• torch.repeat()


    
    >>> import torch
    >>> 
    >>> a = torch.randn(33, 55)
    >>> a.size()
    torch.Size([33, 55])
    #repeat(repeat_counts_axis_0,repeat_counts_axis_1)
    >>> 
    #(arg1,arg2,arg3,...,axis = 0, axis=1),除了最后两个参数是指在相应维度上复制到的结果维度,其余都是在tensor.size前追加维度
    >>> a.repeat(1, 1).size()
    torch.Size([33, 55])
    >>> 
    >>> a.repeat(2,1).size()
    torch.Size([66, 55])
    >>> 
    >>> a.repeat(1,2).size()
    torch.Size([33, 110])
    >>>
    >>> a.repeat(1,1,1).size()
    torch.Size([1, 33, 55])
    >>>
    >>> a.repeat(2,1,1).size()
    torch.Size([2, 33, 55])
    >>>
    >>> a.repeat(1,2,1).size()
    torch.Size([1, 66, 55])
    >>>
    >>> a.repeat(1,1,2).size()
    torch.Size([1, 33, 110])
    >>>
    >>> a.repeat(1,1,1,1).size()
    torch.Size([1, 1, 33, 55])
    >>> 
    >>> # repeat()的参数的个数,不能少于被操作的张量的维度的个数
    #即repeat的参数最少是tensor的维度个数
    >>> # 下面是一些错误示例
    >>> a.repeat(2).size()  # 1D < 2D, error
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
    RuntimeError: Number of dimensions of repeat dims can not be smaller than number of dimensions of tensor
    >>>
    >>> b = torch.randn(5,6,7)
    >>> b.size() # 3D
    torch.Size([5, 6, 7])
    >>> 
    >>> b.repeat(2).size() # 1D < 3D, error
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
    RuntimeError: Number of dimensions of repeat dims can not be smaller than number of dimensions of tensor
    >>>
    >>> b.repeat(2,1).size() # 2D < 3D, error
    Traceback (most recent call last):
      File "<stdin>", line 1, in <module>
    RuntimeError: Number of dimensions of repeat dims can not be smaller than number of dimensions of tensor
    >>>
    >>> b.repeat(2,1,1).size() # 3D = 3D, okay
    torch.Size([10, 6, 7])
    >>>
    
    
  • 相关阅读:
    setlocale
    c++的 程序实体 作用域 名空间 变量生存期概念理解
    本人的Ubuntu 10.04配置清单
    hadoopStreamming 编程 Angels
    级联 mapreduce (Cascading Mapreduce) Angels
    委托
    OPC和XML/SOAP/Web Services
    实例管理2(会话服务)
    实例管理3()
    操作(Operation)
  • 原文地址:https://www.cnblogs.com/Henry-ZHAO/p/13857361.html
Copyright © 2020-2023  润新知