o ?e0@sdZddlmZddlmZddlmZddlmZddlmZddlmZddlm Z dd lm Z dd l m Z dd l m Z dd l mZdd lmZddlmZddZedgdGddde jZe eedddZdS)z#The Categorical distribution class.) constant_op)dtypes)ops) tensor_shape) array_ops)math_ops)nn_ops) random_ops) distribution)kullback_leibler)util) deprecation) tf_exportcCs|jjrn|jjrtj|tjd}ntd||j j duo,|j dd o,|j }|r9|j dd|j krn|t j |dt jf|jd9}t |dd}|t j||jd9}|j j durn|t|j dd||fS)z0Broadcasts the event or distribution parameters.dtypez8`value` should have integer `dtype` or `self.dtype` ({})N.)r is_integerZ is_floatingrcastrint32 TypeErrorformatshapendimsis_fully_definedrZ ones_likeZnewaxisZones set_shaper TensorShape)eventparams base_dtypeZshape_known_staticallyZ params_shaperp/home/www/facesmatcher.com/pyenv/lib/python3.10/site-packages/tensorflow/python/ops/distributions/categorical.py_broadcast_cat_event_and_params s*  r!zdistributions.Categorical)v1cseZdZdZejddddddejdddffdd Ze d d Z e d d Z e ddZ ddZ ddZddZddZd"ddZddZddZddZd d!ZZS)# Categoricalas Categorical distribution. The Categorical distribution is parameterized by either probabilities or log-probabilities of a set of `K` classes. It is defined over the integers `{0, 1, ..., K}`. The Categorical distribution is closely related to the `OneHotCategorical` and `Multinomial` distributions. The Categorical distribution can be intuited as generating samples according to `argmax{ OneHotCategorical(probs) }` itself being identical to `argmax{ Multinomial(probs, total_count=1) }`. #### Mathematical Details The probability mass function (pmf) is, ```none pmf(k; pi) = prod_j pi_j**[k == j] ``` #### Pitfalls The number of classes, `K`, must not exceed: - the largest integer representable by `self.dtype`, i.e., `2**(mantissa_bits+1)` (IEEE 754), - the maximum `Tensor` index, i.e., `2**31-1`. In other words, ```python K <= min(2**31-1, { tf.float16: 2**11, tf.float32: 2**24, tf.float64: 2**53 }[param.dtype]) ``` Note: This condition is validated only when `self.validate_args = True`. #### Examples Creates a 3-class distribution with the 2nd class being most likely. ```python dist = Categorical(probs=[0.1, 0.5, 0.4]) n = 1e4 empirical_prob = tf.cast( tf.histogram_fixed_width( dist.sample(int(n)), [0., 2], nbins=3), dtype=tf.float32) / n # ==> array([ 0.1005, 0.5037, 0.3958], dtype=float32) ``` Creates a 3-class distribution with the 2nd class being most likely. Parameterized by [logits](https://en.wikipedia.org/wiki/Logit) rather than probabilities. ```python dist = Categorical(logits=np.log([0.1, 0.5, 0.4]) n = 1e4 empirical_prob = tf.cast( tf.histogram_fixed_width( dist.sample(int(n)), [0., 2], nbins=3), dtype=tf.float32) / n # ==> array([0.1045, 0.5047, 0.3908], dtype=float32) ``` Creates a 3-class distribution with the 3rd class being most likely. The distribution functions can be evaluated on counts. ```python # counts is a scalar. p = [0.1, 0.4, 0.5] dist = Categorical(probs=p) dist.prob(0) # Shape [] # p will be broadcast to [[0.1, 0.4, 0.5], [0.1, 0.4, 0.5]] to match counts. counts = [1, 0] dist.prob(counts) # Shape [2] # p will be broadcast to shape [3, 5, 7, 3] to match counts. counts = [[...]] # Shape [5, 7, 3] dist.prob(counts) # Shape [5, 7, 3] ``` z 2019-01-01zThe TensorFlow Distributions library has moved to TensorFlow Probability (https://github.com/tensorflow/probability). You should update all references to use `tfp.distributions` instead of `tf.distributions`.T)Z warn_onceNFc stt}tj|||gd}tj|||d|d\|_|_|r&t|j|_|j d}|j durAtj |j dt jdd|_ntjddt|jd|_Wdn1s[wYtj|jd d} t|d durtj |jd jt jd d|_ntjd d| |j|_Wdn1swY|dd rtj|dd t jd d|_ntjd d| dd |_Wdn1swYWdn1swYtt|j|t j!||||j|jg|d dS)aInitialize Categorical distributions using class log-probabilities. Args: logits: An N-D `Tensor`, `N >= 1`, representing the log probabilities of a set of Categorical distributions. The first `N - 1` dimensions index into a batch of independent distributions and the last dimension represents a vector of logits for each class. Only one of `logits` or `probs` should be passed in. probs: An N-D `Tensor`, `N >= 1`, representing the probabilities of a set of Categorical distributions. The first `N - 1` dimensions index into a batch of independent distributions and the last dimension represents a vector of probabilities for each class. Only one of `logits` or `probs` should be passed in. dtype: The type of the event samples (default: int32). validate_args: Python `bool`, default `False`. When `True` distribution parameters are checked for validity despite possibly degrading runtime performance. When `False` invalid inputs may silently render incorrect outputs. allow_nan_stats: Python `bool`, default `True`. When `True`, statistics (e.g., mean, mode, variance) use the value "`NaN`" to indicate the result is undefined. When `False`, an exception is raised if one or more of the statistic's batch members are undefined. name: Python `str` name prefixed to Ops created by this class. valuesT)logitsprobs validate_argsZmultidimensionalnameNZ batch_rank)rr)r) logits_shaper event_size batch_shape)rZreparameterization_typer(allow_nan_stats parametersZ graph_parentsr))"dictlocalsr name_scopedistribution_utilZget_logits_and_probs_logits_probsZ#embed_check_categorical_event_shape get_shapeZwith_rank_at_leastrconvert_to_tensorrr _batch_rankrZrankrrZdimension_valuedimsvalue _event_sizerrconstantas_list_batch_shape_valsuperr#__init__r ZNOT_REPARAMETERIZED) selfr&r'rr(r/r)r0Zlogits_shape_staticr, __class__rr rAsn (      ( zCategorical.__init__cC|jS)z-Scalar `int32` tensor: the number of classes.)r<rBrrr r-zCategorical.event_sizecCrE)z Vector of coordinatewise logits.)r5rFrrr r&rGzCategorical.logitscCrE)z'Vector of coordinatewise probabilities.)r6rFrrr r'rGzCategorical.probscCs t|jSN)ridentityr?rFrrr _batch_shape_tensors zCategorical._batch_shape_tensorcCs|jddS)Nr)r&r7rFrrr _batch_shapeszCategorical._batch_shapecCstjgtjdS)Nr)rr=rrrFrrr _event_shape_tensorszCategorical._event_shape_tensorcCs tgSrH)rrrFrrr _event_shapes zCategorical._event_shapecCs|jjdkr |j}n t|jd|jg}|jjdkrtj ntj }t j ||||d}tt |t|g|gd}t||jS)Nr)seedZ output_dtyper)r&r7rrreshaper-rsizerZint64rr Z multinomialZ transposeconcatZbatch_shape_tensorrr)rBnrPZ logits_2dZ sample_dtypeZdrawsrrr _sample_n szCategorical._sample_ncCstj|dd}|jrtj|tjd}t||j|j j d\}}t |d|j f}t |dg}t t ||j |t |}tj|dd}t |t |S)Nkr+Z target_dtyperrZaxis)rr8r(r4"embed_check_integer_casting_closedrrr!r'rrrrQr<whereZ sequence_maskZ zeros_liker reduce_sumr)rBrVr'Zbatch_flattened_probsZbatch_flattened_kZ to_sum_overZbatch_flattened_cdfrrr _cdfs&   zCategorical._cdfcCsLtj|dd}|jrtj|tjd}t||j|j j d\}}t j ||d S)NrVr+rWrX)labelsr&) rr8r(r4rZrrr!r&rrrZ(sparse_softmax_cross_entropy_with_logits)rBrVr&rrr _log_prob.s  zCategorical._log_probcCstjt|j|jdd S)NrrY)rr\r log_softmaxr&r'rFrrr _entropy;szCategorical._entropycCs0tj|j|jd}t||j}||j|S)NrY)rZargmaxr&r9rrrr.)rBretrrr _mode?s zCategorical._moderH)__name__ __module__ __qualname____doc__r deprecatedrrrApropertyr-r&r'rJrKrLrMrUr]r_rarc __classcell__rrrCr r#:s:Y S      r#NcCsntj|d|j|jgd!t|jt|j}tjt|j|ddWdS1s0wYdS)aMCalculate the batched KL divergence KL(a || b) with a and b Categorical. Args: a: instance of a Categorical distribution object. b: instance of a Categorical distribution object. name: (optional) Name to use for created operations. default is "kl_categorical_categorical". Returns: Batchwise KL(a || b) Zkl_categorical_categoricalr$rrYN)rr3r&rr`rr\Zsoftmax)abr)Zdelta_log_probs1rrr _kl_categorical_categoricalFs   $rmrH)rgZtensorflow.python.frameworkrrrrZtensorflow.python.opsrrrr Z#tensorflow.python.ops.distributionsr r r r4Ztensorflow.python.utilr Z tensorflow.python.util.tf_exportrr! Distributionr#Z RegisterKLrmrrrr s*