o i e@sJdZddlZddlmmZddlmZedgdGdddeZ dS)zThe DetermisticRandomTestTool. (from www.tensorflow.org/guide/migrate/validate_correctness) is a tool used to make random number generation semantics match between TF1.x graphs/sessions and eager execution. N) keras_exportz'keras.utils.DeterministicRandomTestTool)v1c@sBeZdZdZd defddZeddZejd dZd d Z d S)DeterministicRandomTestToolaDeterministicRandomTestTool is a testing tool. This tool is used to validate random number generation semantics match between TF1.x graphs/sessions and eager execution. This is useful when you are migrating from TF 1.x to TF2 and need to make sure your computation is still happening correctly along the way. See the validating correctness migration guide for more info: https://www.tensorflow.org/guide/migrate/validate_correctness The following DeterministicRandomTestTool object provides a context manager scope() that can make stateful random operations use the same seed across both TF1 graphs/sessions and eager execution,The tool provides two testing modes: - constant which uses the same seed for every single operation no matter how many times it has been called and, - num_random_ops which uses the number of previously-observed stateful random operations as the operation seed. The num_random_ops mode serves as a more sensitive validation check than the constant mode. It ensures that the random numbers initialization does not get accidentaly reused.(for example if several weights take on the same initializations), you can use the num_random_ops mode to avoid this. In the num_random_ops mode, the generated random numbers will depend on the ordering of random ops in the program. This applies both to the stateful random operations used for creating and initializing variables, and to the stateful random operations used in computation (such as for dropout layers). Args: mode: Set mode to 'constant' or 'num_random_ops'. Defaults to 'constant'. seed: The random seed to use. *constantseedcCsL|dvr tdd|tjtjjjj|_||_ ||_ d|_ t |_ dS)N>rnum_random_opsz1Mode arg must be 'constant' or 'num_random_ops'. zGot: r) ValueErrorsysmodulestfcompatrget_seed __module__seed_implementation_mode_seedoperation_seedset_observed_seeds)selfrmoderY/var/www/myenv/lib/python3.10/site-packages/keras/src/legacy_tf_layers/migration_utils.py__init__5s z$DeterministicRandomTestTool.__init__cCs|jSN_operation_seedrrrrrAsz*DeterministicRandomTestTool.operation_seedcCs ||_dSrr)rvaluerrrrEs cs6tjjfdd}tjjjjjj j d|dS)zset random seed.csxj}jdkrtj|n(|jvr*tdd|dddddd d j|jd 7_j|fS) aWraps TF get_seed to make deterministic random generation easier. This makes a variable's initialization (and calls that involve random number generation) depend only on how many random number generations were used in the scope so far, rather than on how many unrelated operations the graph contains. Returns: Random seed tuple. rzBThis `DeterministicRandomTestTool` object is trying to re-use the zalready-used operation seed z. z.It cannot guarantee random numbers will match z2between eager and sessions when an operation seed zis reused. You most likely set z2`operation_seed` explicitly but used a value that z1caused the naturally-incrementing operation seed z/sequences to overlap with an already-used seed.) rrr randomset_seedrr addr)_op_seedrrr _get_seedNs0     z4DeterministicRandomTestTool.scope.._get_seedr)wraps) r r!r"rr rtestmockpatchobjectr)rr&rrrscopeIs  $z!DeterministicRandomTestTool.scopeN)rr) __name__r __qualname____doc__intrpropertyrsetterr,rrrrrs#   r) r/r tensorflow.compat.v2r v2r tensorflow.python.util.tf_exportrr+rrrrrs