\;j<&ddlZddlmZddlmZddlmZddlm Z ddl m Z ddl m Z gZeed  d d Zy) N) static_only) LayerHelper) templatedoc) ParamAttr)Assign)check_variable_and_dtypence)op_typec #td%it#t|dddgdt|ddgd|jdk7rt d|jd |j d } |j d } #j #j|| gd |j }i}#jr0#j #j|d gd |j }||d<#j|j}#j|j}#j|j}||d<||d<||d<||ng|d<|dk(rd}n-|dk(rd }n$|dk(r| J|}dg|z}dg|z}g}g}t|D]L}| ||z}|dz dkDr|j||f'd|z dkDr|j||fC|||<d||<Nt|rt|r|jd}|jd}|d}|d }|d ||d<|||d<|d |d zd z }|dz dkDr|j||fn&d|z dkDr|j||fn |||<d||<t|r t|rt|r!|jd}d||d<d||d<t|r!|jd}d||d<d||d<#fd} | tj | j#d|d<| tj |j#d|d<| tj |j#d|d<d}n t%d|d }n t'|}| }!t)d!t'||| || |!d"}"#j+d||||d#|"$||d zz S)&a :api_attr: Static Graph ${comment} Args: input (Tensor): Input tensor, 2-D tensor with shape [batch_size, dim], and data type is float32 or float64. label (Tensor): Input label, 2-D tensor with shape [batch_size, num_true_class], and data type is int64. num_total_classes (int):${num_total_classes_comment}. sample_weight (Tensor|None): A Tensor of shape [batch_size, 1] storing a weight for each sample. The default weight for each sample is 1.0. param_attr (ParamAttr|None): To specify the weight parameter attribute. Default: None, which means the default weight parameter property is used. See usage for details in :ref:`api_paddle_ParamAttr` . bias_attr (ParamAttr|None): To specify the bias parameter attribute. Default: None, which means the default bias parameter property is used. See usage for details in :ref:`api_paddle_ParamAttr` . num_neg_samples (int): ${num_neg_samples_comment}. name(str|None): For detailed information, please refer to :ref:`api_guide_Name` . Usually name is no need to set and None by default. sampler (str, optional): The sampler used to sample class from negative classes. It can be 'uniform', 'log_uniform' or 'custom_dist'. default: 'uniform'. custom_dist (nd.array|None): A numpy ndarray with size=num_total_classes. It is used when sampler is set to 'custom_dist'. custom_dist[i] is the probability of i-th class to be sampled. default: None. seed (int, optional): The seed used in sampler. Default 0, means no random seed. is_sparse(bool, optional): The flag indicating whether to use sparse update, the weight@GRAD and bias@GRAD will be changed to SelectedRows. Default False. Returns: Tensor: The output nce loss. Examples: .. code-block:: python >>> import paddle >>> import numpy as np >>> paddle.enable_static() >>> window_size = 5 >>> words = [] >>> for i in range(window_size): ... words.append(paddle.static.data( ... name='word_{0}'.format(i), shape=[-1, 1], dtype='int64')) >>> dict_size = 10000 >>> label_word = int(window_size / 2) + 1 >>> embs = [] >>> for i in range(window_size): ... if i == label_word: ... continue ... ... emb = paddle.static.nn.embedding(input=words[i], size=[dict_size, 32], ... param_attr='embed', is_sparse=True) ... embs.append(emb) >>> embs = paddle.concat(x=embs, axis=1) # concat from 4 * [(-1, 1, 32)] to (-1, 4, 32) >>> embs = paddle.reshape(x=embs, shape=(-1, 4 * 32)) # reshape to (batch_size = -1, dim = 4*32) >>> loss = paddle.static.nn.nce(input=embs, label=words[label_word], ... num_total_classes=dict_size, param_attr='nce.w_0', ... bias_attr='nce.b_0') # or use custom distribution >>> dist = np.array([0.05,0.5,0.1,0.3,0.05]) >>> loss = paddle.static.nn.nce(input=embs, label=words[label_word], ... num_total_classes=5, param_attr='nce.w_1', ... bias_attr='nce.b_1', ... num_neg_samples=3, ... sampler="custom_dist", ... custom_dist=dist) r inputfloat32float64labelint64z,The rank of `input` must be 2, but received .F)attrshapeis_biasdtypeTBias)rInputLabelWeight SampleWeightuniformr log_uniform custom_distg?cjt|j|jt |}d|_|S)N)rrrdefault_initializerT)create_parameterrrrr stop_gradient) numpy_arrayrethelpers ^G:\00. PROJECTS\API\Inventory\templateJSON\kerjaOCR\Lib\site-packages\paddle/static/nn/loss.py_init_by_numpy_arrayz!nce.._init_by_numpy_arraysF))[!''!''$*;$7 *C !%C JCustomDistProbsint32CustomDistAliasCustomDistAliasProbszUnsupported sampler type. zWWith sparse mode, if your models has only small parameter prefetch may cause speed down)num_total_classesnum_neg_samplesseedsampler is_sparseremote_prefetch)Cost SampleLogits SampleLabels)typeinputsoutputsattrs)r )rlocalsr ndim ValueErrorrr$ param_attrr bias_attr"create_variable_for_type_inferencerangeappendlenpopnparrayastype Exceptionintprint append_op)$r rr1 sample_weightrArBr2namer4r r3r5dimnum_true_classwr;bcost sample_logits sample_labelscustom_dist_len alias_probs_alias_bigslittlesi normal_probbiglittlebig_idxbig_probbig_leftr*r6r=r(s$ @r)r r !sq| +&( +FUGi-CUKUGgY> zzQ:5::,a H   ++a.C[[^N    #&kk  A F   # #!!$a(++ $  v  4 45;; 4 GD==EKK=PM==EKK=PMF7OF7OF8.;.G]RF>) M ! M !&&&+s_, &'A%a.?:KS 1$ Q ,-{"Q&;/0"- Qq ($iCL((1+C[[^F!fG1vH&,QiL # 'F6!9 1vq )A-H#~! Wh/0x!#23(0 W%"$w!$iCL$ t9((1+C#&LQ F3q6N w<[[^F&)L # "F6!9  %9 HH[ ! ( ( 3%  !%9 HHV  # #G ,%  !*> HH\ " ) )) 4* %&344o.O a !!23**  E  ))   ?Q& ''r+) NNNNNrNrF)numpyrHpaddle.base.frameworkrpaddle.base.layer_helperr+paddle.base.layers.layer_function_generatorrpaddle.base.param_attrrpaddle.nn.initializerrbase.data_feederr __all__r r+r)rmsc-1C,(8    U   d( d(r+