oneflow.Tensor¶
OneFlow Tensor Class¶
-
class
oneflow.
Tensor
¶ -
abs
()¶ See
oneflow.abs()
-
acos
()¶ See
oneflow.acos()
-
acosh
()¶ See
oneflow.acosh()
-
add
(other)¶ Computes the addition of input by other for each element, scalar and broadcast promotation are supported. The formula is:
\[out = input + other\]For example:
>>> import numpy as np >>> import oneflow as flow # element-wise add >>> x = flow.Tensor(np.random.randn(2,3)) >>> y = flow.Tensor(np.random.randn(2,3)) >>> out = flow.add(x, y).numpy() >>> out.shape (2, 3) # scalar add >>> x = 5 >>> y = flow.Tensor(np.random.randn(2,3)) >>> out = flow.add(x, y).numpy() >>> out.shape (2, 3) # broadcast add >>> x = flow.Tensor(np.random.randn(1,1)) >>> y = flow.Tensor(np.random.randn(2,3)) >>> out = flow.add(x, y).numpy() >>> out.shape (2, 3)
-
add_
(y)¶ In-place version of
oneflow.Tensor.add()
.
-
addmm
(mat1, mat2, alpha=1, beta=1)¶ See
oneflow.addmm()
-
arccos
()¶ See
oneflow.arccos()
-
arccosh
()¶
-
arcsin
()¶ See
oneflow.asin()
-
arcsinh
()¶ See
oneflow.asinh()
-
arctan
()¶ See
oneflow.arctan()
-
arctanh
()¶
-
argmax
(dim=None, keepdim=None)¶ See
oneflow.argmax()
-
argmin
(dim=None, keepdim=None)¶ See
oneflow.argmin()
-
argsort
(dim=None, descending=None)¶ This operator sorts the input Tensor at specified dim and return the indices of the sorted Tensor.
- Parameters
input (oneflow.Tensor) – The input Tensor.
dim (int, optional) – dimension to be sorted. Defaults to the last dim (-1).
descending (bool, optional) – controls the sorting order (ascending or descending).
- Returns
The indices of the sorted Tensor.
- Return type
For example:
>>> import numpy as np >>> import oneflow as flow >>> x = np.array([[10, 2, 9, 3, 7], ... [1, 9, 4, 3, 2]]).astype("float32") >>> input = flow.Tensor(x) >>> output = flow.argsort(input) >>> output tensor([[1, 3, 4, 2, 0], [0, 4, 3, 2, 1]], dtype=oneflow.int32) >>> output = flow.argsort(input, descending=True) >>> output tensor([[0, 2, 4, 3, 1], [1, 2, 3, 4, 0]], dtype=oneflow.int32) >>> output = flow.argsort(input, dim=0) >>> output tensor([[1, 0, 1, 0, 1], [0, 1, 0, 1, 0]], dtype=oneflow.int32)
-
asin
()¶ See
oneflow.asin()
-
asinh
()¶ See
oneflow.asinh()
-
atan
()¶ See
oneflow.atan()
-
atan2
(other)¶ See
oneflow.atan2()
-
atanh
()¶ See
oneflow.atanh()
-
backward
(gradient=None, retain_graph=False, create_graph=False)¶ The interface is consistent with PyTorch. The documentation is referenced from: https://pytorch.org/docs/stable/generated/torch.Tensor.backward.html#torch.Tensor.backward.
Computes the gradient of current tensor w.r.t. graph leaves.
The graph is differentiated using the chain rule. If the tensor is non-scalar (i.e. its data has more than one element) and requires gradient, the function additionally requires specifying gradient. It should be a tensor of matching type and location, that contains the gradient of the differentiated function w.r.t. self.
This function accumulates gradients in the leaves - you might need to zero .grad attributes or set them to None before calling it. See Default gradient layouts for details on the memory layout of accumulated gradients.
Note
If you run any forward ops, create gradient, and/or call backward in a user-specified CUDA stream context, see Stream semantics of backward passes.
Note
When inputs are provided and a given input is not a leaf, the current implementation will call its grad_fn (though it is not strictly needed to get this gradients). It is an implementation detail on which the user should not rely. See https://github.com/pytorch/pytorch/pull/60521#issuecomment-867061780 for more details.
- Parameters
gradient (Tensor or None) – Gradient w.r.t. the tensor. If it is a tensor, it will be automatically converted to a Tensor that does not require grad unless create_graph is True. None values can be specified for scalar Tensors or ones that don’t require grad. If a None value would be acceptable then this argument is optional.
retain_graph (bool, optional) – If False, the graph used to compute the grads will be freed. Note that in nearly all cases setting this option to True is not needed and often can be worked around in a much more efficient way. Defaults to the value of create_graph.
create_graph (bool, optional) – If True, graph of the derivative will be constructed, allowing to compute higher order derivative products. Defaults to False.
-
bmm
(other)¶ See
oneflow.bmm()
-
cast
(dtype)¶ See
oneflow.cast()
-
ceil
()¶ See
oneflow.ceil()
-
chunk
(chunks=None, dim=None)¶ See
oneflow.chunk()
-
clamp
(min=None, max=None)¶ See
oneflow.clamp()
-
clip
(min=None, max=None)¶ See
oneflow.clamp()
-
clone
(self: oneflow._oneflow_internal.Tensor) → oneflow._oneflow_internal.Tensor¶
-
copy_
(other: Union[oneflow._oneflow_internal.Tensor, numpy.ndarray])¶ The interface is consistent with PyTorch.
Tensor.copy_(src, non_blocking=False) → Tensor
Copies the elements from src into self tensor and returns self.
The src tensor must be broadcastable with the self tensor. It may be of a different data type or reside on a different device.
- Parameters
src (Tensor) – the source tensor to copy from
non_blocking (bool) – if True and this copy is between CPU and GPU, the copy may occur asynchronously with respect to the host. For other cases, this argument has no effect.
-
cos
()¶ Returns a new tensor with the cosine of the elements of
input
.\[\text{out}_{i} = \cos(\text{input}_{i})\]- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> import numpy as np >>> arr = np.array([1.4309, 1.2706, -0.8562, 0.9796]) >>> input = flow.tensor(arr, dtype=flow.float32) >>> output = flow.cos(input).numpy()
-
cosh
()¶ Returns a new tensor with the hyperbolic cosine of the elements of
input
.\[\text{out}_{i} = \cosh(\text{input}_{i})\]- Parameters
input (Tensor) – the input tensor.
For example:
>>> import numpy as np >>> import oneflow as flow >>> arr = np.array([ 0.1632, 1.1835, -0.6979, -0.7325]) >>> input = flow.tensor(arr, dtype=flow.float32) >>> output = flow.cosh(input).numpy() >>> output array([1.0133467, 1.7859949, 1.2535787, 1.2804903], dtype=float32)
-
cpu
()¶ Returns a copy of this object in CPU memory. If this object is already in CPU memory and on the correct device, then no copy is performed and the original object is returned.
For example:
>>> import oneflow as flow >>> input = flow.tensor([1, 2, 3, 4, 5], device=flow.device("cuda")) >>> output = input.cpu() >>> output.device device(type='cpu', index=0)
-
cuda
(device: Optional[Union[oneflow.nn.modules.tensor_ops.int, str, oneflow._oneflow_internal.device]] = None)¶ Returns a copy of this object in CUDA memory. If this object is already in CUDA memory and on the correct device, then no copy is performed and the original object is returned.
- Parameters
device (flow.device) – The destination GPU device. Defaults to the current CUDA device.
For example:
>>> import oneflow as flow >>> input = flow.Tensor([1, 2, 3, 4, 5]) >>> output = input.cuda() >>> output.device device(type='cuda', index=0)
-
property
data
¶
-
detach
(self: oneflow._oneflow_internal.Tensor) → oneflow._oneflow_internal.Tensor¶
-
property
device
¶
-
diag
(diagonal=0)¶ See
oneflow.diag()
-
diagonal
(offset=0, dim1=0, dim2=1)¶
-
dim
()¶ Tensor.dim() → int
Returns the number of dimensions of self tensor.
-
div
(other)¶ Computes the division of input by other for each element, scalar and broadcast promotation are supported. The formula is:
\[out = \frac{input}{other}\]- Parameters
input (Union[int, float, flow.Tensor]) – input.
other (Union[int, float, flow.Tensor]) – other.
For example:
>>> import numpy as np >>> import oneflow as flow # element-wise divide >>> input = flow.Tensor(np.random.randn(2,3)) >>> other = flow.Tensor(np.random.randn(2,3)) >>> out = flow.div(input,other).numpy() >>> out.shape (2, 3) # scalar divide >>> input = 5 >>> other = flow.Tensor(np.random.randn(2,3)) >>> out = flow.div(input,other).numpy() >>> out.shape (2, 3) # broadcast divide >>> input = flow.Tensor(np.random.randn(1,1)) >>> other = flow.Tensor(np.random.randn(2,3)) >>> out = flow.div(input,other).numpy() >>> out.shape (2, 3)
-
double
()¶ Tensor.double() is equivalent to Tensor.to(flow.float64). See to().
- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> import numpy as np >>> input = flow.tensor(np.random.randn(1, 2, 3), dtype=flow.int) >>> input = input.double() >>> input.dtype oneflow.float64
-
property
dtype
¶
-
element_size
()¶ Tensor.element_size() → int
Returns the size in bytes of an individual element.
-
eq
(other)¶ Computes element-wise equality. The second argument can be a number or a tensor whose shape is broadcastable with the first argument.
- Parameters
input (oneflow.Tensor) – the tensor to compare
other (oneflow.Tensor, float or int) – the target to compare
- Returns
A boolean tensor that is True where
input
is equal toother
and False elsewhere
For example:
>>> import oneflow as flow >>> import numpy as np >>> input = flow.tensor(np.array([2, 3, 4, 5]), dtype=flow.float32) >>> other = flow.tensor(np.array([2, 3, 4, 1]), dtype=flow.float32) >>> y = flow.eq(input, other) >>> y tensor([1, 1, 1, 0], dtype=oneflow.int8)
-
erf
()¶ See
oneflow.erf()
-
erfc
()¶ See
oneflow.erfc()
-
exp
()¶ See
oneflow.exp()
-
expand
(*sizes)¶ This operator expand the input tensor to a larger size.
Passing -1 as the size for a dimension means not changing the size of that dimension.
Tensor can be also expanded to a larger number of dimensions and the new ones will be appended at the front.
For the new dimensions, the size cannot be set to -1.
- Parameters
input (oneflow.Tensor) – The input Tensor.
*sizes (oneflow.Size or int) – The desired expanded size.
- Returns
The result Tensor.
- Return type
For example:
>>> import oneflow as flow >>> import numpy as np >>> x = np.array([[[[0, 1]], ... [[2, 3]], ... [[4, 5]]]]).astype(np.int32) >>> input = flow.Tensor(x) >>> input.shape oneflow.Size([1, 3, 1, 2]) >>> out = input.expand(1, 3, 2, 2) >>> out.shape oneflow.Size([1, 3, 2, 2])
-
expand_as
(other) → Tensor¶ Expand this tensor to the same size as
other
.self.expand_as(other)
is equivalent toself.expand(other.size())
.Please see
expand()
for more information aboutexpand
.- Parameters
other (
oneflow.Tensor
) – The result tensor has the same size asother
.
-
expm1
()¶ See
oneflow.expm1()
-
fill_
(value)¶ Tensor.fill_(value) → Tensor
Fills self tensor with the specified value.
-
flip
(dims)¶ See
oneflow.flip()
-
float
()¶ Tensor.float() is equivalent to Tensor.to(flow.float32). See to().
- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> import numpy as np >>> input = flow.tensor(np.random.randn(1, 2, 3), dtype=flow.int) >>> input = input.float() >>> input.dtype oneflow.float32
-
floor
()¶ See
oneflow.floor()
-
fmod
(other)¶ See
oneflow.fmod()
-
gather
(dim, index) → Tensor¶ See
oneflow.gather()
-
ge
(other)¶ See
oneflow.ge()
-
gelu
()¶ See
oneflow.gelu()
-
get_device
() -> Device ordinal (Integer)¶ For CUDA tensors, this function returns the device ordinal of the GPU on which the tensor resides. For CPU tensors, an error is thrown.
-
property
grad
¶
-
property
grad_fn
¶
-
gt
(other)¶ See
oneflow.gt()
-
int
()¶ Tensor.int() is equivalent to Tensor.to(flow.int32). See to().
- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> import numpy as np >>> input = flow.tensor(np.random.randn(1, 2, 3), dtype=flow.float32) >>> input = input.int() >>> input.dtype oneflow.int32
-
property
is_consistent
¶
-
is_contiguous
(self: oneflow._oneflow_internal.Tensor) → bool¶
-
property
is_cuda
¶
-
is_floating_point
()¶ Returns True if the data type of input is a floating point data type i.e., one of flow.float64, flow.float32, flow.float16.
- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> input = flow.tensor([1, 2, 3, 4, 5], dtype=flow.int) >>> output = flow.is_floating_point(input) >>> output False
-
property
is_lazy
¶
-
property
is_leaf
¶
-
item
()¶ Returns the value of this tensor as a standard Python number. This only works for tensors with one element. For other cases, see tolist().
This operation is not differentiable.
- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> x = flow.tensor([1.0]) >>> x.item() 1.0
-
le
(other)¶ Returns the truth value of \(input <= other\) element-wise.
- Parameters
input (oneflow.Tensor) – A Tensor
other (oneflow.Tensor) – A Tensor
- Returns
A Tensor with int8 type.
- Return type
For example:
>>> import numpy as np >>> import oneflow as flow >>> input1 = flow.tensor(np.array([1, 2, 3]).astype(np.float32), dtype=flow.float32) >>> input2 = flow.tensor(np.array([1, 1, 4]).astype(np.float32), dtype=flow.float32) >>> out = flow.le(input1, input2) >>> out tensor([1, 0, 1], dtype=oneflow.int8)
-
log
()¶ Returns a new tensor with the natural logarithm of the elements of
input
.\[y_{i} = \log_{e} (x_{i})\]- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> import numpy as np >>> arr = np.random.randn(2, 3, 4, 5) >>> input = flow.tensor(arr, dtype=flow.float32) >>> output = flow.log(input)
-
log1p
()¶ See
oneflow.log1p()
-
long
()¶ Tensor.long() is equivalent to Tensor.to(flow.int64). See to().
- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> import numpy as np >>> input = flow.tensor(np.random.randn(1, 2, 3), dtype=flow.float32) >>> input = input.long() >>> input.dtype oneflow.int64
-
lt
(other)¶ Returns the truth value of \(input < other\) element-wise.
- Parameters
input (oneflow.Tensor) – A Tensor
other (oneflow.Tensor) – A Tensor
- Returns
A Tensor with int8 type.
- Return type
For example:
>>> import numpy as np >>> import oneflow as flow >>> input1 = flow.tensor(np.array([1, 2, 3]).astype(np.float32), dtype=flow.float32) >>> input2 = flow.tensor(np.array([1, 2, 4]).astype(np.float32), dtype=flow.float32) >>> out = flow.lt(input1, input2) >>> out tensor([0, 0, 1], dtype=oneflow.int8)
-
masked_fill
(mask, value)¶ Fills elements of
self
tensor withvalue
wheremask
is True. The shape ofmask
must be broadcastable with the shape of the underlying tensor.- Parameters
mask (BoolTensor) – the boolean mask
value (float) – the value to fill in with
For example:
>>> import oneflow as flow >>> import numpy as np >>> in_arr = np.array( ... [[[-0.13169311, 0.97277078, 1.23305363, 1.56752789], ... [-1.51954275, 1.87629473, -0.53301206, 0.53006478], ... [-1.38244183, -2.63448052, 1.30845795, -0.67144869]], ... [[ 0.41502161, 0.14452418, 0.38968 , -1.76905653], ... [ 0.34675095, -0.7050969 , -0.7647731 , -0.73233418], ... [-1.90089858, 0.01262963, 0.74693893, 0.57132389]]] ... ) >>> fill_value = 8.7654321 # random value e.g. -1e9 3.1415 >>> input = flow.tensor(in_arr, dtype=flow.float32) >>> mask = flow.tensor((in_arr > 0).astype(np.int8), dtype=flow.int) >>> output = flow.masked_fill(input, mask, fill_value) # tensor([[[-0.1317, 8.7654, 8.7654, 8.7654], # [-1.5195, 8.7654, -0.533 , 8.7654], # [-1.3824, -2.6345, 8.7654, -0.6714]], # [[ 8.7654, 8.7654, 8.7654, -1.7691], # [ 8.7654, -0.7051, -0.7648, -0.7323], # [-1.9009, 8.7654, 8.7654, 8.7654]]], dtype=oneflow.float32)
-
masked_select
(mask)¶
-
matmul
(other)¶ See
oneflow.matmul()
-
max
(dim, index) → Tensor¶ See
oneflow.max()
-
mean
(dim, index) → Tensor¶ See
oneflow.mean()
-
min
(dim, index) → Tensor¶ See
oneflow.min()
-
mish
()¶ See
oneflow.mish()
-
mul
(other)¶ See
oneflow.mul()
-
narrow
(dimension, start, length)¶ See
oneflow.narrow()
-
property
ndim
¶
-
ndimension
()¶ Tensor.dim() → int
Returns the number of dimensions of self tensor.
-
ne
(other)¶ Computes element-wise not equality. The second argument can be a number or a tensor whose shape is broadcastable with the first argument.
- Parameters
input (oneflow.Tensor) – the tensor to compare
other (oneflow.Tensor, float or int) – the target to compare
- Returns
A boolean tensor that is True where
input
is not equal toother
and False elsewhere
For example:
>>> import oneflow as flow >>> import numpy as np >>> input = flow.tensor(np.array([2, 3, 4, 5]), dtype=flow.float32) >>> other = flow.tensor(np.array([2, 3, 4, 1]), dtype=flow.float32) >>> y = flow.ne(input, other) >>> y tensor([0, 0, 0, 1], dtype=oneflow.int8)
-
negative
()¶
-
nelement
()¶ Tensor.nelement() → int
Alias for numel()
-
new_ones
(size=None, dtype=None, device=None, placement=None, sbp=None, requires_grad=False)¶ Returns a Tensor of size size filled with 1. By default, the returned Tensor has the same torch.dtype and torch.device as this tensor.
- Parameters
size (int...) – a list, tuple, or flow.Size of integers defining the shape of the output tensor.
dtype (flow.dtype, optional) – the desired type of returned tensor. Default: if None, same flow.dtype as this tensor.
device (flow.device, optional) – the desired device of returned tensor. Default: if None, same flow.device as this tensor.
placement (flow.placement, optional) – the desired placement of returned consistent tensor. Default: if None, the returned tensor is local one using the argument device.
sbp (flow.sbp.sbp or tuple of flow.sbp.sbp, optional) – the desired sbp descriptor of returned consistent tensor. Default: if None, the returned tensor is local one using the argument device.
requires_grad (bool, optional) – If autograd should record operations on the returned tensor. Default: False.
For example:
>>> import numpy as np >>> import oneflow as flow >>> x = flow.Tensor(np.ones((1, 2, 3))) >>> y = x.new_ones((2, 2)) >>> y tensor([[1., 1.], [1., 1.]], dtype=oneflow.float32)
-
norm
(input, ord=None, dim=None, keepdim=False, *, dtype=None, out=None) → Tensor¶ Returns the matrix norm or vector norm of a given tensor.
This function can calculate one of eight different types of matrix norms, or one of an infinite number of vector norms, depending on both the number of reduction dimensions and the value of the ord parameter.
- Parameters
input (Tensor) – The input tensor. If dim is None, input must be 1-D or 2-D, unless
ord
is None. If bothdim
andord
are None, the 2-norm of the input flattened to 1-D will be returned. Its data type must be either a floating point or complex type. For complex inputs, the norm is calculated on of the absolute values of each element. If the input is complex and neitherdtype
norout
is specified, the result’s data type will be the corresponding floating point type (e.g. float ifinput
is complexfloat).ord (int, inf, -inf, 'fro', 'nuc', optional) –
order of norm. Default: ‘None’ The following norms can be calculated:
ord
norm for matrices
norm for vectors
None
Frobenius norm
2-norm
’fro’
Frobenius norm
– not supported –
‘nuc’
– not supported yet –
– not supported –
inf
max(sum(abs(x), dim=1))
max(abs(x))
-inf
min(sum(abs(x), dim=1))
min(abs(x))
0
– not supported –
sum(x != 0)
1
max(sum(abs(x), dim=0))
as below
-1
min(sum(abs(x), dim=0))
as below
2
– not supported yet –
as below
-2
– not supported yet –
as below
other
– not supported –
sum(abs(x)^{ord})^{(1 / ord)}
where inf refers to float(‘inf’), NumPy’s inf object, or any equivalent object.
dim (int, 2-tuple of ints, 2-list of ints, optional) – If
dim
is an int, vector norm will be calculated over the specified dimension. Ifdim
is a 2-tuple of ints, matrix norm will be calculated over the specified dimensions. Ifdim
is None, matrix norm will be calculated when the input tensor has two dimensions, and vector norm will be calculated when the input tensor has one dimension. Default:None
keepdim (bool, optional) – If set to True, the reduced dimensions are retained in the result as dimensions with size one. Default:
False
out (Tensor, optional) – The output tensor.
For example:
>>> import oneflow as flow >>> from oneflow import linalg as LA >>> import numpy as np >>> a = flow.tensor(np.arange(9, dtype=np.float32) - 4) >>> a tensor([-4., -3., -2., -1., 0., 1., 2., 3., 4.], dtype=oneflow.float32) >>> b = a.reshape(3, 3) >>> b tensor([[-4., -3., -2.], [-1., 0., 1.], [ 2., 3., 4.]], dtype=oneflow.float32) >>> LA.norm(a) tensor(7.7460, dtype=oneflow.float32) >>> LA.norm(b) tensor(7.7460, dtype=oneflow.float32) >>> LA.norm(b, 'fro') tensor(7.7460, dtype=oneflow.float32) >>> LA.norm(a, float('inf')) tensor(4., dtype=oneflow.float32) >>> LA.norm(b, float('inf')) tensor(9., dtype=oneflow.float32) >>> LA.norm(a, -float('inf')) tensor(0., dtype=oneflow.float32) >>> LA.norm(b, -float('inf')) tensor(2., dtype=oneflow.float32) >>> LA.norm(a, 1) tensor(20., dtype=oneflow.float32) >>> LA.norm(b, 1) tensor(7., dtype=oneflow.float32) >>> LA.norm(a, -1) tensor(0., dtype=oneflow.float32) >>> LA.norm(b, -1) tensor(6., dtype=oneflow.float32) >>> LA.norm(a, 2) tensor(7.7460, dtype=oneflow.float32) >>> LA.norm(a, -2) tensor(0., dtype=oneflow.float32) >>> LA.norm(a, 3) tensor(5.8480, dtype=oneflow.float32) >>> LA.norm(a, -3) tensor(0., dtype=oneflow.float32) >>> c = flow.tensor([[1., 2., 3.], ... [-1, 1, 4]]) >>> LA.norm(c, dim=0) tensor([1.4142, 2.2361, 5.0000], dtype=oneflow.float32) >>> LA.norm(c, dim=1, keepdim = True) tensor([[3.7417], [4.2426]], dtype=oneflow.float32) >>> LA.norm(c, ord=1, dim=1) tensor([6., 6.], dtype=oneflow.float32)
-
normal_
(mean=0, std=1, *, generator=None) → Tensor¶ Fills
self
tensor with elements samples from the normal distribution parameterized bymean
andstd
.
-
numel
()¶ See
oneflow.numel()
-
numpy
()¶ Tensor.numpy() → numpy.ndarray
Returns self tensor as a NumPy ndarray. This tensor and the returned ndarray share the same underlying storage. Changes to self tensor will be reflected in the ndarray and vice versa.
-
permute
(*dims)¶
-
property
placement
¶
-
pow
(b)¶ See
oneflow.pow()
-
prod
(dim, index) → Tensor¶ See
oneflow.prod()
-
reciprocal
()¶ Computes the safe reciprocal of x. If x is zero, the reciprocal will be also set to zero.
For example:
>>> import numpy as np >>> import oneflow as flow >>> x = flow.Tensor(np.array([[1, 2, 3], [4, 5, 6]])) >>> out = flow.reciprocal(x) >>> out.numpy() array([[1. , 0.5 , 0.33333334], [0.25 , 0.2 , 0.16666667]], dtype=float32)
-
register_hook
(self: oneflow._oneflow_internal.Tensor, arg0: Callable[[oneflow._oneflow_internal.Tensor], oneflow._oneflow_internal.Tensor]) → None¶
-
relu
(inplace=False)¶ See
oneflow.relu()
-
repeat
(*sizes)¶ This operator repeat the input tensor to a larger size along the specified dimensions.
- Parameters
x (oneflow.Tensor) – The input Tensor.
*size (flow.Size or int) – The number of times to repeat this tensor along each dimension
- Returns
The result Tensor.
- Return type
For example:
>>> import oneflow as flow >>> import numpy as np >>> x = np.array([[[[0, 1]], ... [[2, 3]], ... [[4, 5]]]]).astype(np.int32) >>> input = flow.Tensor(x) >>> out = input.repeat(1, 1, 2, 2) >>> out.shape oneflow.Size([1, 3, 2, 4])
-
property
requires_grad
¶
-
requires_grad_
(self: oneflow._oneflow_internal.Tensor, requires_grad: bool = True) → oneflow._oneflow_internal.Tensor¶
-
reshape
(*shape)¶ This operator reshapes a Tensor.
We can set one dimension in shape as -1, the operator will infer the complete shape.
- Parameters
x – A Tensor.
*shape – tuple of python::ints or int…
- Returns
A Tensor has the same type as x.
For example:
>>> import numpy as np >>> import oneflow as flow >>> x = np.array( ... [[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], [13, 14, 15, 16]] ... ).astype(np.float32) >>> input = flow.Tensor(x) >>> y = input.reshape(2, 2, 2, -1).shape >>> y oneflow.Size([2, 2, 2, 2])
-
retain_grad
(self: oneflow._oneflow_internal.Tensor) → None¶
-
roll
(shifts, dims=None)¶ See
oneflow.roll()
-
round
()¶ See
oneflow.round()
-
rsqrt
()¶ Returns a new tensor with the reciprocal of the square-root of each of the elements of
input
.\[\text{out}_{i} = \frac{1}{\sqrt{\text{input}_{i}}}\]- Parameters
input – the input tensor.
>>> import oneflow as flow >>> import numpy as np >>> a = flow.Tensor(np.array([1.0, 2.0, 3.0])) >>> out = flow.rsqrt(a).numpy() >>> out array([1. , 0.70710677, 0.57735026], dtype=float32)
-
selu
()¶ See
oneflow.selu()
-
property
shape
¶
-
sigmoid
()¶
-
sign
()¶ See
oneflow.sign()
-
silu
()¶ See
oneflow.silu()
-
sin
() → Tensor¶ See
oneflow.sin()
-
sin_
()¶ In-place version of
oneflow.sin()
-
sinh
()¶ See
oneflow.sinh()
-
size
(idx=None)¶ The interface is consistent with PyTorch.
Returns the size of the self tensor. If dim is not specified, the returned value is a torch.Size, a subclass of tuple. If dim is specified, returns an int holding the size of that dimension.
- Parameters
idx (int, optional) – The dimension for which to retrieve the size.
-
softmax
(dim=None)¶
-
softplus
()¶
-
softsign
()¶
-
sort
(dim: int = - 1, descending: bool = False)¶ Sorts the elements of the input tensor along a given dimension in ascending order by value.
- Parameters
input (oneflow.Tensor) – The input Tensor.
dim (int, optional) – dimension to be sorted. Defaults to the last dim (-1).
descending (bool, optional) – controls the sorting order (ascending or descending).
- Returns
A tuple of (values, indices), where where the values are the sorted values and the indices are the indices of the elements in the original input tensor.
- Return type
Tuple(oneflow.Tensor, oneflow.Tensor(dtype=int32))
For example:
>>> import oneflow as flow >>> import numpy as np >>> x = np.array([[1, 3, 8, 7, 2], [1, 9, 4, 3, 2]], dtype=np.float32) >>> input = flow.Tensor(x) >>> (values, indices) = flow.sort(input) >>> values tensor([[1., 2., 3., 7., 8.], [1., 2., 3., 4., 9.]], dtype=oneflow.float32) >>> indices tensor([[0, 4, 1, 3, 2], [0, 4, 3, 2, 1]], dtype=oneflow.int32) >>> (values, indices) = flow.sort(input, descending=True) >>> values tensor([[8., 7., 3., 2., 1.], [9., 4., 3., 2., 1.]], dtype=oneflow.float32) >>> indices tensor([[2, 3, 1, 4, 0], [1, 2, 3, 4, 0]], dtype=oneflow.int32) >>> (values, indices) = flow.sort(input, dim=0) >>> values tensor([[1., 3., 4., 3., 2.], [1., 9., 8., 7., 2.]], dtype=oneflow.float32) >>> indices tensor([[0, 0, 1, 1, 0], [1, 1, 0, 0, 1]], dtype=oneflow.int32)
-
split
(split_size_or_sections=None, dim=None)¶ See
oneflow.split()
-
sqrt
()¶ Returns a new tensor with the square-root of the elements of
input
.\[\text{out}_{i} = \sqrt{\text{input}_{i}}\]- Parameters
input – the input tensor.
>>> import oneflow as flow >>> import numpy as np >>> arr = np.array([1.0, 2.0, 3.0]) >>> input = flow.Tensor(arr) >>> output = flow.sqrt(input).numpy() >>> output array([1. , 1.4142135, 1.7320508], dtype=float32)
-
square
()¶ Returns a new tensor with the square of the elements of
input
.\[\text{out}_{i} = \sqrt{\text{input}_{i}}\]- Parameters
input – the input tensor.
>>> import oneflow as flow >>> import numpy as np >>> arr = np.array([1.0, 2.0, 3.0]) >>> input = flow.Tensor(arr) >>> output = flow.square(input).numpy() >>> output array([1., 4., 9.], dtype=float32)
-
squeeze
(dim=None)¶
-
std
(dim=None, unbiased=True, keepdim=False)¶ See
oneflow.std()
-
stride
(self: oneflow._oneflow_internal.Tensor) → tuple¶
-
sub
(other)¶ Computes the subtraction of input by other for each element, scalar and broadcast promotation are supported. The formula is:
\[out = input - other\]For example:
>>> import numpy as np >>> import oneflow as flow # element-wise subtract >>> input = flow.Tensor(np.random.randn(2,3)) >>> other = flow.Tensor(np.random.randn(2,3)) >>> out = flow.sub(input,other).numpy() >>> out.shape (2, 3) # scalar subtract >>> input = 5 >>> other = flow.Tensor(np.random.randn(2,3)) >>> out = flow.sub(input,other).numpy() >>> out.shape (2, 3) # broadcast subtract >>> input = flow.Tensor(np.random.randn(1,1)) >>> other = flow.Tensor(np.random.randn(2,3)) >>> out = flow.sub(input,other).numpy() >>> out.shape (2, 3)
-
tan
()¶ See
oneflow.tan()
-
tanh
()¶ See
oneflow.tanh()
-
tile
(reps)¶ The interface is consistent with PyTorch. The documentation is referenced from: https://pytorch.org/docs/stable/generated/torch.tile.html
Constructs a tensor by repeating the elements of
input
. Thereps
argument specifies the number of repetitions in each dimension.If
reps
specifies fewer dimensions thaninput
has, then ones are prepended toreps
until all dimensions are specified. For example, ifinput
has shape (8, 6, 4, 2) andreps
is (2, 2), thenreps
is treated as (1, 1, 2, 2).Analogously, if
input
has fewer dimensions thanreps
specifies, theninput
is treated as if it were unsqueezed at dimension zero until it has as many dimensions asreps
specifies. For example, ifinput
has shape (4, 2) andreps
is (3, 3, 2, 2), theninput
is treated as if it had the shape (1, 1, 4, 2).Note
This function is similar to NumPy’s tile function.
- Parameters
input (oneflow.Tensor) – the tensor whose elements to repeat.
reps (tuple) – the number of repetitions per dimension.
For example:
>>> import oneflow as flow >>> import numpy as np >>> x = np.array([1, 2]).astype(np.int32) >>> input = flow.tensor(x, dtype=flow.int32) >>> out = input.tile(reps=(2,)) >>> out tensor([1, 2, 1, 2], dtype=oneflow.int32) >>> x = np.random.randn(5, 2, 1) >>> input = flow.Tensor(x) >>> out = input.tile(reps=(3, 4)) >>> out.size() oneflow.Size([5, 6, 4])
-
to
(*args, **kwargs)¶ - Performs Tensor dtype and/or device conversion.
A flow.dtype and flow.device are inferred from the arguments of input.to(*args, **kwargs).
Note
If the
input
Tensor already has the correctflow.dtype
andflow.device
, theninput
is returned. Otherwise, the returned tensor is a copy ofinput
with the desired.- Parameters
input (oneflow.Tensor) – An input tensor.
*args (oneflow.Tensor or oneflow.device or oneflow.dtype) – Positional arguments
**kwargs (oneflow.device or oneflow.dtype) – Key-value arguments
- Returns
A Tensor.
- Return type
For example:
>>> import numpy as np >>> import oneflow as flow >>> arr = np.random.randint(1, 9, size=(1, 2, 3, 4)) >>> input = flow.Tensor(arr) >>> output = input.to(dtype=flow.float32) >>> np.array_equal(arr.astype(np.float32), output.numpy()) True
-
to_consistent
(placement=None, sbp=None, grad_sbp=None)¶ Cast a local tensor to consistent tensor or cast a consistent tensor to another consistent tensor with different sbp or placement
- Parameters
input (Tensor) – the input tensor.
placement (flow.placement, optional) – the desired placement of returned consistent tensor. Default: if None, the input tensor must be consistent one and use its own placement.
sbp (flow.sbp.sbp or tuple of flow.sbp.sbp, optional) – the desired sbp descriptor of returned consistent tensor. Default: if None, the input tensor must be consistent one and use its own sbp.
For example:
>>> import oneflow as flow >>> import numpy as np >>> np_arr = np.array([0.5, 0.6, 0.7]).astype(np.float32) >>> input = flow.Tensor(np_arr) >>> placement = flow.placement("cpu", {0:range(1)}) >>> output_tensor = input.to_consistent(placement, [flow.sbp.split(0)]) >>> output_tensor.is_consistent True
-
to_local
()¶ Returns the local tensor of a consistent tensor.
- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> import numpy as np >>> np_arr = np.array([0.5, 0.6, 0.7]).astype(np.float32) >>> input = flow.tensor(np_arr, dtype=flow.float32) >>> placement = flow.placement("cpu", {0:range(1)}) >>> consistent_tensor = input.to_consistent(placement, [flow.sbp.split(0)]) >>> consistent_tensor.to_local() tensor([0.5000, 0.6000, 0.7000], dtype=oneflow.float32)
-
tolist
()¶ Returns the tensor as a (nested) list. For scalars, a standard Python number is returned, just like with item(). Tensors are automatically moved to the CPU first if necessary.
This operation is not differentiable.
- Parameters
input (Tensor) – the input tensor.
For example:
>>> import oneflow as flow >>> input = flow.tensor([[1,2,3], [4,5,6]]) >>> input.tolist() [[1, 2, 3], [4, 5, 6]]
-
topk
(k, dim: Optional[int] = None, largest: bool = True, sorted: bool = True)¶ Finds the values and indices of the k largest entries at specified axis.
- Parameters
input (oneflow.Tensor) – Input Tensor
k (int) – the k in “top-k”
dim (int, optional) – the dimension to sort along. Defaults to the last dim (-1)
largest (bool, optional) – controls whether to return largest or smallest elements
sorted (bool, optional) – controls whether to return the elements in sorted order (Only Support True Now!)
- Returns
A tuple of (values, indices), where the indices are the indices of the elements in the original input tensor.
- Return type
Tuple(oneflow.Tensor, oneflow.Tensor(dtype=int32))
For example:
>>> import oneflow as flow >>> import numpy as np >>> x = np.array([[1, 3, 8, 7, 2], [1, 9, 4, 3, 2]], dtype=np.float32) >>> (values, indices) = flow.topk(flow.Tensor(x), k=3, dim=1) >>> values tensor([[8., 7., 3.], [9., 4., 3.]], dtype=oneflow.float32) >>> indices tensor([[2, 3, 1], [1, 2, 3]], dtype=oneflow.int64) >>> values.shape oneflow.Size([2, 3]) >>> indices.shape oneflow.Size([2, 3]) >>> (values, indices) = flow.topk(flow.Tensor(x), k=2, dim=1, largest=False) >>> values tensor([[1., 2.], [1., 2.]], dtype=oneflow.float32) >>> indices tensor([[0, 4], [0, 4]], dtype=oneflow.int64) >>> values.shape oneflow.Size([2, 2]) >>> indices.shape oneflow.Size([2, 2])
-
transpose
(dim0, dim1)¶
-
tril
(diagonal=0)¶ See
oneflow.tril()
-
triu
(diagonal=0)¶ See
oneflow.triu()
-
type_as
(target)¶ - Returns this tensor cast to the type of the given tensor.
This is a no-op if the tensor is already of the correct type.
- Parameters
For example:
>>> import oneflow as flow >>> import numpy as np >>> input = flow.tensor(np.random.randn(1, 2, 3), dtype=flow.float32) >>> target = flow.tensor(np.random.randn(4, 5, 6), dtype = flow.int32) >>> input = input.type_as(target) >>> input.dtype oneflow.int32
-
unfold
(dimension, size, step)¶ The interface is consistent with PyTorch. The documentation is referenced from: https://pytorch.org/docs/stable/generated/torch.Tensor.unfold.html#torch.Tensor.unfold.
Returns a view of the original tensor which contains all slices of size size from self tensor in the dimension dimension.
Step between two slices is given by step.
If sizedim is the size of dimension dimension for self, the size of dimension dimension in the returned tensor will be (sizedim - size) / step + 1.
An additional dimension of size size is appended in the returned tensor.
- Parameters
dimension (int) – dimension in which unfolding happens
size (int) – the size of each slice that is unfolded
step (int) – the step between each slice
For example:
>>> import numpy as np >>> import oneflow as flow >>> x = flow.arange(1., 8) >>> x tensor([ 1., 2., 3., 4., 5., 6., 7.]) >>> x.unfold(0, 2, 1) tensor([[ 1., 2.], [ 2., 3.], [ 3., 4.], [ 4., 5.], [ 5., 6.], [ 6., 7.]]) >>> x.unfold(0, 2, 2) tensor([[ 1., 2.], [ 3., 4.], [ 5., 6.]])
-
uniform_
(a=0, b=1)¶ Tensor.uniform_(from=0, to=1) → Tensor
Fills self tensor with numbers sampled from the continuous uniform distribution:
\[P(x)=1/(to-from)\]
-
unsqueeze
(dim)¶
-
var
(dim=None, unbiased=True, keepdim=False)¶ See
oneflow.var()
-
view
(*shape)¶ The interface is consistent with PyTorch. The documentation is referenced from: https://pytorch.org/docs/stable/generated/torch.Tensor.view.html
Returns a new tensor with the same data as the
self
tensor but of a differentshape
.The returned tensor shares the same data and must have the same number of elements, but may have a different size. For a tensor to be viewed, the new view size must be compatible with its original size and stride, i.e., each new view dimension must either be a subspace of an original dimension, or only span across original dimensions \(d, d+1, \dots, d+k\) that satisfy the following contiguity-like condition that \(\forall i = d, \dots, d+k-1\),
\[\text{stride}[i] = \text{stride}[i+1] \times \text{size}[i+1]\]Otherwise, it will not be possible to view
self
tensor asshape
without copying it (e.g., viacontiguous()
). When it is unclear whether aview()
can be performed, it is advisable to usereshape()
, which returns a view if the shapes are compatible, and copies (equivalent to callingcontiguous()
) otherwise.- Parameters
input – A Tensor.
*shape – flow.Size or int…
- Returns
A Tensor has the same type as input.
For example:
>>> import numpy as np >>> import oneflow as flow >>> x = np.array( ... [[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], [13, 14, 15, 16]] ... ).astype(np.float32) >>> input = flow.Tensor(x) >>> y = input.view(2, 2, 2, -1).numpy().shape >>> y (2, 2, 2, 2)
-
where
(x=None, y=None)¶ Return a tensor of elements selected from either
x
ory
, depending oncondition
. If the element in condition is larger than 0,it will take the x element, else it will take the y element
Note
If
x
is None andy
is None, flow.where(condition) is identical to flow.nonzero(condition, as_tuple=True).The tensors
condition
,x
,y
must be broadcastable.- Parameters
- Returns
A tensor of shape equal to the broadcasted shape of
condition
,x
,y
- Return type
For example:
>>> import numpy as np >>> import oneflow as flow >>> x = flow.tensor( ... np.array([[-0.4620, 0.3139], [0.3898, -0.7197], [0.0478, -0.1657]]), ... dtype=flow.float32, ... ) >>> y = flow.tensor(np.ones(shape=(3, 2)), dtype=flow.float32) >>> condition = flow.tensor(np.array([[0, 1], [1, 0], [1, 0]]), dtype=flow.int32) >>> out = condition.where(x, y) >>> out tensor([[1.0000, 0.3139], ... [0.0478, 1.0000]], dtype=oneflow.float32)
-
zeros_
(self: oneflow._oneflow_internal.Tensor) → None¶
-