如何在Julia中读取包含类型为'<class'numpy.float32'>'的稀疏矩阵的泡菜文件？

如何解决如何在Julia中读取包含类型为'<class'numpy.float32'>'的稀疏矩阵的泡菜文件？

我正在尝试读取Julia中最初由python创建的泡菜文件。这是我所做的：

f3=open("filename.pickle");

r3 = pickle.load(f3)

这将返回以下内容：

PyObject <41302x1425 sparse matrix of type '<class 'numpy.float32'>'
    with 1602890 stored elements in Compressed Sparse Row format>

1. 如何访问矩阵元素？

2. 假设我在Julia中有一个稀疏矩阵，如何将数据存储到相同格式的pickle文件中？

仅供参考，我已经做了以下操作来解决有关找不到scipy模块的错误：

using Conda

Conda.add("scipy")

解决方法

从Julia到Python泡菜：

julia> using PyCall

julia> a = rand(Float32,2,2)
2×2 Array{Float32,2}:
 0.943764  0.726961
 0.9184    0.422781

julia> pickle = pyimport("pickle");

julia> open("pyt.pickle","w") do f
         pickle.dump(a,f)
       end

在Python中阅读上述泡菜：

>>> import pickle,numpy
>>> f=open("pyt.pickle","rb")
>>> a = pickle.load(f)
>>> f.close()
>>> a
array([[0.94376445,0.72696066],[0.91840017,0.42278147]],dtype=float32)
>>> type(a)
<class 'numpy.ndarray'>

准备一个新的泡菜，这次将在Julia中阅读：

>>> b = numpy.ones((2,3),dtype='float32')
>>> b
array([[1.,1.,1.],[1.,1.]],dtype=float32)
>>> f=open("pyt2.pickle","wb")
>>> pickle.dump(b,f)
>>> f.close()

在Julia中读取Python创建的泡菜：

julia> using PyCall

julia> pickle = pyimport("pickle");

julia> open("pyt2.pickle","r") do f
         pickle.load(f)
       end
2×3 Array{Float32,2}:
 1.0  1.0  1.0
 1.0  1.0  1.0

此介绍之后，让我们做一个稀疏数组。我们从Python设置开始：

>>> import scipy
>>> a = scipy.sparse.rand(4,4,0.25,dtype="float32")
>>> a
<4x4 sparse matrix of type '<class 'numpy.float32'>'
        with 4 stored elements in COOrdinate format>
>>> f=open("pyt3.pickle","wb")
>>> pickle.dump(a,f)
>>> f.close()
>>> print(a)
  (0,3)        0.30552787
  (3,0)        0.810103
  (2,1)        0.691249
  (2,2)        0.63436085

让我们现在在朱莉娅中阅读它：

julia> a=open("pyt3.pickle","r") do f
                pickle.load(f)
                       end
PyObject <4x4 sparse matrix of type '<class 'numpy.float64'>'
        with 4 stored elements in COOrdinate format>
julia> using SparseArrays;      
julia> res = spzeros(Float32,a.shape...);
julia> sp = pyimport("scipy.sparse");
julia> i,j,vals = sp.find(a);

julia> setindex!.(Ref(res),vals,i .+ 1,j .+ 1); #we copy the data to Julia structure


julia> res
4×4 SparseMatrixCSC{Float32,Int64} with 4 stored entries:
  [4,1]  =  0.810103
  [3,2]  =  0.691249
  [3,3]  =  0.634361
  [1,4]  =  0.305528