mindspore.Tensor.mean

Tensor.mean(dim=None, keepdim=False, *, dtype=None) → Tensor[source]

Reduces all dimensions of a tensor by averaging all elements in the dimension, by default. And reduce a dimension of self along the specified dim. keepdim determines whether the dimensions of the output and self are the same.

Warning

Non-backward-compatible change after version 2.9.0: overload mean(axis=None, keep_dims=False) will be removed, and only mean(dim=None, keepdim=False, *, dtype=None) will be supported.

Note

The dim with tensor type is only used for compatibility with older versions and is not recommended.

Parameters

  • dim (Union[int, tuple(int), list(int), Tensor], optional) – The dimensions to reduce. Default: None , reduce all dimensions. Only constant value is allowed. Assume the rank of self is r, and the value range is [-r,r).

  • keepdim (bool, optional) – If True , keep these reduced dimensions and the length is 1. If False , don't keep these dimensions. Default: False .

Keyword Arguments

dtype (mindspore.dtype, optional) – The desired data type of returned Tensor. Default: None .

Returns

Tensor, has the same data type as self tensor.

  • If dim is None , and keepdim is False , the output is a 0-D tensor representing the mean of all elements in the self tensor.

  • If dim is int, set as 1, and keepdim is False , the shape of output is \((x_0, x_2, ..., x_R)\).

  • If dim is tuple(int), set as (1, 2), and keepdim is False , the shape of output is \((x_0, x_3, ..., x_R)\).

  • If dim is 1-D Tensor, set as [1, 2], and keepdim is False , the shape of output is \((x_0, x_3, ..., x_R)\).

Raises

  • TypeError – If dim is not one of the following: int, tuple, list or Tensor.

  • TypeError – If keepdim is not a bool.

  • ValueError – If dim is out of range.

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore
>>> import numpy as np
>>> from mindspore import Tensor
>>> x = Tensor(np.random.randn(3, 4, 5, 6).astype(np.float32))
>>> output = Tensor.mean(x, 1, keepdim=True)
>>> result = output.shape
>>> print(result)
(3, 1, 5, 6)
>>> # case 1: Reduces a dimension by averaging all elements in the dimension.
>>> x = Tensor(np.array([[[2, 2, 2, 2, 2, 2], [2, 2, 2, 2, 2, 2], [2, 2, 2, 2, 2, 2]],
... [[4, 4, 4, 4, 4, 4], [5, 5, 5, 5, 5, 5], [6, 6, 6, 6, 6, 6]],
... [[6, 6, 6, 6, 6, 6], [8, 8, 8, 8, 8, 8], [10, 10, 10, 10, 10, 10]]]),
... mindspore.float32)
>>> output = Tensor.mean(x)
>>> print(output)
5.0
>>> print(output.shape)
()
>>> # case 2: Reduces a dimension along the dim 0
>>> output = Tensor.mean(x, 0, True)
>>> print(output)
[[[4. 4. 4. 4. 4. 4.]
  [5. 5. 5. 5. 5. 5.]
  [6. 6. 6. 6. 6. 6.]]]
>>> # case 3: Reduces a dimension along the dim 1
>>> output = Tensor.mean(x, 1, True)
>>> print(output)
[[[2. 2. 2. 2. 2. 2.]]
 [[5. 5. 5. 5. 5. 5.]]
 [[8. 8. 8. 8. 8. 8.]]]
>>> # case 4: Reduces a dimension along the dim 2
>>> output = Tensor.mean(x, 2, True)
>>> print(output)
[[[ 2.]
  [ 2.]
  [ 2.]]
 [[ 4.]
  [ 5.]
  [ 6.]]
 [[ 6.]
  [ 8.]
  [10.]]]
Tensor.mean(axis=None, keep_dims=False) → Tensor[source]

For details, please refer to mindspore.ops.mean() .