a  d@sdZddlmZmZmZmZmZddlZddlmZddl m Z zddl m Z dZWneyjd ZYn0dd lmZmZmZmZdd lmZdd lmZmZgd ZeeeeeeehZzddlmZeeWneyYn0eejdddZeejdddZ ddZ!e"Z#edddZ$edddZ%ddZ&ejeee'e'fee'fdddZ Gd d!d!ejZ(Gd"d#d#ejZ)dS)$zm PyTorch FX Based Feature Extraction Helpers Using https://pytorch.org/vision/stable/feature_extraction.html )CallableListDictUnionTypeN)nn)_get_feature_info)create_feature_extractorTF) Conv2dSameScaledStdConv2dSame CondConv2d StdConv2dSame)BilinearAttnTransform) MaxPool2dSame AvgPool2dSame) register_notrace_moduleis_notrace_moduleget_notrace_modulesregister_notrace_functionis_notrace_functionget_notrace_functionsr FeatureGraphNetGraphExtractNet) InplaceAbnmodulecCst||S)zq Any module not under timm.models.layers should get this decorator if we don't want to trace through it. ) _leaf_modulesaddrra/var/www/html/stable-diffusion-webui/venv/lib/python3.9/site-packages/timm/models/_features_fx.pyr+s rcCs|tvSN)rrrrr r3srcCsttSr!)listrrrrr r7srfunccCst||S)zF Decorator for functions which ought not to be traced through )_autowrap_functionsrr#rrr r?s rcCs|tvSr!)r%r#rrr rGsrcCsttSr!)r"r%rrrr rKsrmodel return_nodescCs(ts Jdt||ttttddS)NKPlease update to PyTorch 1.10+, torchvision 0.11+ for FX feature extraction)Z leaf_modulesZautowrap_functions)Z tracer_kwargs)has_fx_feature_extraction_create_feature_extractorr"rr%r&rrr r Os  r cs*eZdZdZdfdd ZddZZS)rzX A FX Graph based feature extractor that works with the model feature_info metadata NcshttsJdt||_dur>ttks>Jfddt|jD}t|||_dS)Nr)cs6i|].\}}|vr|ddur*|n|dqS)rNr).0iinfo out_indicesout_maprr `sz,FeatureGraphNet.__init__..) super__init__r*r Z feature_infolen enumerater graph_module)selfr'r0r1r( __class__r/r r4Zs    zFeatureGraphNet.__init__cCst||Sr!)r"r7values)r8xrrr forwardeszFeatureGraphNet.forward)N)__name__ __module__ __qualname____doc__r4r= __classcell__rrr9r rWs rcs\eZdZdZd eeeefeefedfdd Z eee j e j fdddZ Z S) ra A standalone feature extraction wrapper that maps dict -> list or single tensor NOTE: * one can use feature_extractor directly if dictionary output is desired * unlike FeatureGraphNet, this is intended to be used standalone and not with model feature_info metadata for builtin feature extraction mode * create_feature_extractor can be used directly if dictionary output is acceptable Args: model: model to extract features from return_nodes: node names to return features from (dict or list) squeeze_out: if only one output, and output in list format, flatten to single tensor T)r( squeeze_outcs t||_t|||_dSr!)r3r4rCr r7)r8r'r(rCr9rr r4vs zGraphExtractNet.__init__)returncCs0t||}|jr,t|dkr,|dS|S)Nrr)r"r7r;rCr5)r8r<outrrr r={szGraphExtractNet.forward)T)r>r?r@rArrstrrboolr4torchZTensorr=rBrrr9r ris *r)*rAtypingrrrrrrHrZ _featuresr Z%torchvision.models.feature_extractionr r+r* ImportErrorZ timm.layersr r r rZtimm.layers.non_local_attnrZtimm.layers.pool2d_samerr__all__rrrModulerrrsetr%rrrrFrrrrrr s@         &