# Copyright 2022-2023 XProbe Inc.
# derived from copyright 1999-2021 Alibaba Group Holding Ltd.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import numpy as np
from ... import opcodes as OperandDef
from ..array_utils import as_same_device, device
from ..datasource import tensor as astensor
from .core import TensorReduction, TensorReductionMixin, numel
class TensorMean(TensorReduction, TensorReductionMixin):
_op_type_ = OperandDef.MEAN
def __init__(self, axis=None, keepdims=None, combine_size=None, stage=None, **kw):
stage = self._rewrite_stage(stage)
super().__init__(
_axis=axis,
_keepdims=keepdims,
_combine_size=combine_size,
stage=stage,
**kw
)
@classmethod
def execute_agg(cls, ctx, op):
axis = cls.get_axis(op.axis)
a = ctx[op.inputs[0].key]
if not isinstance(a, (list, tuple)):
(inp,), device_id, xp = as_same_device(
[a], device=op.device, ret_extra=True
)
with device(device_id):
ctx[op.outputs[0].key] = xp.mean(
inp, axis=axis, dtype=op.dtype, keepdims=bool(op.keepdims)
)
else:
(_data, _count), device_id, xp = as_same_device(
a, device=op.device, ret_extra=True
)
with device(device_id):
chunk_count = xp.sum(
_count, axis=axis, dtype=op.dtype, keepdims=bool(op.keepdims)
)
chunk_sum = xp.sum(
_data, axis=axis, dtype=op.dtype, keepdims=bool(op.keepdims)
)
ctx[op.outputs[0].key] = xp.true_divide(
chunk_sum, chunk_count, dtype=op.dtype
)
@classmethod
def execute_map(cls, ctx, op):
(in_chunk,), device_id, xp = as_same_device(
[ctx[c.key] for c in op.inputs], device=op.device, ret_extra=True
)
axis = cls.get_axis(op.axis)
with device(device_id):
chunk_count = numel(
in_chunk, axis=axis, dtype=np.int64, keepdims=bool(op.keepdims)
)
chunk_sum = xp.sum(
in_chunk, axis=axis, dtype=op.dtype, keepdims=bool(op.keepdims)
)
ctx[op.outputs[0].key] = (chunk_sum, chunk_count)
@classmethod
def execute_combine(cls, ctx, op):
axis = cls.get_axis(op.axis)
(_data, _count), device_id, xp = as_same_device(
ctx[op.inputs[0].key], device=op.device, ret_extra=True
)
with device(device_id):
chunk_count = xp.sum(
_count, axis=axis, dtype=np.int64, keepdims=bool(op.keepdims)
)
chunk_sum = xp.sum(
_data, axis=axis, dtype=op.dtype, keepdims=bool(op.keepdims)
)
ctx[op.outputs[0].key] = (chunk_sum, chunk_count)
[docs]def mean(a, axis=None, dtype=None, out=None, keepdims=None, combine_size=None):
"""
Compute the arithmetic mean along the specified axis.
Returns the average of the array elements. The average is taken over
the flattened tensor by default, otherwise over the specified axis.
`float64` intermediate and return values are used for integer inputs.
Parameters
----------
a : array_like
Tensor containing numbers whose mean is desired. If `a` is not an
tensor, a conversion is attempted.
axis : None or int or tuple of ints, optional
Axis or axes along which the means are computed. The default is to
compute the mean of the flattened array.
If this is a tuple of ints, a mean is performed over multiple axes,
instead of a single axis or all the axes as before.
dtype : data-type, optional
Type to use in computing the mean. For integer inputs, the default
is `float64`; for floating point inputs, it is the same as the
input dtype.
out : Tensor, optional
Alternate output tensor in which to place the result. The default
is ``None``; if provided, it must have the same shape as the
expected output, but the type will be cast if necessary.
See `doc.ufuncs` for details.
keepdims : bool, optional
If this is set to True, the axes which are reduced are left
in the result as dimensions with size one. With this option,
the result will broadcast correctly against the input tensor.
If the default value is passed, then `keepdims` will not be
passed through to the `mean` method of sub-classes of
`Tensor`, however any non-default value will be. If the
sub-classes `sum` method does not implement `keepdims` any
exceptions will be raised.
combine_size: int, optional
The number of chunks to combine.
Returns
-------
m : Tensor, see dtype parameter above
If `out=None`, returns a new tensor containing the mean values,
otherwise a reference to the output array is returned.
See Also
--------
average : Weighted average
std, var, nanmean, nanstd, nanvar
Notes
-----
The arithmetic mean is the sum of the elements along the axis divided
by the number of elements.
Note that for floating-point input, the mean is computed using the
same precision the input has. Depending on the input data, this can
cause the results to be inaccurate, especially for `float32` (see
example below). Specifying a higher-precision accumulator using the
`dtype` keyword can alleviate this issue.
By default, `float16` results are computed using `float32` intermediates
for extra precision.
Examples
--------
>>> import mars.tensor as mt
>>> a = mt.array([[1, 2], [3, 4]])
>>> mt.mean(a).execute()
2.5
>>> mt.mean(a, axis=0).execute()
array([ 2., 3.])
>>> mt.mean(a, axis=1).execute()
array([ 1.5, 3.5])
In single precision, `mean` can be inaccurate:
>>> a = mt.zeros((2, 512*512), dtype=mt.float32)
>>> a[0, :] = 1.0
>>> a[1, :] = 0.1
>>> mt.mean(a).execute()
0.54999924
Computing the mean in float64 is more accurate:
>>> mt.mean(a, dtype=mt.float64).execute()
0.55000000074505806
"""
a = astensor(a)
if dtype is None:
dtype = np.mean(np.empty((1,), dtype=a.dtype)).dtype
op = TensorMean(
axis=axis, dtype=dtype, keepdims=keepdims, combine_size=combine_size
)
return op(a, out=out)