ó ¾÷Xc@@sÞddlmZddlZddlZddlmZddlm Z ddlm Z de fd„ƒYZ d e fd „ƒYZ d e fd „ƒYZd e fd„ƒYZde fd„ƒYZde fd„ƒYZde fd„ƒYZde fd„ƒYZde fd„ƒYZde fd„ƒYZdd„Zdd„Zdd„Zdd„Zdd„Ze ZZeZZeZ eZ!Z"eZ#Z$eZ%eZ&eZ'd d!„Z(d"„Z)dd#„Z*d$„Z+dS(%i(tabsolute_importNi(tbackend(tserialize_keras_object(tdeserialize_keras_objectt InitializercB@s2eZdZdd„Zd„Zed„ƒZRS(sFInitializer base class: all initializers inherit from this class. cC@s t‚dS(N(tNotImplementedError(tselftshapetdtype((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt__call__ scC@siS(N((R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt get_configscC@s ||S(N((tclstconfig((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt from_configsN(t__name__t __module__t__doc__tNoneR R t classmethodR (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR s  tZeroscB@seZdZdd„ZRS(s4Initializer that generates tensors initialized to 0.cC@stjdd|d|ƒS(NiRR(tKtconstant(RRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR sN(RRRRR (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyRstOnescB@seZdZdd„ZRS(s4Initializer that generates tensors initialized to 1.cC@stjdd|d|ƒS(NiRR(RR(RRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR "sN(RRRRR (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyRstConstantcB@s/eZdZdd„Zdd„Zd„ZRS(s“Initializer that generates tensors initialized to a constant value. # Arguments value: float; the value of the generator tensors. icC@s ||_dS(N(tvalue(RR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt__init__-scC@stj|jd|d|ƒS(NRR(RRR(RRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR 0scC@si|jd6S(NR(R(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR 3sN(RRRRRR R (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR&s  t RandomNormalcB@s5eZdZdddd„Zdd„Zd„ZRS(smInitializer that generates tensors with a normal distribution. # Arguments 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 seed the random generator. ggš™™™™™©?cC@s||_||_||_dS(N(tmeantstddevtseed(RRRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyRBs  cC@s(tj||j|jd|d|jƒS(NRR(Rt random_normalRRR(RRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR GscC@s"i|jd6|jd6|jd6S(NRRR(RRR(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR Ks  N(RRRRRR R (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR7s  t RandomUniformcB@s5eZdZdddd„Zdd„Zd„ZRS(s¢Initializer that generates tensors with a uniform distribution. # Arguments minval: A python scalar or a scalar tensor. Lower bound of the range of random values to generate. maxval: A python scalar or a scalar tensor. Upper bound of the range of random values to generate. Defaults to 1 for float types. seed: A Python integer. Used to seed the random generator. gš™™™™™©¿gš™™™™™©?cC@s||_||_||_dS(N(tminvaltmaxvalR(RR R!R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR^s  cC@s(tj||j|jd|d|jƒS(NRR(Rtrandom_uniformR R!R(RRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR cscC@s"i|jd6|jd6|jd6S(NR R!R(R R!R(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR gs  N(RRRRRR R (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyRSs  tTruncatedNormalcB@s5eZdZdddd„Zdd„Zd„ZRS(slInitializer 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. # Arguments 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 seed the random generator. ggš™™™™™©?cC@s||_||_||_dS(N(RRR(RRRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyRs  cC@s(tj||j|jd|d|jƒS(NRR(Rttruncated_normalRRR(RRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR „scC@s"i|jd6|jd6|jd6S(NRRR(RRR(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR ˆs  N(RRRRRR R (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR#os tVarianceScalingcB@s8eZdZddddd„Zdd„Zd„ZRS(sËInitializer capable of adapting its scale to the shape of weights. With `distribution="normal"`, samples are drawn from a truncated normal distribution centered on zero, with `stddev = sqrt(scale / n)` where n is: - number of input units in the weight tensor, if mode = "fan_in" - number of output units, if mode = "fan_out" - average of the numbers of input and output units, if mode = "fan_avg" With `distribution="uniform"`, samples are drawn from a uniform distribution within [-limit, limit], with `limit = sqrt(3 * scale / n)`. # Arguments scale: Scaling factor (positive float). mode: One of "fan_in", "fan_out", "fan_avg". distribution: Random distribution to use. One of "normal", "uniform". seed: A Python integer. Used to seed the random generator. # Raises ValueError: In case of an invalid value for the "scale", mode" or "distribution" arguments. gð?tfan_intnormalcC@s©|dkrtd|ƒ‚n|jƒ}|dddhkrQtd|ƒ‚n|jƒ}|ddhkrtd |ƒ‚n||_||_||_||_dS( Ngs&`scale` must be a positive float. Got:R&tfan_outtfan_avgsPInvalid `mode` argument: expected on of {"fan_in", "fan_out", "fan_avg"} but gotR'tuniformsNInvalid `distribution` argument: expected one of {"normal", "uniform"} but got(t ValueErrortlowertscaletmodet distributionR(RR-R.R/R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR¨s        cC@st|ƒ\}}|j}|jdkr@|td|ƒ:}nF|jdkre|td|ƒ:}n!|tdt||ƒdƒ:}|jdkrÆtj|ƒ}tj |d|d|d|j ƒStjd |ƒ}tj || |d|d|j ƒSdS( NR&gð?R(iR'gRRg@( t _compute_fansR-R.tmaxtfloatR/tnptsqrtRR$RR"(RRRR&R(R-Rtlimit((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR ½s !cC@s,i|jd6|jd6|jd6|jd6S(NR-R.R/R(R-R.R/R(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR Ïs    N(RRRRRR R (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR%s  t OrthogonalcB@s2eZdZddd„Zdd„Zd„ZRS(sInitializer that generates a random orthogonal matrix. # Arguments gain: Multiplicative factor to apply to the orthogonal matrix. seed: A Python integer. Used to seed the random generator. # References Saxe et al., http://arxiv.org/abs/1312.6120 gð?cC@s||_||_dS(N(tgainR(RR7R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyRãs c C@sìd}x|d D]}||9}qW|d}||f}|jdk r`tjj|jƒntjjdd|ƒ}tjj|dtƒ\}} } |j|kr®|n| } | j |ƒ} |j | d|d…d|d…fS(Niiÿÿÿÿggð?t full_matricesi( RRR3trandomR'tlinalgtsvdtFalseRtreshapeR7( RRRtnum_rowstdimtnum_colst flat_shapetatut_tvtq((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR çs  !cC@si|jd6|jd6S(NR7R(R7R(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR ös N(RRRRRR R (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR6Øs  tIdentitycB@s/eZdZdd„Zdd„Zd„ZRS(s°Initializer that generates the identity matrix. Only use for square 2D matrices. # Arguments gain: Multiplicative factor to apply to the identity matrix. gð?cC@s ||_dS(N(R7(RR7((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyRscC@sQt|ƒdks&|d|dkr5tdƒ‚n|jtj|dƒSdS(NiiisDIdentity matrix initializer can only be used for 2D square matrices.(tlenR+R7R3tidentity(RRR((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR s&cC@si|jd6S(NR7(R7(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR sN(RRRRRR R (((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyRGýs  c C@stddddddd|ƒS(sÅLeCun uniform initializer. It draws samples from a uniform distribution within [-limit, limit] where `limit` is `sqrt(3 / fan_in)` where `fan_in` is the number of input units in the weight tensor. # Arguments seed: A Python integer. Used to seed the random generator. # Returns An initializer. # References LeCun 98, Efficient Backprop, http://yann.lecun.com/exdb/publis/pdf/lecun-98b.pdf R-gð?R.R&R/R*R(R%(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt lecun_uniforms c C@stddddddd|ƒS(sDGlorot normal initializer, also called Xavier normal initializer. It draws samples from a truncated normal distribution centered on 0 with `stddev = sqrt(2 / (fan_in + fan_out))` where `fan_in` is the number of input units in the weight tensor and `fan_out` is the number of output units in the weight tensor. # Arguments seed: A Python integer. Used to seed the random generator. # Returns An initializer. # References Glorot & Bengio, AISTATS 2010 http://jmlr.org/proceedings/papers/v9/glorot10a/glorot10a.pdf R-gð?R.R)R/R'R(R%(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt glorot_normal-s c C@stddddddd|ƒS(sIGlorot uniform initializer, also called Xavier uniform initializer. It draws samples from a uniform distribution within [-limit, limit] where `limit` is `sqrt(6 / (fan_in + fan_out))` where `fan_in` is the number of input units in the weight tensor and `fan_out` is the number of output units in the weight tensor. # Arguments seed: A Python integer. Used to seed the random generator. # Returns An initializer. # References Glorot & Bengio, AISTATS 2010 http://jmlr.org/proceedings/papers/v9/glorot10a/glorot10a.pdf R-gð?R.R)R/R*R(R%(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pytglorot_uniformEs c C@stddddddd|ƒS(sHe normal initializer. It draws samples from a truncated normal distribution centered on 0 with `stddev = sqrt(2 / fan_in)` where `fan_in` is the number of input units in the weight tensor. # Arguments seed: A Python integer. Used to seed the random generator. # Returns An initializer. # References He et al., http://arxiv.org/abs/1502.01852 R-g@R.R&R/R'R(R%(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt he_normal]s c C@stddddddd|ƒS(s¤He uniform variance scaling initializer. It draws samples from a uniform distribution within [-limit, limit] where `limit` is `sqrt(6 / fan_in)` where `fan_in` is the number of input units in the weight tensor. # Arguments seed: A Python integer. Used to seed the random generator. # Returns An initializer. # References He et al., http://arxiv.org/abs/1502.01852 R-g@R.R&R/R*R(R%(R((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt he_uniformss t channels_lastcC@s t|ƒdkr)|d}|d}nÚt|ƒdddhkrÓ|dkr‚tj|dƒ}|d|}|d|}q|dkrÀtj|d ƒ}|d |}|d |}qtd |ƒ‚n0tjtj|ƒƒ}tjtj|ƒƒ}||fS( sóComputes the number of input and output units for a weight shape. # Arguments shape: Integer shape tuple. data_format: Image data format to use for convolution kernels. Note that all kernels in Keras are standardized on the `channels_last` ordering (even when inputs are set to `channels_first`). # Returns A tuple of scalars, `(fan_in, fan_out)`. # Raises ValueError: in case of invalid `data_format` argument. iiiiiitchannels_firstROiþÿÿÿiÿÿÿÿsInvalid data_format: (RHR3tprodR+R4(Rt data_formatR&R(treceptive_field_size((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyR0—s     cC@s t|ƒS(N(R(t initializer((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt serialize¿scC@st|dtƒd|ddƒS(Ntmodule_objectstcustom_objectstprintable_module_nameRT(Rtglobals(R RW((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyt deserializeÃs  cC@srt|tƒrt|ƒSt|tjƒrOit|ƒd6id6}t|ƒSt|ƒr_|Std|ƒ‚dS(Nt class_nameR s+Could not interpret initializer identifier:(t isinstancetdictRZtsixt string_typeststrtcallableR+(t identifierR ((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pytgetÊs   (,t __future__RtnumpyR3R^tRRtutils.generic_utilsRRtobjectRRRRRRR#R%R6RGRRJRKRLRMRNtzerotzerostonetonesRR*R"R'RR$RIt orthogonalR0RURZRc(((s1/tmp/pip-build-isqEY4/keras/keras/initializers.pyts@  !H%          (