_codegen)r[ZPRIORITIZED_OPSZ new_graphr"rTqueuecurr)r^rar rsT              )'typingrrrrrrrr rs dataclassesr rcr.Ztorch.fxZ torch.fx.noder Ztorch.fx._compatibilityr __all__r/ZTensorsZTensorOrTensorsrWrZrYrHrZNamesr%r nnModulerboolrrrXrrrrr s2(       $#