Tell me, is there any function that lets you know how much memory is occupied by a matrix or an array of numpy?

    1 answer 1

    You can use the nbytes method:

    In [5]: a = np.random.rand(10**6) In [6]: a.nbytes Out[6]: 8000000 In [7]: a.dtype Out[7]: dtype('float64')

    this also works for multidimensional matrices:

     In [11]: a = np.random.randint(0, 10, (10**3, 10**3, 10**3), dtype=np.int8) In [12]: a.shape Out[12]: (1000, 1000, 1000) In [13]: a.nbytes Out[13]: 1000000000 In [14]: a.dtype Out[14]: dtype('int8')

    You can also use the standard function: sys.getsizeof ()

     In [33]: a = np.random.rand(10**6) In [34]: sys.getsizeof(a) Out[34]: 8000096

    which will call __sizeof__ for the callee:

     In [35]: a.__sizeof__() Out[35]: 8000096

    From the documentation:

    getsizeof () calls the sizeof __ sizeof __ method adds the additional garbage collector overhead.

    • one
      In the simple case, numpy.array can be considered as a piece of memory with the corresponding metadata (type, size, where / what lies). .nbytes returns the size of a piece of memory (the space occupied by the elements of the array — in the idealized model: perhaps ignoring alignment, auxiliary structures), disregarding the metadata. sys.getsizeof() promises to return the size of the Python object ( numpy.array() ) in bytes, which may include space for metadata (sometimes garbage is returned). These numbers should be considered as estimates of the required memory, and the actual memory can be recognized by measurements. - jfs