oneflow.Tensor¶
OneFlow Tensor Class¶
-
class
oneflow.
Tensor
¶ -
property
T
¶ Is this Tensor with its dimensions reversed.
If n is the number of dimensions in x, x.T is equivalent to x.permute(n-1, n-2, …, 0).
-
abs
()¶ See
oneflow.abs()
-
acos
()¶ See
oneflow.acos()
-
acosh
()¶ See
oneflow.acosh()
-
add
(other)¶ See
oneflow.add()
-
add_
(other)¶ In-place version of
oneflow.Tensor.add()
.
-
addmm
(mat1, mat2, alpha=1, beta=1)¶
-
arccos
()¶ See
oneflow.arccos()
-
arccosh
()¶
-
arcsin
()¶ See
oneflow.arcsin()
-
arcsinh
()¶
-
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)
-
argwhere
()¶
-
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()
.
-
clamp_
(min=None, max=None)¶ Inplace version of
oneflow.Tensor.clamp()
.
-
clip
(min=None, max=None)¶ Alias for
oneflow.Tensor.clamp()
.
-
clip_
(min=None, max=None)¶ Alias for
oneflow.Tensor.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
()¶ See
oneflow.cos()
-
cosh
()¶
-
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[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)¶ See
oneflow.div()
-
div_
(value) → Tensor¶ In-place version of
oneflow.Tensor.div()
.
-
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)¶ See
oneflow.eq()
-
erf
() → Tensor¶ See
oneflow.erf()
-
erfc
() → Tensor¶ See
oneflow.erfc()
-
erfinv
()¶ See
oneflow.erfinv()
-
erfinv_
()¶ Inplace version of
oneflow.erfinv()
-
exp
()¶ See
oneflow.exp()
-
expand
() → Tensor¶ See
oneflow.expand()
-
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()
-
floor_
()¶ In-place version of
oneflow.floor()
-
fmod
(other) → Tensor¶ 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
-
is_contiguous
(self: oneflow._oneflow_internal.Tensor) → bool¶
-
property
is_cuda
¶
-
is_floating_point
()¶
-
property
is_global
¶
-
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)¶ See
oneflow.le()
-
log
()¶
-
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)¶ See
oneflow.lt()
-
masked_fill
(mask, fill_value)¶
-
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
(value) → Tensor¶ See
oneflow.mul()
-
mul_
(value) → Tensor¶ In-place version of
oneflow.Tensor.mul()
.
-
narrow
(dimension, start, length)¶ See
oneflow.narrow()
-
property
ndim
¶
-
ndimension
()¶ Tensor.dim() → int
Returns the number of dimensions of self tensor.
-
ne
(other)¶ See
oneflow.ne()
-
negative
()¶
-
nelement
()¶ Tensor.nelement() → int
Alias for numel()
-
nms
(scores, iou_threshold: float)¶ See
oneflow.nms()
-
norm
(p=None, dim=None, keepdim=False, dtype=None)¶
-
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
()¶
-
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
(*size) → Tensor¶ See
oneflow.repeat()
-
property
requires_grad
¶
-
requires_grad_
(self: oneflow._oneflow_internal.Tensor, requires_grad: bool = True) → oneflow._oneflow_internal.Tensor¶
-
reshape
(*shape)¶
-
retain_grad
(self: oneflow._oneflow_internal.Tensor) → None¶
-
roll
(shifts, dims=None)¶ See
oneflow.roll()
-
round
()¶ See
oneflow.round()
-
rsqrt
()¶
-
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 oneflow.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)¶ See
oneflow.sort()
-
split
(split_size_or_sections=None, dim=0)¶ See
oneflow.split()
-
sqrt
()¶
-
square
()¶
-
squeeze
(dim=None)¶
-
std
(dim=None, unbiased=True, keepdim=False)¶ See
oneflow.std()
-
stride
(self: oneflow._oneflow_internal.Tensor) → tuple¶
-
sub
(other)¶ See
oneflow.sub()
-
sub_
(value) → Tensor¶ In-place version of
oneflow.Tensor.sub()
.
-
sum
(dim, index) → Tensor¶ See
oneflow.sum()
-
swapaxes
(dim0, dim1)¶
-
t
()¶ Tensor.t() → Tensor
See
oneflow.t()
-
tan
()¶ See
oneflow.tan()
-
tanh
()¶ See
oneflow.tanh()
-
tile
(*dims) → Tensor¶ See
oneflow.tile()
-
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_local
()¶ Returns the local tensor of a global 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", ranks=[0]) >>> global_tensor = input.to_global(placement, [flow.sbp.split(0)]) >>> global_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)¶ See
oneflow.topk()
-
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)¶ See
oneflow.where()
-
zeros_
(self: oneflow._oneflow_internal.Tensor) → Maybe[void]¶
-
property