\;je,ddlmZmZmZmZddlmZddlmZm Z ddl m Z ddl m Z ddlmZmZmZgZedd d d dd ZddZdZddZy))_C_ops _legacy_C_opsin_dynamic_modetensor) deprecated) check_typecheck_variable_and_dtype) LayerHelper)Variable)convert_np_dtype_to_dtype_corein_dynamic_or_pir_modez2.5.2zpaddle.diag_embedz>> import paddle >>> lengths = paddle.to_tensor([10, 9, 8]) >>> mask = paddle.nn.functional.sequence_mask(lengths) >>> print(mask) Tensor(shape=[3, 10], dtype=int64, place=Place(cpu), stop_gradient=True, [[1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1, 1, 1, 0], [1, 1, 1, 1, 1, 1, 1, 1, 0, 0]]) N out_dtypemaxlenT sequence_maskdtypeX MaxLenTensorYtypeinputsoutputsattrs)r )r isinstancerVarDescVarTyper eagerTensorrr stop_gradientr locals"create_variable_for_type_inferencer"r append_op)xrr"namer*outhelperr(s rr r +s#x%!5!56.u5E  &$**"3"34$e,#11!VDeD$eXv>#11!TBEB $C J  5FH 5F  3 3% 3 @CA3ZF #)) $E  fh '%+F> "$E(O  Vc3ZuC Jrc|jdk7r td|j|jk7r tdtrtj||St d it }t|dddgdt|dddgd|j|j }|jd||d d |i |S)a To be used after beam search. After beam search, we get selected ids at each time step and the corresponding parents in the search tree. Both ids and parents have the layout :attr:`[max_time, batch_size, beam_size]`. Then :attr:`gather_tree` is used to backtrace from the last time step and generate the full sequences by collecting selected ids. Here is an example: .. code-block:: text Given: ids = [[[2 2] [6 1]] [[3 9] [6 1]] [[0 1] [9 0]]] parents = [[[0 0] [1 1]] [[1 0] [1 0]] [[0 0] [0 1]]] Then: gather_tree(ids, parents) = [[[2 2] [1 6]] [[3 3] [6 1]] [[0 1] [9 0]]] Args: ids(Tensor): A Tensor with shape :attr:`[length, batch_size, beam_size]` and data type :attr:`int32` or :attr:`int64`. It contains the selected ids of all time steps. parents(Tensor): A Tensor with the same shape and data type as :attr:`ids`, It contains the parents corresponding to selected ids when searching among beams. Returns: A Tensor with the same shape and data type as :attr:`ids`. \ It contains the full sequences. The sequences are collected from \ :attr:`ids` by backtracing according to :attr:`parents`. Examples: .. code-block:: python >>> import paddle >>> ids = paddle.to_tensor([[[2, 2], [6, 1]], [[3, 9], [6, 1]], [[0, 1], [9, 0]]]) >>> parents = paddle.to_tensor([[[0, 0], [1, 1]], [[1, 0], [1, 0]], [[0, 0], [0, 1]]]) >>> final_sequences = paddle.nn.functional.gather_tree(ids, parents) >>> [[[2, 2], [1, 6]], [[3, 3], [6, 1]], [[0, 1], [9, 0]]] >>> final_sequences = paddle.nn.functional.gather_tree(ids, parents) >>> print(final_sequences) Tensor(shape=[3, 2, 2], dtype=int64, place=Place(cpu), stop_gradient=True, [[[2, 2], [1, 6]], [[3, 3], [6, 1]], [[0, 1], [9, 0]]]) rzLThe input ids must be a 3D tensor with shape [length, batch_size, beam_size]z4The ids's shape must be the same as parents' shape. gather_treeidsint32int64parentsr!)IdsParentsOut)r'r(r))r9) ndim ValueErrorrrr9r r1r r2r"r3)r:r=r7r6s rr9r9sN xx1} Z   xx7<<OPP!!#w//7fh7 egw-?O Y' 2M 77cii7H73CL   rc |dvrtd|dtrtj||||St dit }t |dgddt|dtdt|dtd|j|j }t|ts td |jdd |id |i|||d |S)a **Temporal Shift Operator** Calculate the temporal shifting features for Input(X). Input(X) should be in shape of [N*T, C, H, W] or [N*T, H, W, C], while N is the batch size, T is the temporal segment number specified by :attr:`seg_num`, C is the channel number, H and W is the height and width of features. Temporal Shifting is calculated as follows when data format is NCHW: Step 1: Reshape Input(X) to [N, T, C, H, W]. Step 2: Pad 0 to reshaping result in the 2nd(T) dimension with padding width as 1 on each side, padding result will be in shape of [N, T+2, C, H, W]. Step 3: Assume :attr:`shift_ratio` is :math:`1/4`, slice padding result as follows: $$ slice1 = x[:, :T, :C/4, :, :] $$ $$ slice2 = x[:, 2:T+2, C/4:C/2, :, :] $$ $$ slice3 = x[:, 1:T+1, C/2:, :, :] $$ Step 4: Concatenate three slices along the 3rd(C) dimension and reshape result to [N*T, C, H, W]. For details of temporal shifting, please refer to paper: `Temporal Shift Module `_ . Args: x(Tensor): ${x_comment} seg_num(int): ${seg_num_comment} shift_ratio(float): ${shift_ratio_comment} name(str, optional): For detailed information, please refer to :ref:`api_guide_Name`. Usually name is no need to set and None by default. data_format(str, optional): Data format that specifies the layout of input. It can be "NCHW" or "NHWC". Default: "NCHW". Returns: out(Tensor): The temporal shifting result is a tensor with the same shape and same data type as the input. Examples: .. code-block:: python >>> import paddle >>> import paddle.nn.functional as F >>> input = paddle.randn([6, 4, 2, 2]) >>> out = F.temporal_shift(x=input, seg_num=2, shift_ratio=0.2) )NCHWNHWCzJAttr(data_format) should be 'NCHW' or 'NHWC'. Received Attr(data_format): .temporal_shiftr4)float16uint16float32float64seg_num shift_ratior!zseg_num must be int type.r#r@)rLrM data_formatr&)rG)rBrrrGr r1r r intfloatr2r"r+ TypeErrorr3)r4rLrMr5rNr7r6s rrGrGs|** ++6-q :  $$QkJJ::  7   7Is,<=; u6FG77agg7F'3'78 8!8CL"**   r)r)Nr<N)g?NrD)paddlerrrr paddle.utilsrbase.data_feederr r base.layer_helperr common_ops_importr frameworkr rr__all__rr r9rGrrr\sd$BA#D,)  ! I  8  8Yx^B_r