0hINdZddlmcmZddlmZejjjZ ejjjZ ejjjZ ejjjZejjj Zejjj$j&Zejjj*Zejjj.Zeddgde edd gde ed d gde ed gdeed dgdeeddgdeedgdeedgdeegdGddejjj2ZegdGddejjj6ZeddgGddejjj:ZedgGdd ejjjZed!gGd"d#ejjjZ ed$gGd%d&ejjjZ!ed'gGd(d)ejjjZ"y)*zKeras initializers for TF 1.N) keras_exportzkeras.initializers.Zeroszkeras.initializers.zerosT)v1allow_multiple_exportszkeras.initializers.Oneszkeras.initializers.oneszkeras.initializers.Constantzkeras.initializers.constantz"keras.initializers.VarianceScalingzkeras.initializers.Orthogonalzkeras.initializers.orthogonalzkeras.initializers.Identityzkeras.initializers.identityz!keras.initializers.glorot_uniformz keras.initializers.glorot_normal)zkeras.initializers.RandomNormalz keras.initializers.random_normalzkeras.initializers.normal)rc@eZdZdZdddej ffd ZxZS) RandomNormalaInitializer that generates a normal distribution. Args: mean: a python scalar or a scalar tensor. Mean of the random values to generate. stddev: a python scalar or a scalar tensor. Standard deviation of the random values to generate. seed: A Python integer. Used to create random seeds. See `tf.compat.v1.set_random_seed` for behavior. dtype: Default data type, used if no `dtype` argument is provided when calling the initializer. Only floating point types are supported. @compatibility(TF2) Although it is a legacy compat.v1 api, `tf.compat.v1.keras.initializers.RandomNormal` is compatible with eager execution and `tf.function`. To switch to native TF2, switch to using `tf.keras.initializers.RandomNormal` (not from `compat.v1`) and if you need to change the default dtype use `tf.keras.backend.set_floatx(float_dtype)` or pass the dtype when calling the initializer, rather than passing it when constructing the initializer. Random seed behavior: Also be aware that if you pass a seed to the TF2 initializer API it will reuse that same seed for every single initialization (unlike the TF1 initializer) #### Structural Mapping to Native TF2 Before: ```python initializer = tf.compat.v1.keras.initializers.RandomNormal( mean=mean, stddev=stddev, seed=seed, dtype=dtype) weight_one = tf.Variable(initializer(shape_one)) weight_two = tf.Variable(initializer(shape_two)) ``` After: ```python initializer = tf.keras.initializers.RandomNormal( mean=mean, # seed=seed, # Setting a seed in the native TF2 API # causes it to produce the same initializations # across multiple calls of the same initializer. stddev=stddev) weight_one = tf.Variable(initializer(shape_one, dtype=dtype)) weight_two = tf.Variable(initializer(shape_two, dtype=dtype)) ``` #### How to Map Arguments | TF1 Arg Name | TF2 Arg Name | Note | | :---------------- | :-------------- | :------------------------- | | `mean` | `mean` | No change to defaults | | `stddev` | `stddev` | No change to defaults | | `seed` | `seed` | Different random number generation | : : : semantics (to change in a : : : : future version). If set, the TF2 version : : : : will use stateless random number : : : : generation which will produce the exact : : : : same initialization even across multiple : : : : calls of the initializer instance. the : : : : `compat.v1` version will generate new : : : : initializations each time. Do not set : : : : a seed if you need different : : : : initializations each time. Instead : : : : either set a global tf seed with : : : : `tf.random.set_seed` if you need : : : : determinism, or initialize each weight: : : : with a separate initializer instance : : : : and a different seed. : | `dtype` | `dtype` | The TF2 native api only takes it | : : : as a `__call__` arg, not a constructor arg. : | `partition_info` | - | (`__call__` arg in TF1) Not supported | #### Example of fixed-seed behavior differences `compat.v1` Fixed seed behavior: >>> initializer = tf.compat.v1.keras.initializers.RandomNormal(seed=10) >>> a = initializer(shape=(2, 2)) >>> b = initializer(shape=(2, 2)) >>> tf.reduce_sum(a - b) == 0 After: >>> initializer = tf.keras.initializers.RandomNormal(seed=10) >>> a = initializer(shape=(2, 2)) >>> b = initializer(shape=(2, 2)) >>> tf.reduce_sum(a - b) == 0 @end_compatibility 皙?Nc,t|||||y)Nmeanstddevseeddtypesuper__init__selfr r rr __class__s c/var/www/html/engine/venv/lib/python3.12/site-packages/tf_keras/src/initializers/initializers_v1.pyrzRandomNormal.__init__s d6EJ__name__ __module__ __qualname____doc__tffloat32r __classcell__rs@rrr@s%gR 4rzzKKrr)z keras.initializers.RandomUniformz!keras.initializers.random_uniformzkeras.initializers.uniformc@eZdZdZdddej ffd ZxZS) RandomUniforma?Initializer that generates tensors with a uniform distribution. Args: minval: A python scalar or a scalar tensor. Lower bound of the range of random values to generate. Defaults to `-0.05`. maxval: A python scalar or a scalar tensor. Upper bound of the range of random values to generate. Defaults to `0.05`. seed: A Python integer. Used to create random seeds. See `tf.compat.v1.set_random_seed` for behavior. dtype: Default data type, used if no `dtype` argument is provided when calling the initializer. @compatibility(TF2) Although it is a legacy `compat.v1` api, `tf.compat.v1.keras.initializers.RandomUniform` is compatible with eager execution and `tf.function`. To switch to native TF2, switch to using `tf.keras.initializers.RandomUniform` (not from `compat.v1`) and if you need to change the default dtype use `tf.keras.backend.set_floatx(float_dtype)` or pass the dtype when calling the initializer, rather than passing it when constructing the initializer. Random seed behavior: Also be aware that if you pass a seed to the TF2 initializer API it will reuse that same seed for every single initialization (unlike the TF1 initializer) #### Structural Mapping to Native TF2 Before: ```python initializer = tf.compat.v1.keras.initializers.RandomUniform( minval=minval, maxval=maxval, seed=seed, dtype=dtype) weight_one = tf.Variable(initializer(shape_one)) weight_two = tf.Variable(initializer(shape_two)) ``` After: ```python initializer = tf.keras.initializers.RandomUniform( minval=minval, maxval=maxval, # seed=seed, # Setting a seed in the native TF2 API # causes it to produce the same initializations # across multiple calls of the same initializer. ) weight_one = tf.Variable(initializer(shape_one, dtype=dtype)) weight_two = tf.Variable(initializer(shape_two, dtype=dtype)) ``` #### How to Map Arguments | TF1 Arg Name | TF2 Arg Name | Note | | :---------------- | :-------------- | :------------------------- | | `minval` | `minval` | No change to defaults | | `maxval` | `maxval` | No change to defaults | | `seed` | `seed` | Different random number generation | : : : semantics (to change in a : : : : future version). If set, the TF2 version : : : : will use stateless random number : : : : generation which will produce the exact : : : : same initialization even across multiple : : : : calls of the initializer instance. the : : : : `compat.v1` version will generate new : : : : initializations each time. Do not set : : : : a seed if you need different : : : : initializations each time. Instead : : : : either set a global tf seed with : : : `tf.random.set_seed` if you need : : : : determinism, or initialize each weight : : : : with a separate initializer instance : : : : and a different seed. : | `dtype` | `dtype` | The TF2 native api only takes it | : : : as a `__call__` arg, not a constructor arg. : | `partition_info` | - | (`__call__` arg in TF1) Not supported | #### Example of fixed-seed behavior differences `compat.v1` Fixed seed behavior: >>> initializer = tf.compat.v1.keras.initializers.RandomUniform(seed=10) >>> a = initializer(shape=(2, 2)) >>> b = initializer(shape=(2, 2)) >>> tf.reduce_sum(a - b) == 0 After: >>> initializer = tf.keras.initializers.RandomUniform(seed=10) >>> a = initializer(shape=(2, 2)) >>> b = initializer(shape=(2, 2)) >>> tf.reduce_sum(a - b) == 0 @end_compatibility gr Nc,t|||||y)N)minvalmaxvalrrr)rr$r%rrrs rrzRandomUniform.__init__)s vDNrrr s@rr"r"s%jX$Dt2::OOrr"z"keras.initializers.TruncatedNormalz#keras.initializers.truncated_normalc@eZdZdZdddej ffd ZxZS)TruncatedNormalaInitializer that generates a truncated normal distribution. These values are similar to values from a `random_normal_initializer` except that values more than two standard deviations from the mean are discarded and re-drawn. This is the recommended initializer for neural network weights and filters. Args: mean: a python scalar or a scalar tensor. Mean of the random values to generate. stddev: a python scalar or a scalar tensor. Standard deviation of the random values to generate. seed: A Python integer. Used to create random seeds. See `tf.compat.v1.set_random_seed` for behavior. dtype: Default data type, used if no `dtype` argument is provided when calling the initializer. Only floating point types are supported. @compatibility(TF2) Although it is a legacy compat.v1 api, `tf.compat.v1.keras.initializers.TruncatedNormal` is compatible with eager execution and `tf.function`. To switch to native TF2, switch to using `tf.keras.initializers.TruncatedNormal` (not from `compat.v1`) and if you need to change the default dtype use `tf.keras.backend.set_floatx(float_dtype)` or pass the dtype when calling the initializer, rather than passing it when constructing the initializer. Random seed behavior: Also be aware that if you pass a seed to the TF2 initializer API it will reuse that same seed for every single initialization (unlike the TF1 initializer) #### Structural Mapping to Native TF2 Before: ```python initializer = tf.compat.v1.keras.initializers.TruncatedNormal( mean=mean, stddev=stddev, seed=seed, dtype=dtype) weight_one = tf.Variable(initializer(shape_one)) weight_two = tf.Variable(initializer(shape_two)) ``` After: ```python initializer = tf.keras.initializers.TruncatedNormal( mean=mean, # seed=seed, # Setting a seed in the native TF2 API # causes it to produce the same initializations # across multiple calls of the same initializer. stddev=stddev) weight_one = tf.Variable(initializer(shape_one, dtype=dtype)) weight_two = tf.Variable(initializer(shape_two, dtype=dtype)) ``` #### How to Map Arguments | TF1 Arg Name | TF2 Arg Name | Note | | :---------------- | :-------------- | :------------------------- | | `mean` | `mean` | No change to defaults | | `stddev` | `stddev` | No change to defaults | | `seed` | `seed` | Different random number generation | : : : semantics (to change in a : : : : future version). If set, the TF2 version : : : : will use stateless random number : : : : generation which will produce the exact : : : : same initialization even across multiple : : : : calls of the initializer instance. the : : : : `compat.v1` version will generate new : : : : initializations each time. Do not set : : : : a seed if you need different : : : : initializations each time. Instead : : : : either set a global tf seed with : : : `tf.random.set_seed` if you need : : : : determinism, or initialize each weight : : : : with a separate initializer instance : : : : and a different seed. : | `dtype` | `dtype` | The TF2 native api only takes it | : : : as a `__call__` arg, not a constructor arg. : | `partition_info` | - | (`__call__` arg in TF1) Not supported | #### Example of fixed-seed behavior differences `compat.v1` Fixed seed behavior: >>> initializer = tf.compat.v1.keras.initializers.TruncatedNormal(seed=10) >>> a = initializer(shape=(2, 2)) >>> b = initializer(shape=(2, 2)) >>> tf.reduce_sum(a - b) == 0 After: >>> initializer = tf.keras.initializers.TruncatedNormal(seed=10) >>> a = initializer(shape=(2, 2)) >>> b = initializer(shape=(2, 2)) >>> tf.reduce_sum(a - b) == 0 @end_compatibility rr Nc,t|||||y)aMInitializer that generates a truncated normal distribution. Args: mean: a python scalar or a scalar tensor. Mean of the random values to generate. stddev: a python scalar or a scalar tensor. Standard deviation of the random values to generate. seed: A Python integer. Used to create random seeds. See `tf.compat.v1.set_random_seed` for behavior. dtype: Default data type, used if no `dtype` argument is provided when calling the initializer. Only floating point types are supported. r Nrrs rrzTruncatedNormal.__init__s d6EJrrr s@rr'r'-s%l\ 4rzzKKrr'zkeras.initializers.lecun_normalc&eZdZdfd ZdZxZS) LecunNormalc,t|ddd|y)N?fan_intruncated_normalscalemode distributionrrrrrs rrzLecunNormal.__init__ H3ED  rcd|jiSNrrrs r get_configzLecunNormal.get_config ""rNrrrrr9rr s@rr*r*  #rr*z keras.initializers.lecun_uniformc&eZdZdfd ZdZxZS) LecunUniformc,t|ddd|y)Nr,r-uniformr/rr3s rrzLecunUniform.__init__ H94  rcd|jiSr6r7r8s rr9zLecunUniform.get_configr:rr;r<r s@rr?r?r=rr?zkeras.initializers.he_normalc&eZdZdfd ZdZxZS)HeNormalc,t|ddd|y)N@r-r.r/rr3s rrzHeNormal.__init__r4rcd|jiSr6r7r8s rr9zHeNormal.get_configr:rr;r<r s@rrErEr=rrEzkeras.initializers.he_uniformc&eZdZdfd ZdZxZS) HeUniformc,t|ddd|y)NrGr-rAr/rr3s rrzHeUniform.__init__rBrcd|jiSr6r7r8s rr9zHeUniform.get_configr:rr;r<r s@rrJrJr=rrJ)#rtensorflow.compat.v2compatv2r tensorflow.python.util.tf_exportrrzeros_initializer_v1_zeros_initializerones_initializer_v1_ones_initializerconstant_initializer_v1_constant_initializervariance_scaling_initializer _v1_variance_scaling_initializerorthogonal_initializer_v1_orthogonal_initializer initializersidentity _v1_identityglorot_uniform_initializer_v1_glorot_uniform_initializerglorot_normal_initializer_v1_glorot_normal_initializerrandom_normal_initializerrrandom_uniform_initializerr"truncated_normal_initializerr'r*r?rErJrrrfs6#"!: 66yy||4499<<<<#%99<<#L#L YY\\@@yy||((11 !#!H!H " F F "$>? !#<= %'DE ,-d"$ ')HI %'DE +,T " *+D! kK299<<99kKkK\nOBIILL;;nOnOb,- }Kbiill??}K  }K@345#")),,;;#6#456#299<<<<#7#012#ryy||88#3#123# 99#4#r