mxnet.numpy_extension.utils

Util functions for the numpy module.

Functions

load(file)	Load arrays from .npy, .npz or legacy MXNet file format.
save(file, arr)	Save an array to a binary file in NumPy .npy format.
savez(file, *args, **kwds)	Save several arrays into a single file in uncompressed .npz format.

mxnet.numpy_extension.utils.from_dlpack(dlpack)

Returns a np.ndarray backed by a dlpack tensor.

Parameters:

dlpack (PyCapsule (the pointer of DLManagedTensor)) – input data

Returns:

an ndarray backed by a dlpack tensor

Return type:

np.ndarray

Examples

>>> x = mx.np.ones((2,3))
>>> y = mx.npx.to_dlpack_for_read(x)
>>> type(y)
<class 'PyCapsule'>
>>> z = mx.npx.from_dlpack(y)
>>> type(z)
<class 'mxnet.numpy.ndarray'>
>>> z
array([[1., 1., 1.],
       [1., 1., 1.]])

>>> w = mx.npx.to_dlpack_for_write(x)
>>> type(w)
<class 'PyCapsule'>
>>> u = mx.npx.from_dlpack(w)
>>> u += 1
>>> x
array([[2., 2., 2.],
       [2., 2., 2.]])

mxnet.numpy_extension.utils.from_numpy(ndarray, zero_copy=True)

Returns an MXNet’s np.ndarray backed by numpy’s ndarray. When zero_copy is set to be true, this API consumes numpy’s ndarray and produces MXNet’s np.ndarray without having to copy the content. In this case, we disallow users to modify the given numpy ndarray, and it is suggested not to read the numpy ndarray as well for internal correctness.

Parameters:

• ndarray (np.ndarray) – input data

• zero_copy (bool) – Whether we use DLPack’s zero-copy conversion to convert to MXNet’s np.ndarray. This is only available for c-contiguous arrays, i.e. array.flags[C_CONTIGUOUS] == True.

Returns:

a np.ndarray backed by a dlpack tensor

Return type:

np.ndarray

mxnet.numpy_extension.utils.load(file)

Load arrays from .npy, .npz or legacy MXNet file format.

See more details in save.

Parameters:

file (str) – The filename.

Returns:

result – Data stored in the file.

Return type:

list of ndarrays or dict of str -> ndarray

Notes

This function can only be called within numpy semantics, i.e., npx.is_np_shape() and npx.is_np_array() must both return true.

mxnet.numpy_extension.utils.save(file, arr)

Save an array to a binary file in NumPy .npy format.

Parameters:

• file (str) – File or filename to which the data is saved. If file is a file-object, then the filename is unchanged.

• arr (ndarray) – Array data to be saved. Sparse formats are not supported. Please use savez function to save sparse arrays.

See also

savez

Save several arrays into a .npz archive

Notes

For a description of the .npy format, see numpy.lib.format.

mxnet.numpy_extension.utils.savez(file, *args, **kwds)

Save several arrays into a single file in uncompressed .npz format.

If arguments are passed in with no keywords, the corresponding variable names, in the .npz file, are ‘arr_0’, ‘arr_1’, etc. If keyword arguments are given, the corresponding variable names, in the .npz file will match the keyword names.

Parameters:

• file (str) – Either the filename (string) or an open file (file-like object) where the data will be saved.

• args (Arguments, optional) – Arrays to save to the file. Since it is not possible for Python to know the names of the arrays outside savez, the arrays will be saved with names “arr_0”, “arr_1”, and so on. These arguments can be any expression.

• kwds (Keyword arguments, optional) – Arrays to save to the file. Arrays will be saved in the file with the keyword names.

Return type:

None

See also

save

Save a single array to a binary file in NumPy format.

Notes

The .npz file format is a zipped archive of files named after the variables they contain. The archive is not compressed and each file in the archive contains one variable in .npy format. For a description of the .npy format, see numpy.lib.format.

When opening the saved .npz file with load a dictionary object mapping file-names to the arrays themselves.

When saving dictionaries, the dictionary keys become filenames inside the ZIP archive. Therefore, keys should be valid filenames. E.g., avoid keys that begin with / or contain ..

mxnet.numpy_extension.utils.to_dlpack_for_read(data)

Returns a reference view of np.ndarray that represents as DLManagedTensor until all previous write operations on the current array are finished.

data: np.ndarray

input data.

PyCapsule (the pointer of DLManagedTensor)

a reference view of ndarray that represents as DLManagedTensor.

>>> x = mx.np.ones((2,3))
>>> y = mx.npx.to_dlpack_for_read(x)
>>> type(y)
<class 'PyCapsule'>
>>> z = mx.npx.from_dlpack(y)
>>> z
array([[1., 1., 1.],
       [1., 1., 1.]])

mxnet.numpy_extension.utils.to_dlpack_for_write(data)

Returns a reference view of ndarray that represents as DLManagedTensor until all previous read/write operations on the current array are finished.

data: np.ndarray

input data.

PyCapsule (the pointer of DLManagedTensor)

a reference view of np.ndarray that represents as DLManagedTensor.

>>> x = mx.np.ones((2,3))
>>> w = mx.npx.to_dlpack_for_write(x)
>>> type(w)
<class 'PyCapsule'>
>>> u = mx.npx.from_dlpack(w)
>>> u += 1
>>> x
array([[2., 2., 2.],
       [2., 2., 2.]])