BVh=9jdZddlZddlZddlZddlmZd dZ d dZdZ ddZ ddZ d Z d Z d Zy)z/Utilities related to layer/model functionality.N)nestct|ds|S||r|j\}}}|js|gS|j|}|jrt j |j Sg}|jD]A\}}}}t|||}|D](tfd|Ds|j*C|S)aReturns the list of input tensors necessary to compute `tensor`. Output will always be a list of tensors (potentially with 1 element). Args: tensor: The tensor to start from. layer: Origin layer of the tensor. Will be determined via tensor._keras_history if not provided. node_index: Origin node index of the tensor. Returns: List of input tensors. _keras_historyc3&K|]}|u ywN).0txs Y/home/dcms/DCMS/lib/python3.12/site-packages/tensorflow/python/keras/utils/layer_utils.py z$get_source_inputs..;s4!4s) hasattrr_inbound_nodesis_inputrflatten input_tensorsiterate_inboundget_source_inputsallappend)tensorlayer node_index_nodesource_tensorsprevious_sourcesr s @r rrs ) * M ]j!00E:q    8O    +D }} \\$,, --n*.*>*>*@% &%Q,VUJG! %A 4^4 4  ! !! $ %% c |r|y|r t|ryt|tr||vry|rdnd}||rdndz }|d|z }td|d|d|d |d ) z0Validates the correctness of a string-based arg.Nz`None`, za `Callable`, z or one of the following values: zThe z argument of layer z received an invalid value z. Allowed values are: .)callable isinstancestr ValueError) input_dataallowable_strings layer_namearg_name allow_noneallow_callables allowed_argss r validate_string_argr-@suJ& 8J/ *c"z5F'F !+:L$R?L.ds@@s!)idvaluesshapeas_listintsum)weightswunique_weights weight_shapesw_istandardized_weight_shapess r count_paramsr?Vs'..BqE1H.557..<=177??$=-=8E 34A.SCKqS .   S@%?@ @ AA /=. s"B!B B B +BBct|jjdk(rd}n|jsd}nd}|jj }g}|D]U}t |dkDs.print_rows} D 3v; / QCRy3 c&)nd -9Q< d cYq\CI-..d /  TNrz Model: "{}"r=c$ |j}|j}|jj }|j st|ddsd}n|j}|dz|zdz||g}|y#t$rd}Yst$rd}YwxYw) zQPrints a summary for a single layer. Args: layer: target layer. multiple?_is_graph_networkFz 0 (unused) ()N) output_shapeAttributeError RuntimeErrorname __class____name__builtgetattrr?)rr[r^cls_nameparamsrNrOrSs r print_layer_summaryz*print_summary..print_layer_summarys ''l ::D''H ;;wu.A5If!!#fTkH$s*L& AF fi   l ls A55 B BBc  |j}g}|jD]O}r|vr |jD]3\}}}}|j dj |j ||5Q|j }|jj} |sd} n|d} |dz| zdz||j| g} | t|dkDr-tdt|D]} ddd|| g} | yy#t$rd}YwxYw) zuPrints a summary for a single layer (including topological connections). Args: layer: target layer. rVz {}[{}][{}]r rrYrZrBN) r[r\rrrformatr^r_r`r?rMrL)rr[ connectionsr inbound_layerr tensor_indexrr^rcfirst_connectionrNrQrOrSrelevant_nodess r $print_layer_summary_with_connectionsz;print_summary..print_layer_summary_with_connectionssF  ''lK$$> D68<8L8L8N> 4-\1<..}/A/A:/;= >> > ::D''H $Q t h$l .Ffi  ;!QK()%!b"k!n-&)$%/  l s C== D  D _collected_trainable_weightszTotal params: {:,}zTrainable params: {:,}zNon-trainable params: {:,})printr_r`rX_nodes_by_depthr4rMrr keras_inputslayersrr7rgr^rLrr?rntrainable_weightsnon_trainable_weights)model line_lengthrOrRsequential_likenodes_by_depthnodesvrflagrr2 to_displayrermrrrQtrainable_countnon_trainable_countrSrls `` @@r print_summaryrgs H __-O  " "OO**113N E  a&1*#a&A+dll1Q4+<+<=>B   qje<< %(( D U] %od    #K+^I}1:;A3{Q';i channels_lastchannels_firstrBrN)rrB)rBrr) get_weightsrLr5reshaper0 transposer1 set_weights) denseprevious_feature_map_shapetarget_data_formatkernelbiasrQchr:original_fm_shapekis r !convert_dense_weights_data_formatrs, B BB B""$,&$ a ! Ja--*gaAa) !Q$<   1 2b <<I &b*gaAa) !Q$<   1 2b <<I &b::b277+E#F"HIF1a4L JVTN#rc^t|ddsy|jdk7xr|jdk7S)N_keras_api_namesF)zkeras.layers.Layer)rbr_keras_api_names_v1)rs r is_builtin_layerr6s= *D 1   $; ; ?  # #'> >@rc|tjtjfd}|_|S)a' Lightweight decorator for caching lazily constructed properties. When to use: This decorator provides simple caching with minimal overhead. It is designed for properties which are expensive to compute and static over the life of a class instance, and provides no mechanism for cache invalidation. Thus it is best suited for lazily exposing derived properties of other static data. For classes with custom getattr / setattr behavior (such as trackable objects), storing cache results as object attributes is not performant. Instead, a specialized cache can significantly reduce property lookup overhead. (While still allowing the decorated property to be lazily computed.) Consider the following class: ``` class MyClass(object): def __setattr__(self, key, value): # Some expensive class specific code # ... # ... super(MyClass, self).__setattr__(key, value) @property def thing(self): # `thing` is expensive to compute (and may not even be requested), so we # want to lazily compute it and then cache it. output = getattr(self, '_thing', None) if output is None: self._thing = output = compute_thing(self) return output ``` It's also worth noting that ANY overriding of __setattr__, even something as simple as: ``` def __setattr__(self, key, value): super(MyClass, self).__setattr__(key, value) ``` Slows down attribute assignment by nearly 10x. By contrast, replacing the definition of `thing` with the following sidesteps the expensive __setattr__ altogether: ''' @property @tracking.cached_per_instance def thing(self): # `thing` is expensive to compute (and may not even be requested), so we # want to lazily compute it and then cache it. return compute_thing(self) ''' Performance: The overhead for this decorator is ~0.4 us / call. A much lower overhead implementation (~0.085 us / call) can be achieved by using a custom dict type: ``` def dict_based_cache(f): class Cache(dict): __slots__ = () def __missing__(self, key): self[key] = output = f(key) return output return property(Cache().__getitem__) ``` However, that implementation holds class instances as keys, and as a result blocks garbage collection. (And modifying it to use weakref's as keys raises the lookup overhead to ~0.4 us) As a result, the WeakKeyDictionary implementation below turns out to be more prudent. Args: f: The function to cache. Returns: f decorated with simple caching behavior. cHj|}| |x|<}|Sr)get)itemoutputcachefs r wrappedz$cached_per_instance..wrappeds, YYt_F ~tW$eDkF Mr)weakrefWeakKeyDictionary functoolswrapsr)rrrs` @r cached_per_instancer@s=d  # # %% ??1 '- .rc#>Kt}|ddd}|r|j}t||vr |jt|t |drt |t s|n(t|ddxsg}|j|ddd|ryyw)z4Filter out empty Layer-like containers and uniquify.NrE _is_layerrr) setpopr3addrr#typerbextend) layer_listexistingto_visitobj sub_layerss r filter_empty_layer_containersrsU( " ( ,,.C #w( LLCsK C)> i3$/52jooj2&' s BBB)NN)FF)NNN)r)__doc__rrnumpyr0tensorflow.python.utilrrr-r?rrrrrrrr rsU 6'#T$)(- G,B"eT:J$$N@\~(r